Convergence of per capita carbon emissions in the Yangtze River Economic Belt,China

Abstract

The convergence of carbon emissions in important regions is a key prerequisite for reaching China’s carbon emissions peak target. The Yangtze River Economic Belt (YREB) is the latest strategic economic area in China. In this study, a method of investigating the convergence of per capita carbon emissions in the YREB is proposed, and it includes the following models: $σ$ convergence, stochastic convergence and $β$ convergence. Besides, the five control variables per capita Gross Domestic Product (GDP), energy consumption structure, industrial structure, population density and urbanization rate are considered in the test models. In addition to testing the whole sample of 74 cities belonging to the Economic Belt, the convergence of three subgroups (municipalities, capital and sub-provincial cities and prefecture-level cities) is also tested by considering spatial effects. The following results are obtained: (1) although $σ$ convergence and $β$ conditional convergence are observed, stochastic convergence is not observed for the 74 cities; (2) the per capita GDP and energy structure are the key factors influencing the cities’ convergence, whereas the urbanization rate has a limited impact; and (3) improving the industrial structure and reducing the population density are effective methods of promoting the convergence of per capita carbon emissions among the capital and sub-provincial cities and municipalities, respectively.

Keywords

Carbon emission convergence Yangtze River Economic Belt spatial effect reduction cities

Introduction

To deal with climate change, China has made a commitment to reducing its carbon intensity by 60%–65% relative to 2005 by 2030, and achieving its peak emissions around 2030 or earlier in the Intended Nationally Determined Contributions (INDCs). Reaching convergence or a steady state of regional energy-related carbon emissions is an important precondition for China to achieve its carbon emissions reduction target and peak emissions by 2030. The convergence of regional per capita carbon emissions means the shrinking differences in emission level among different areas or a relatively stable state of carbon emissions. If regional emissions did not converge at a relatively lower stable level, but showed divergent trends, sustained growth in carbon emissions will be made, as well as the gradually increasing inequality which make difficult for China to reach its targets of carbon intensity reduction and peak in total amount.

To achieve sustainable economic development in China, an important economic development strategy for the Yangtze River Valley named the Development and Planning Outline of Yangtze River Economic Belt (YREB) was issued by China in 2016.¹ The new economic development zone will establish a new developmental pattern based on the concept of “Economical Priority and Green Development”. The YREB contains 11 provinces and municipalities and traverses the heart of China from west to east, covering an area of approximately 2.05 million km² which accounts for nearly 21% of China. In 2016, the proportion of population and GDP of this region in total is 42.8% and 45.3%, respectively.² Moreover, the estimated energy-related CO₂ emissions of the belt were 3.09 billion tons in 2016. Undoubtedly, the belt has become the most important economic region of China. Therefore, investigating the convergence of carbon emissions in the economic belt would not only help us to understand the characteristics and dynamic evolution of carbon emissions, but also contribute to the formulation of a targeted regional carbon emissions reduction policy. Such an investigation could also provide decision-making support for the achievement of China’s carbon emissions reduction targets and peak.

Therefore, the present study conducted sigma ( $σ$ ), stochastic and beta ( $β$ ) convergence tests by using the data of 74 cities (2 municipalities, 10 capital and sub-provincial cities and 62 prefecture-level cities) of YREB. The main contributions of the present study are as follows. First, the convergence of carbon emissions in the latest economic zone of China, the YREB, is investigated for the first time. Exploring the evolution of carbon emissions in this new special economic zone is of great significance for promoting green development and environmental protection for the zone. Second, the effects on the convergence of CO₂ emissions of five control variables (energy consumption structure, industrial structure, per capita GDP, population density and urbanization rate) are simultaneously discussed in our study. Previous studies have primarily focused on the industrial structure, GDP per capita and population density,^3–5 and they have rarely considered the energy structure, which is an important factor affecting the convergence of carbon emissions. Third, the influence of spatial factors was detected in the beta absolute and conditional convergence tests. Finally, the convergence of emissions according to different city levels (municipalities, capital and sub-provincial cities and prefecture-level cities) is further investigated to formulate refined emission reduction policies.

Literature review

Literature on convergence of CO₂ emissions

The concept of convergence hypothesis, which indicates a steady growth in the long run, initially appeared in Solow⁶ growth model of new classical economics. In addition to clarifying giant differences of economic growth^7–10 or income performance¹¹ across different economies, carbon emissions convergence, as the great attention from world on climate change and energy security, has attracted eyes of policy makers and academic researchers.

Intensive studies have explored the convergence of per capita carbon emissions or carbon intensity among countries,^12–15 regions^10,16–19 and sectors.^20–22

Above literatures supported the existence of carbon emissions convergence, whereas some scholars concluded that there is no convergent trend based on different objects, samples or methods. For example, Barassi et al.²³ applied the panel unit root tests and found no existence of per capita CO₂ emissions convergence across Organization for Economic Co-operation and Development (OECD) countries in consideration of plausible nonlinearities during the sample period from 1950 to 2002. Camarero et al.²⁴ also came out the similar conclusion that no robust convergence of carbon intensity existed in 22 OECD countries during 1870–2006.

In most studies, the carbon convergence in subsample has been confirmed while that in full sample not been found. Lee and Chang²⁵ examined per capita carbon emissions of 21 OECD members with seemingly unrelated panel regressions Augmented Dickey-Fuller(ADF) unit root tests, and the result showed that convergence only appeared among seven countries. In order to study per capita CO2 emissions of 128 countries globally, Panopoulou and Pantelidis²⁶ utilized the club convergence technique proposed by Phillips and Sul²⁷ and found two convergence clubs in later years of sample period. By using the same club convergence method and conditional convergence test in the consideration of structural factors, Burnett²⁸ examined the convergence of per capita CO₂ emissions among a panel of 48 U.S. states and observed one converging club consisting of 26 states, what is more, he also found that the convergence rate of the convergent group was higher than that of whole sample. By employing the TAR (threshold autoregressive) panel unit root approach and panel data covered G7 countries from 1960 to 2005, Yavuz and Yilanci²⁹ found that conditional convergence existed in the first regime and divergence existed in the second regime. Moutinho et al.²¹ confirmed that the carbon intensities of Portuguese industry and energy sectors from 1996 to 2009 were converged towards two clubs. Besides, based on an environmental performance index method, Yu et al.²² found that three converging clubs across the carbon intensities of 24 industrial sectors in China between 1995 and 2015. In addition, Aldy,³⁰ Barassi et al.,³¹ Apergis and Payne³² and etc. reported the similar subsample convergence results.

However, the conclusions might also be inconsistent in different types of convergence tests. For instance, Presno et al.³³ investigated the per capita carbon emission convergence of 28 OECD members during the period 1991–2009, and the result implied the existence of stochastic convergence. However, the β convergence test showed the tendency of divergence in full sample and developed countries. Li et al.⁴ discovered that stochastic convergence and β convergence of carbon intensity existed among Yangtze River Delta cities. And the result of σ convergence indicated divergence between 2002 and 2004.

Through our limited review, we find that the previous studies about carbon convergence exist mainly across countries, provinces or states. However, less attention has been focused on the convergence of carbon emissions among cities, especially cities belonging to a specific economic zone. Specific economic areas may be geographically adjacent, economically interdependent and present strong radiation effects on policies. In the latest regional development strategy proposed by China, the YREB region will break administrative jurisdiction boundaries of 11 provinces and municipalities for the first time, explore new green development pattern and achieve the integration of relevant policies. Therefore, exploring the convergence of carbon emissions in the Economic Belt could help reveal the evolution of carbon emissions in the region, identify key factors that affect the steady-state changes in carbon emissions and provide policy-making support for the development of targeted emission reduction policies in the zone.

Literatures on green development model of the Yangtze River Economic Belt

The Chinese government decided to implement a development model of ecological priority and green development for the YREB. Thus, lots of studies focused on investigating how to conduct the green development pattern for the region. For instance, Wu and Huang³⁴ constructed an index system of industrial green development and made an objective evaluation on industrial green development level of Yangtze River Belt by Entropy-TOPSIS model. Similarly, taking Yangtze River Belt zone as the research object, Wang et al.³⁵ and Jiang³⁶ established eco-environmental health and environmental justice assessment system, respectively. Chen et al.³⁷ investigated the environmental efficiency of 131 cities of YREB by employing the super efficiency DEA method and panel Tobit model, and the results suggested that the degree of opening up and industrial structure have positive influence on environmental efficiency while GDP per capita has opposite impact. Employing three factors transcendental production function, Wu and Du³⁸ calculated total factor energy efficiency of 11 provinces and municipalities along the Yangtze River and proved that biased technology change is conductive to total factor energy efficiency promotion. By using DEA, super efficiency SBM-DEA and SBM model respectively, Wang et al.,³⁹ Zhao et al.⁴⁰ and Xing et al.⁴¹ investigated total factor industrial ecological efficiency or energy ecological efficiency. Taking the YREB region as object, some scholars studied the influencing factors of carbon emissions,^42–44 urban comprehensive carrying capacity of YREB area^45–47 and water transfer network⁴⁸ as well. However, according to our limited knowledge, there is no research examining per capita CO₂ convergence characteristic at city-level of YREB area.

Literatures on spatial effects of carbon emissions

In recent empirical researches on regional science and economics, spatial factors are receiving widespread attention as the unobservable heterogeneity and mutual effects of one region with its neighbors cannot be ignored.^49–54

Focusing on regional carbon emissions, many researches fully considered the influence of spatial dependence and spillover effects. Meng et al.⁵⁵ decomposed carbon emission growth from a spatial perspective, by applying a spatial structural decomposition analysis method based on the 2007 and 2010 China’s inter-regional input-output tables. On the basis of STIRPAT framework and Sys-GMM method with the consideration of spatial characteristics, Su et al.⁵⁶ identified the influential factors of energy-related carbon emissions of Chinese cities from 1992 to 2013. Similarly, Chuai et al.,⁵⁷ Zhang et al.⁵⁸ and Zhang and Zhao⁵⁹ also analyzed the impact factors of carbon emissions by spatial panel regressions and geographical detector method. To investigate the existence of Environmental Kuznets Curve (EKC), Wang and Ye⁶⁰ found the monotonously increasing relationship between economic development and carbon emissions at Chinese city level by utilizing spatial lag model and spatial error model. By using the spatial Durbin model, Meng and Huang⁶¹ also drew the same conclusion. Marbuah and Amuakwa-Mensah⁵³ verified EKC hypothesis and spatial effects of seven types of air emissions in 290 Swedish municipalities. The obvious neighborhood effects as well as significant economic spillovers within and inter-municipality were found. Besides, Tang et al.,⁶² Li et al.⁶³ and Chen et al.⁶⁴ also confirmed that regional carbon emissions exhibit significant spatial interaction.

According to the reviews of the section Literature on convergence of CO₂ emissions, we found that most literatures, which examined carbon emission convergence across regions, ignored the spatial heterogeneity and the dependence of the space. Only a few studies added the spatial effect into the regression model. For example, Huang and Meng⁶⁵ proposed a spatial-temporal model and discovered that per capita CO₂ emissions in urban China converged from 1985 to 2008. Zhao et al.¹⁹ examined carbon intensity of 30 provinces in China from 1990 to 2010 by using a spatial dynamic panel data model. Li et al.⁴ considered spatial spillover effect when they investigated the carbon intensity convergence of Yangtze River Delta cities. Therefore, in present study, spatial effects are also taken into consideration in the beta convergence test.

Methodology

In this empirical research, $σ$ convergence, stochastic convergence and $β$ convergence models are proposed to explore the convergence of the per capita carbon emissions of cities in the YREB. In addition, spatial effects are considered in the $β$ convergence test.

$σ$ convergence

If the dispersion degree of per capita carbon emissions in the YREB cities decreases with time, the $σ$ convergence of per capita carbon emissions is implied among these cities. $σ$ convergence can be tested by the deviation, the standard deviation (SD), the coefficient of variation (CV), etc.⁶⁶ The CV is the most popular method of measuring the $σ$ convergence, which can be expressed as follows

C V_{t} = \frac{1}{\bar{P C E_{t}}} \sqrt{\frac{1}{n} \sum_{i = 1}^{n} (P C E_{i, t} - {\bar{P C E_{t}}}^{2})}

(1)

where

i

and t represent the city and year, respectively;

C V

is the coefficient of variation of per capita carbon emissions;

n

indicates the number of research samples in YREB; and

P C E_{i, t}

and

\bar{P C E_{t}}

are the per capita carbon emissions of city

i

in year

t

and average per capita carbon emissions of all cities in year

t

, respectively.

Stochastic convergence

Stochastic convergence requires that the gap of per capita carbon emissions among various cities should be close to zero, which is stricter than that of $σ$ convergence.⁶⁷ The unit root test is generally adopted to examine whether stochastic convergence has occurred. If the time series of per capita carbon emissions in different cities of the YREB is stationary, stochastic convergence has occurred. Namely, the shock on per capita carbon emissions relative to the average level is temporary and will diminish over time. The unit root test methods of the Harris and Tzavlis (HT) test and Im, Pesaran, and Shin (IPS) test for short panels proposed by Harris and Tzavalis⁶⁸ and Im et al.,⁶⁹ respectively, are applied in this paper. The HT test requires that each city has the same auto-regression coefficient (common root), whereas the IPS test does not impose this requirement.⁴

$β$ convergence

Although the $σ$ convergence and stochastic convergence models are easy to run and readily comprehensible, they do not include the initial conditions and structural characteristics of cities. As a result, too much information is ignored, and little decision support can be provided to decision makers.⁷⁰ $β$ convergence represents one of the most important methods for testing convergence, and it can compensate for the shortages of $σ$ convergence and stochastic convergence and offer more information in the results. If $β$ convergence occurs, the per capita carbon emissions of cities with higher initial per capita carbon emissions would grow more slowly than those of cities with lower initial per capita carbon emissions.

$β$ convergence includes absolute $β$ convergence and conditional $β$ convergence. On the basis of the absolute $β$ convergence model proposed by Barro,⁷ this study establishes an absolute convergence model with a lag period of one year as shown in equation (2)

\ln (P C E_{i, t} / P C E_{i, t - 1}) = α + β \ln P C E_{i, t - 1} + ε_{i, t}

(2)

where

P C E_{i, t - 1}

is per capita carbon emissions of city

i

in period

t - 1

\ln (P C E_{i, t} / P C E_{i, t - 1})

is the relative growth rate of per capita carbon emissions of city

i

in period

t

α

and

β

represent the intercept term and the estimated coefficient of

\ln P C E_{i, t - 1}

, respectively, and

ε_{i, t}

is the stochastic error term.

Control variables have important impacts on the conditional $β$ convergence. Different selections and combinations of variables will result in different estimated results.¹⁶ Therefore, considering the control variables can help identify the key factors that affect the convergence of per capita carbon emissions. Among the factors affecting carbon emissions, GDP is widely discussed and has a positive relationship with carbon emissions.⁷¹ The proportion of carbon-intensive fuel (especially coal) in primary energy consumption directly affects the amount of carbon emissions under the same energy consumption. Industrial sectors are the main sources of carbon dioxide emissions in China.⁷² Therefore, a decline of the share of GDP of secondary industries contributes to a decrease in per capita carbon emissions.¹² Population density, reflects the carbon emissions associated with residential energy consumption to a large extent.^3–5,73 Urbanization rate is an important indicator for measuring the development of a city⁷⁴ and has been frequently used by previous studies as an important factor for exploring the driving forces of carbon emissions.⁷³ Therefore, in our study, per capita GDP, the energy consumption structure, industrial structure, population density and urbanization rate are considered as control variables in the regression model of conditional $β$ convergence to explore their effects on the convergence of per capita carbon emissions.

By considering the above five control variables, the convergence model equation (2) can be revised as equations (3) and (4).

\ln (P C E_{i, t} / P C E_{i, t - 1}) = α_{i} + β \ln P C E_{i, t - 1} + γ_{c} x_{c, i, t} + ε_{i, t}

(3)

γ_{c} x_{c, i, t} = γ_{1} \ln P G D P_{i, t} + γ_{2} \ln E S_{i, t} + γ_{3} \ln I S_{i, t} + γ_{4} \ln P D_{i, t} + γ_{5} \ln U R_{i, t}

(4)

where

x_{c, i, t}

and

γ_{c}

represent control variables and their coefficients (

c = 1, 2…, 5

), respectively. For city

i

in year

t

P G D P_{i, t}

is the per capita GDP,

E S_{i, t}

indicates the energy consumption structure (the proportion of coal consumption accounting for total energy consumption),

I S_{i, t}

refers to the industrial structure (the proportion of added value of secondary industries accounting for the total GDP),

P D_{i, t}

is the population density,

U R_{i, t}

is the urbanization rate and

ε_{i, t}

represents the stochastic error term. The other variables have the same meaning as in equation (2).

For equations (2) and (3), if $β$ is negative and significant simultaneously, then a $β$ convergence of per capita carbon emissions exists across YREB cities. In other words, the growth rate of per capita carbon emissions in cities with comparatively high initial per capita carbon emissions is lower than that in cities with comparatively low initial per capita carbon emissions. In addition, for all cities, if $α_{i} = α$ and $γ_{c} = 0$ are significant, $β$ absolute convergence occurs, which means that the differences of per capita carbon emissions across sample cities decrease with time and eventually converge to the same equilibrium state. If not, $β$ conditional convergence occurs, which means that the carbon emissions growth paths are different and do not necessarily converge to the same equilibrium state.

Regional carbon emissions show significant spatial characteristics.⁶³ The ignorance of spatial effects may result in giant estimation bias and mistaken conclusion. It assumes that per capita carbon emission of YREB cities might show obvious spatial correlation and spillover effect. Therefore, spatial effects are further considered. Equations (5) and (6) can be obtained.

\ln (P C E_{i, t} / P C E_{i, t - 1}) = α_{i} + \sum_{j = 1}^{N} ρ w_{i j} \ln (P C E_{i, t} / P C E_{i, t - 1}) + β \ln P C E_{i, t - 1} + \sum_{j = 1}^{N} δ_{i} w_{i j} \ln P C E_{i, t - 1} + u_{i t}

(5)

\begin{array}{l} \ln (P C E_{i, t} / P C E_{i, t - 1}) = α_{i} + \sum_{j = 1}^{N} ρ w_{i j} \ln (P C E_{i, t} / P C E_{i, t - 1}) + β \ln P C E_{i, t - 1} + γ_{c} x_{c, i, t} \\ + \sum_{j = 1}^{N} δ_{i} w_{i j} (\ln P C E_{i, t - 1} + x_{c, i, t}) + u_{i t} \end{array}

(6)

where

u_{i t} = λ \sum_{j = 1}^{N} w_{i j} u_{i t} + ε_{i t}

(7)

If $ρ \neq 0$ and $λ = δ = 0$ are significant, the spatial lag model (SLM) applies. If $λ \neq 0$ and $ρ = δ = 0$ , the spatial error model (SEM) applies. If $ρ \neq 0$ , $λ \neq 0$ and $δ = 0$ , the spatial autocorrelation model (SAC) applies. Finally, if $ρ \neq 0$ and $δ \neq 0$ are significant, a spatial Durbin model (SDM) applies. As indicated, the SLM is a special case of the SDM. W_ij is the spatial unit of spatial weight matrix, representing the spatial correlation between city i and city j.

Data management

Cities in the Yangtze River Economic Belt

The YREB area involves 11 provincial regions that include 112 cities with different levels, including 2 municipalities, 10 capital and sub-provincial cities and 100 prefecture-level cities (excluding autonomous prefectures). Due to the limited availability of data, only 74 cities in the YREB are selected as research samples in this paper as shown in Figure 1. According to regulations on the division of administrative regions, the 74 cities are divided into three types as shown in online Supplementary Table A1: 2 municipalities, 10 capital and sub-provincial cities and 62 prefecture-level cities.

Figure 1.

Studied cities of the YREB.

Carbon emissions estimation

The present study estimates carbon emissions based on the fossil energy consumption data of each city because of the lack of statistics on carbon emissions in various cities. Except for Shanghai and Chongqing, whose total energy consumption by fuel type can be acquired from their energy balance sheet, the other 72 cities only have physical energy consumption data, which are classified by the energy type for industrial enterprises above a designated size in their statistical yearbooks. To adopt a reasonable and unified method of calculating carbon emissions, provincial energy consumption structures were selected to replace the energy structures of those prefecture-level cities and sub-provincial cities. The carbon dioxide emissions ( $T C_{_{i t}}$ ) for city $i$ in year $t$ can be estimated according to equation (8).

T C_{_{i t}} = E C_{i t} \cdot C F_{i t}, i \in I

(8)

where

I

represents the 11 provincial regions,

j

represents the type of energy,

E C_{i t}

represents the total energy consumption measured as ton of coal equivalent (tce) for city

i

in year

t

, and

C F_{i t}

represents the integrated CO₂ emissions factor per unit energy for city

i

in year

t

C F_{i t}

is expressed as follows

C F_{i t} = \sum_{j} C_{j} \cdot O_{j} \cdot (\frac{E_{I j t} \cdot F_{j}}{\sum_{j} E_{I j t} \cdot F_{j}})

(9)

where

C_{j}

and

O_{j}

represent the carbon emissions factor and the oxidation rate of fuel

j

, respectively;

E_{I j t}

represents the

j

th fuel consumption (in physical) of

I

province in year

t

, for which non-energy use has been removed to avoid a higher estimation bias; and

F_{j}

is the conversion coefficient of the

j

th fuel from the physical quantity to the standard quantity (tce).

E_{I j t}

is obtained from energy balance sheet in the Energy Statistical Yearbook (2006–2016), and

F_{j}

C_{j}

and

O_{j}

reference the China Statistical Yearbook, IPCC Guidelines for National Greenhouse Gas Inventories and Yu et al.,⁷⁵ respectively. Certain cities lack the energy consumption data for the whole society; thus, the “same output value method” is adopted to estimate the data. The core of this method is the assumption that the energy consumption per unit of industrial output value of a city equals to the energy consumption per unit of industrial output value of the province where the city locates, and then the total energy consumption of the province is shared by its subordinated cities according to the proportion of industrial output value of the city.

Energy consumption structure

The energy consumption structure is an important control variable related to carbon emissions. Consisting with most studies, such as Wang et al.^73,76 etc., this research also uses the percentage of coal to represent the energy consumption structure. As mentioned in the previous section, most cities lack energy data by fuel type; hence, the provincial energy structure represents the energy structure of the city belonging to the province. As a result, the related data are obtained from the Energy Statistical Yearbook (2006–2016).

Other variables

In addition to the energy consumption structure discussed above, the utilized control variables include the industrial structure, per capita GDP (in 2010 constant price) and population density. The industrial structure is represented by the proportion of secondary industries in GDP. The population density data are obtained from the China City Statistical Yearbook (2006–2016). The GDP is converted into 2010 constant price according to the GDP index. The urbanization rate is represented by the percentage of the urban population to the total population of one city, and the data are obtained from statistical yearbooks of the province that the cities belong.

Results and discussion

Test of $σ$ convergence

In this study, $σ$ convergence is tested by the $C V_{t}$ , whose variation trend is shown in Figure 2. During the sample period, the $C V_{t}$ of per capita carbon emissions exhibits an overall downward tendency, with an average annual decrease of 0.58%, a maximum of 0.766 and a minimum of 0.708. These results imply that the $σ$ convergence of per capita carbon emissions occurs across the 74 cities in the YREB.

Figure 2.

Coefficient of variation of the per capita carbon emissions.

Test of stochastic convergence

As shown in Table 1, the $Z$ statistic is −0.223 and the corresponding $p$ value is 0.412, which is larger than 0.1 in the HT test. Therefore, the null hypothesis that panel unit root exists cannot be rejected. In the IPS test, the $\bar{t}$ statistic is −1.932 which is higher than the critical value (−2.300) at the 10% significant level, which indicates that the panel unit root exists. That is, the per capita carbon emissions in the YREB are not a stable series and the shock to the average is not temporary. Hence, stochastic convergence does not occur, and the shock of per capita carbon emissions will not diminish over time.

Table 1.

The result of stochastic convergence test.

			Critical values
	Statistic	p value	1%	5%	10%
HT test	−0.2234	0.4116
IPS test	−1.9318		−2.420	−2.340	−2.300

Note: Both time trend and intercept are included. In HT test, small-sample adjustment to T is applied. In IPS test, the $\bar{t}$ statistic and exact critical values that assume both N and T are fixed.

HT: Harris Tzavlis; IPS: Im, Pesaran, and Shin.

Test of $β$ convergence

Spatial model specification

Based on the geographical weight matrix, the global Moran’s I value of the whole panel is 0.2928 and the corresponding $p$ value is 0.000. Moreover, the carbon emissions growth rate shows a significantly positive spatial correlation. To determine the specific form of the spatial model, the Lagrange Multiplier (LM) test and the robust LM test are performed in advance and judged according to the rules proposed by Anselin.^a ^,49 The results are shown in Table 2, which shows that the LM-lag, LM-error, Robust LM-lag and Robust LM-error of all samples are significant at the 1% level. These results imply that the non-spatial effect model should be rejected. However, as shown in Table 2, the significant degrees of these statistics are the same. As a result, the specific form cannot be determined directly by the LM test. Therefore, a comprehensive comparison must be performed, and the results of the LR test and the goodness of fit ( $R^{2}$ ) of each model must be combined.

Table 2.

Results of preliminary tests.

Moran’s I	LM-lag	LM-error	Robust LM-lag	Robust LM-error
0.2928***	945.3682***	849.4354***	3114.7486***	3018.8159***

***

Null hypothesis is rejected at 1% significance level.

Table 3 presents the regression results of the absolute convergence tests for all 74 cities by using the following different methods: Ordinary Least Square (OLS) (non-spatial effects model), SLM, SEM, SAC and SDM. The selection of fixed effects or random effects is determined by the results of the Hausman test.^b As shown in Table 3, the estimated coefficient of $W \cdot \ln P C E_{t - 1}$ is significant at the 1% significance level, and the results of the Likelhood Ratio (LR) test reject simplifying the SDM to the SLM or SEM. Therefore, the SDM with fixed effects is applied when conducting the test of absolute convergence on the whole sample.

Table 3.

Results of absolute convergence in full sample.

	OLS	SLM	SEM	SAC	SDM
$\ln P C E_{t - 1}$	−0.149*** (0.012)	−0.0840*** (0.012)	−0.187*** (0.024)	−0.172*** (0.025)	−0.197*** (0.027)
Constant	0.307*** (0.021)
$W \cdot \ln P C E_{t - 1}$					0.705*** (0.145)
$ρ$		2.928*** (0.198)		1.249** (0.580)	3.215*** (0.133)
$λ$			3.249*** (0.125)	2.974*** (0.180)
Hausman test	127.81***	48.66***	96.07***		92.75***
LR test		351.353***	352.304***	185.833***	151.039***
$R^{2}$	0.1699	0.1148	0.1699	0.1568	0.1739
$N$	740

Note: Robust standard errors are in brackets. The null hypotheses of the SLM, SEM, SAC, and SDM are $ρ = 0$ , $λ = 0$ , $ρ + λ = 0$ , and $δ = 0$ , respectively. All models are maximum likelihood estimations except the non-spatial effect model.

LR: likelhood ratio; OLS: ordinary least square; SLM: spatial lag model; SEM: spatial error model; SAC: spatial autocorrelation model; SDM: spatial Durbin model.

*p < 0.1; **p < 0.05; ***p < 0.01.

Although the cities in the YREB are located along the Yangtze River, the resource endowments, economic development level, technical level and population scale are quite different for various level cities, leading to different statuses and trends of the carbon emissions. To investigate the convergence of per capita carbon emissions among the different city levels, this study divided the 74 cities belonging to the YREB into three groups according to the administrative divisions of China. The first group is prefecture-level cities and includes 62 cities; the second group is capital and sub-provincial cities and includes 10 cities; and the last group is municipalities and includes two cities (Shanghai and Chongqing). Following the same procedures in the last paragraph, the most appropriate models for testing the absolute convergence of the first, second and third subgroups are the SDM with fixed effects, the SEM with random effects and the OLS with fixed effects, respectively. In addition, the SAC with fixed effects, the SAC with fixed effects, the SDM with random effects and the OLS with random effects are chosen for testing conditional convergence of the entire sample, 62 prefecture-level cities, 10 capital and sub-provincial cities and 2 municipalities, respectively. Tables 4 to 7 report the econometric results of the absolute convergence and conditional convergence for the whole sample and three subgroups. The results of the intermediate tests for model selection are presented in online Supplementary Tables A2 to A6.

Table 4.

Regression results for the convergence in the full sample.

	Model I	Model II	Model III	Model IV	Model V	Model VI	Model VII
Method	SDM(FE)	SAC(FE)
$\ln P C E_{t - 1}$	−0.197*** (0.027)	−0.389*** (0.051)	−0.221*** (0.037)	−0.356*** (0.055)	−0.387*** (0.051)	−0.390*** (0.051)	−0.388*** (0.052)
$\ln P G D P$		0.235*** (0.037)		0.175*** (0.041)	0.243*** (0.038)	0.245*** (0.040)	0.239*** (0.035)
$\ln E S$		0.250*** (0.062)	0.114** (0.050)		0.252*** (0.062)	0.257*** (0.064)	0.249*** (0.062)
$\ln I S$		0.0664* (0.040)	0.122*** (0.046)	0.067 (0.044)		0.0653* (0.039)	0.0643 (0.040)
$\ln P D$		0.130 (0.133)	0.315* (0.165)	0.179 (0.145)	0.127 (0.130)		0.129 (0.131)
$\ln U R$		0.0172 (0.050)	0.146*** (0.048)	−0.00657 (0.049)	0.00215 (0.050)	0.0154 (0.051)
$W \cdot \ln P C E_{t - 1}$	0.705*** (0.145)
$ρ$	3.215*** (0.133)	2.568*** (0.291)	1.262** (0.584)	2.376*** (0.352)	2.580*** (0.290)	2.601*** (0.295)	2.569*** (0.291)
$λ$		1.319*** (0.413)	2.886*** (0.216)	1.847*** (0.389)	1.368*** (0.400)	1.326*** (0.424)	1.320*** (0.410)
sigma2_e	0.00329*** (0.0004)	0.00306*** (0.0003)	0.00344*** (0.0004)	0.00325*** (0.00034)	0.00307*** (0.00031)	0.00307*** (0.00032)	0.00306*** (0.0003)
$R^{2}$	0.1739	0.4048	0.2194	0.3255	0.3984	0.3975	0.4045

Note: Robust standard errors are in brackets.

SAC: spatial autocorrelation model; SDM: spatial Durbin model.

*p < 0.1; **p < 0.05; ***p < 0.01.

Absolute convergence and conditional convergence results

In Tables 4 to 7, Model I is an absolute convergence model and Model II is the conditional convergence model that considers all control variables (per capita GDP, energy structure, industrial structure, population density and urbanization rate). In addition, models III–VII are regression equations obtained after separately removing the per capita GDP, energy structure, industrial structure, population density and urbanization rate, and they are used to compare changes in the convergence state caused by a specific variable. Based on the model regression results, we obtain the following findings.

$β$ conditional convergence of per capita carbon emissions occurs in 74 cities in the YREB. As shown in Table 4, for the whole sample, in the absolute convergence model (Model I) and the conditional convergence model (Model II), the estimation coefficient of (first-order) lagged per capita carbon emissions is negative and statistically significant. For different subgroups, the $β$ values are all negative, although they have different significance. As shown in Model I of Tables 5 to 7, the lagged per capita carbon emissions is significant at the 1% level for the 62 prefecture-level cities, insignificant at the 10% level for the 10 capital and sub-provincial cities, and significant at the 10% level for the 2 municipalities. After adding all control variables (Model II) into each group, the $β$ values are all significantly negative at the 1% level. Regardless of whether the full sample or different city-level samples are used, the per capita carbon emissions all show a tendency of conditional convergence. That is, cities with higher initial per capita carbon emissions have slower per capita carbon emissions growth. In addition, the convergence states are influenced by certain factors such as the economic development and the energy consumption structure. Hence, the per capita carbon emissions of 74 cities of the YREB will eventually tend to different steady-state levels.

Table 5.

Regression results for the convergence in the 62 prefecture-level cities.

	Model I	Model II	Model III	Model IV	Model V	Model VI	Model VII
Method	SDM(FE)	SAC(FE)
$\ln P C E_{t - 1}$	−0.201*** (0.030)	−0.413*** (0.058)	−0.225*** (0.038)	−0.361*** (0.065)	−0.414*** (0.058)	−0.415*** (0.058)	−0.412*** (0.059)
$\ln P G D P$		0.255*** (0.044)		0.172*** (0.050)	0.265*** (0.042)	0.263*** (0.048)	0.262*** (0.039)
$\ln E S$		0.300*** (0.075)	0.141** (0.060)		0.303*** (0.075)	0.311*** (0.079)	0.298*** (0.075)
$\ln I S$		0.0506 (0.044)	0.146*** (0.046)	0.0591 (0.047)		0.0542 (0.043)	0.0453 (0.042)
$\ln P D$		0.131 (0.153)	0.279 (0.190)	0.210 (0.163)	0.137 (0.148)		0.132 (0.148)
$\ln U R$		0.0355 (0.066)	0.220*** (0.064)	0.00897 (0.068)	0.018 (0.062)	0.0371 (0.068)
$W \cdot \ln P C E_{t - 1}$	0.782*** (0.182)
$ρ$	3.306*** (0.157)	2.639*** (0.334)	1.358** (0.654)	2.448*** (0.370)	2.672*** (0.318)	2.647*** (0.339)	2.637*** (0.332)
$λ$		1.204** (0.515)	2.826*** (0.290)	1.711*** (0.454)	1.183** (0.483)	1.272** (0.511)	1.210** (0.506)
sigma2_e	0.00338*** (0.0005)	0.00306*** (0.0003)	0.00349*** (0.0004)	0.00333*** (0.0004)	0.00307*** (0.0004)	0.00308*** (0.0004)	0.00307*** (0.0004)
$R^{2}$	0.1673	0.3988	0.2359	0.3206	0.3994	0.3902	0.3985

Note: Robust standard errors are in brackets.

SAC: spatial autocorrelation model; SDM: spatial Durbin model.

*p < 0.1; **p < 0.05; ***p < 0.01.

The per capita GDP, energy structure and industrial structure all have significant positive impacts on the per capita carbon emissions, whereas the population density and urbanization do not have significant effects. In the conditional convergence model (Model II) of the whole sample (Table 4), a 10% increase of the per capita GDP, energy structure and industrial structure will cause the relative growth rate of the per capita carbon emissions to increase by 2.35%, 2.50% and 0.67%, respectively. The positive relationship between the per capita GDP, energy structure, industrial structure and per capita carbon emissions is consistent with the research results of Xu et al.⁷² Thus, cities with higher economic development levels will present a larger proportion of coal use with respect to the total energy consumption. In addition, larger shares of secondary industries in the total GDP correspond to a faster growth of carbon emissions.

A significant relationship is not observed between the population density and the growth rate of per capita carbon emissions, which may be related to the lack of significance of the share of household consumption in the total energy of YREB cities. Sharma,⁷⁷ Sheng and Guo⁷⁸ and others found a significant correlation between urbanization and carbon emissions. However, this study finds that the positive coefficient of urbanization is insignificant, which is consistent with the conclusions of Sadorsky⁷⁹ and Rafiq et al.⁸⁰ for emerging economies. This finding may be related to the balanced consequence between the positive and negative effects of urbanization on carbon emissions. In the full sample, the SAC model with fixed effects is selected, which concurrently takes into account the spatial lag effect and the spatial disturbance effect of the error term. The spatial autoregressive coefficients ( $ρ$ and $λ$ ) are significantly positive at the 1% level, which indicates that spatial correlation and spatial disturbance effects occur in the model; thus, the traditional panel model estimation is biased. Moreover, these results also indicate that the growth of per capita carbon emissions of one city will positively influence those of the other cities in the YREB. Furthermore, a closer distance between cities corresponds to a stronger effect.

In the conditional convergence model (Model II) of the 62 prefecture-level cities, a 10% increase in the per capita GDP and energy structure will lead to a 2.55% and 3.00% growth of per capita carbon emissions, respectively, as shown in Table 5. Similarly, to the overall sample, the population density and urbanization rate are not significant in the prefecture-level samples. Although the relationship between the industrial structure and carbon emissions has been widely recognized,^72,81 in the sample of prefecture-level cities, the effect of the industrial structure on the growth rate of per capita carbon emissions is not significant. This result is mainly because China has made great efforts to implement energy savings and emission reduction policies over the past 10 years and the industrial carbon intensity (carbon emissions per added value) has declined significantly.^82,83 In addition, the share of value added by secondary industries in the total GDP has slowly declined in most cities. Therefore, the effect of the industrial structure on the growth of per capita carbon emissions becomes insignificant. In the sample of prefecture-level cities, the SAC with fixed effects was also used. The results show that $ρ$ and $λ$ are significant at 1% and 5%, respectively, which confirm the existence of spatial effects.

As shown in Table 6, in Model II for the 10 provincial capital and sub-capital cities, a 10% increase of the per capita GDP and energy structure will lead to a per capita carbon emissions growth of 4.98% and 4.13%, respectively. The SDM with random effects is applied in this group. Moreover, spatial lag effects are not only in dependent variable (per capita carbon emission growth rate) but also in the independent variable ( $\ln P C E_{t - 1}$ ). The estimated coefficient of W $\cdot$ lnPCE _t _-1 is −13.49 and the corresponding $p$ value is 0.004. In other words, the urban lagged per capita carbon emissions have a negative impact on the growth rate of per capita carbon emissions in other cities. Higher per capita carbon emissions in the last period will lead to lower per capita carbon emission growth rates in other cities. Similarly, closer distances correspond to stronger impacts.

Table 6.

Regression results for the convergence in 10 capital and sub-provincial cities.

	Model I	Model II	Model III	Model IV	Model V	Model VI	Model VII
Method	SDM(FE)	SAC(FE)
$\ln P C E_{t - 1}$	−0.275 (0.188)	−0.628*** (0.235)	−0.153 (0.239)	−0.613*** (0.237)	−0.572** (0.259)	−0.627** (0.245)	−0.622*** (0.229)
$\ln P G D P$		0.498*** (0.141)		0.388*** (0.134)	0.439*** (0.139)	0.498*** (0.144)	0.495*** (0.145)
$\ln E S$		0.413*** (0.099)	0.0625 (0.054)		0.394*** (0.100)	0.421*** (0.186)	0.409*** (0.111)
$\ln I S$		0.214 (0.288)	−0.220 (0.310)	0.140 (0.298)		0.221 (0.300)	0.205 (0.293)
$\ln P D$		−0.048 (0.139)	0.030 (0.072)	−0.101 (0.163)	−0.050 (0.147)		−0.0476 (0.136)
$\ln U R$		0.020 (0.120)	−0.0887 (0.057)	−0.0564 (0.120)	−0.0206 (0.134)	0.020 (0.127)
Constant	0.706* (0.427)	−1.150 (1.394)	1.169 (2.137)	1.454 (1.366)	−0.207 (1.432)	−1.512 (0.987)	−1.047 (1.649)
$W \cdot \ln P C E_{t - 1}$		−13.49*** (4.746)	−0.256 (4.581)	−8.406 (5.338)	−10.70*** (4.134)	−13.82*** (4.744)	−13.13*** (4.261)
$ρ$		12.73** (6.158)	11.41 (0.778)	12.44** (5.621)	13.77** (5.940)	12.33* (6.579)	12.85** (6.268)
$λ$	17.48*** (5.286)
lgt_theta		−1.568** (0.768)	0.473 (2.899)	−1.618** (0.643)	−1.325 (0.893)	−1.576* (0.858)	−1.534** (0.715)
sigma2_e	0.00784*** (0.0018)	0.0048*** (0.0012)	0.0089*** (0.0018)	0.0054*** (0.0011)	0.0051*** (0.0014)	0.0048*** (0.0013)	0.0049*** (0.0013)
$R^{2}$	0.3800	0.6385	0.3181	0.5804	0.6135	0.6383	0.6362

Note: Robust standard errors are in brackets.

SEM: spatial error model; SDM: spatial Durbin model.

*p < 0.1; **p < 0.05; ***p < 0.01.

Table 7.

Regression results for the convergence in the two municipalities.

	Model I	Model II	Model III	Model IV	Model V	Model VI	Model VII
	OLS(FE)	OLS(RE)
$\ln P C E_{t - 1}$	−0.204* (0.031)	−0.836*** (0.189)	−0.503*** (0.024)	−0.803*** (0.077)	−0.717*** (0.240)	−0.275*** (0.078)	−0.807*** (0.146)
$\ln P G D P$		0.439** (0.219)		0.437** (0.189)	0.177 (0.189)	−0.023*** (0.006)	0.317*** (0.095)
$\ln E S$		0.129** (0.066)	0.123*** (0.016)		0.222*** (0.024)	−0.209 (0.224)	0.139 (0.085)
$\ln I S$		0.335*** (0.003)	−0.026 (0.192)	0.386** (0.030)		0.087 (0.131)	0.212 (0.153)
$\ln P D$		0.352*** (0.067)	0.198*** (0.041)	0.307*** (0.023)	0.292*** (0.045)		0.316*** (0.021)
$\ln U R$		−0.420 (0.426)	0.600*** (0.024)	−0.452 (0.503)	0.327 (0.221)	0.490*** (0.014)
Constant	0.498* (0.068)	−1.245 (1.153)	−3.112** (0.787)	−0.566 (1.708)	−2.899*** (0.934)	−0.910* (0.485)	−2.200*** (0.275)
$R^{2}$	0.2771	0.5302	0.4076	0.5157	0.4621	0.2400	0.5161

Note: Robust standard errors are in brackets.

OLS: ordinary least square.

*p < 0.1; **p < 0.05; ***p < 0.01.

When estimating the two municipalities sample as the process in the previous section, the spatial effects are rejected by our preliminary test. As a result, the panel random effects model is used directly. As shown in Table 7, all variables are significantly positive except for the urbanization rate. A 10% increase in the per capita GDP, energy structure, industrial structure and population density will raise the growth rate of per capita carbon emissions by 4.39%, 1.29%, 3.35% and 3.52%, respectively. For this group, the industrial structure and population density are significantly positive, whereas a significant effect on the growth rate of per capita carbon emissions is not observed in the other two groups. This finding may be related to the relatively slight changes in the industrial structure and population density of the other two groups during the sample period; thus, these factors did not have enough impact on the growth rate of per capita carbon emissions. For the two municipalities, the average of the proportion of the secondary industries accounting for total GDP at the final period decreased by 14.3% compared with that in the initial period, whereas the average of all 74 cities only increased by 5.1%. The average population density increased by 9% during the sample period, whereas the overall increase was only 4.5%.

In addition, the per capita GDP and the ratio of coal to total energy consumption (energy structure) positively contributed to the per capita carbon emissions growth rate in all sub-samples, especially in the provincial capital and sub-capital cities. As shown in Model II of Tables 5 to 7, a 10% increase in the per capita GDP will increase the per capita carbon emission growth rate of the three groups by 2.55%, 4.98% and 4.39%, respectively. A 10% increase in the share of coal consumption in total energy will increase the per capita carbon emissions growth rate by 3.0%, 4.13% and 1.29%, respectively.

Analysis and discussion

The degree of convergence of per capita carbon emissions varies between the different groups. Generally, a higher level of economic development corresponds to a higher degree of convergence of per capita carbon emissions. As shown in Model II of Tables 5 to 7, the $β$ values of the 62 prefecture-level cities, 10 capital and sub-provincial cities and 2 municipalities are −0.413, −0.628 and −0.836, respectively. Cities in the group with the highest level of economic development have larger absolute values of $β$ , which indicates that the degree of convergence of per capita carbon emissions is higher. This finding might appear because the cities with lower development levels still primarily rely on energy-intensive and emission-intensive industries to drive the economy. For example, the per capita GDP of 62 prefecture-level cities averaged 3.35 thousand yuan during the sample period, which accounted for approximately 54% of the average of the 10 capital and sub-provincial cities group and 45% of the average of the two municipalities group. Although the proportion of value added by secondary industry has declined in recent years, at the end of the sample period (2015), the share of secondary industries of prefecture-level cities still accounted for almost 50%. Changing the industrial development patterns of these 62 prefecture-level cities, which presents high energy consumption and emissions, is difficult over a short time. Hence, the challenges faced by these cities in achieving the convergence of per capita carbon emissions are even more difficult. Per capita carbon emissions require a relatively long time to reach a stable equilibrium level, which has resulted in a low degree of convergence. However, the development of cities in provincial and sub-provincial cities and municipalities is comparatively mature. The process of industrialization and urbanization in these cities has basically entered the later stage. Per capita carbon emissions have gradually approached a steady state, and the differences between cities in emissions have also narrowed. Therefore, the convergence degree of these two groups is much higher.

For cities in the YREB, the per capita GDP is the most important factor for the convergence of per capita carbon emissions. Because the estimated coefficient ( $β$ ) of $\ln P C E_{t - 1}$ is a critical condition for determining whether convergence occurs and identifying the degree of convergence, we can obtain the effect of one specific control variable on the convergence state by comparing the $β$ value before and after adding that specific variable. As shown in Table 4, when the five variables per capita GDP, energy structure, industrial structure, population density and urbanization rate are individually removed from Model II, the absolute values of $β$ decrease by 43.2%, 8.5%, 0.51%, −0.26% and 0.26%, respectively. Obviously, per capita GDP is a vital factor for the per capita carbon emissions convergence in the YREB region. However, reducing emissions at the expense of per capita GDP is contrary to the principle of urban development. Therefore, to improve the convergence state of per capita carbon emissions in the YREB, the relationship between urban economic development and carbon emissions reductions must be effectively balanced. Although the energy structure does not affect the convergence state as important as the per capita GDP, optimizing the energy structure is more effective than adjusting the industrial structure and population structure. Therefore, developing clean energy to improve the current energy mix of regions which use coal as the core energy product is an effective method of promoting per capita carbon emissions convergence. The urbanization level and population density have very limited effects on the convergence of the whole sample, and a significant change in the absolute value of $β$ is not observed. These findings might be related to the small proportion of carbon emissions from residential consumption among the total energy consumption. Thus, the impact of population density on convergence is rather limited. In addition, the insignificant action of urbanization indicates that urbanization may not be the main factor that influences the per capita carbon emissions convergence in the YREB cities.

In different sub-samples, the same variable leads to a considerable difference in the convergence degree of per capita carbon emissions. For example, when the per capita GDP is added into the model, $β$ changes from −0.335 to −0.503 in terms of prefecture-level cities as listed in Tables 5 to 7. For the 10 capital and sub-provincial cities, $β$ changes from an insignificant to significant value at the 1% significance level. The $β$ value of the two municipalities changes from −0.503 to −0.836, and the absolute value increases by 66.2%. After considering the energy structure, the $β$ value of the prefecture-level cities sample changes obviously and shows an absolute increase of 14.4%, whereas it only increases by 2.45% and 4.11% in the capital and sub-provincial cities and municipalities, respectively. For prefecture-level cities, the industrial structure and population density are not conducive to the convergence of per capita carbon emissions, although these factors have positive effects on the convergence of two municipalities. In the content of the urbanization rate, the $β$ values of the three sub-samples barely changed and increased by 0.24%, 0.96% and 3.59%.

Conclusions and policy recommendations

Conclusions

By adopting the panel data of 74 cities in the Yangtze River Economic Belt in the period from 2005 to 2015, this study investigates the $σ$ convergence, stochastic convergence and $β$ convergence of per capita carbon emissions. The following conclusions have been obtained.

$σ$ convergence and $β$ conditional convergence are observed in 74 cities, although stochastic convergence has not been observed. During the sample period, the CV of per capita carbon emissions shows a downward trend in general, and the differences in per capita carbon emissions among cities decrease. According to the results of the HT and IPS unit root tests, the per capita carbon emissions are not a stationary series and the shock to the average level is not temporary. For the whole sample and the three subgroups, $β$ conditional convergence is observed for per capita carbon emissions. That is, the per capita carbon emissions of different cities will eventually tend to various steady states.

In addition, spatial lag effects are observed for the per capita carbon emission growth rate and spatial disturbance effects of the error term for the whole sample and the prefecture-level cities. Moreover, for the capital and sub-provincial cities, a spatial lag is observed on the per capita carbon emission growth rate and one lagged per capita carbon emissions.

b. The per capita GDP and energy consumption structure are the key factors that affect the convergence of per capita carbon emissions in the YREB cities, although the urbanization rate has little impact on convergence. According to $β$ conditional convergence, the per capita GDP and energy structure have significant positive impacts on the growth rate of per capita carbon emissions, and $β$ changes dramatically when these two variables are added into the model. However, in the whole sample, an insignificant but positive relationship is observed between the urbanization rate and per capita carbon emissions. Furthermore, the impact of the urbanization rate on convergence is also not significant. It may due to the fact that the related positive and negative effects of urbanization on carbon emissions are mutually offset.

c. Industrial structure and population density play important roles in achieving the per capita carbon emission convergence of the provincial capital and sub-capital cities and municipalities, respectively. Although a positive but not significant relationship is observed between industrial structure and the growth rate of per capita carbon emissions in the capital and sub-provincial cities, the absolute value of $β$ increases by 9.8% after taking the industrial structure into account. However, the population density has a significant effect on the growth rate of per capita carbon emissions only for the two municipalities. For the municipalities, a 10% increase in population density will increase the growth rate of per capita carbon emissions by 3.52%, which is surpassed only by the effect of the per capita GDP (4.39%).

Policy recommendations

To further ensure that the carbon emissions in the YREB region grow stably and achieve the emissions peak target as soon as possible, the following policy recommendations are proposed based on the results in this research.

Strengthen the improvement of the energy consumption structure and moderately control GDP growth for all 74 cities. Per capita GDP and energy structure are the key factors for the convergence of per capita carbon emissions of the YREB. Therefore, although greater efforts should be made to develop renewable clean energy and reduce traditional fossil energy, especially coal, the rate of GDP growth should not be overemphasized because it needs to be properly controlled to ensure the stability of per capita carbon emissions.

Accelerate the adjustment of the industrial structure of the capital and sub-provincial cities. For this group of cities, the share of industry in GDP has a significant impact on the convergence of per capita carbon emissions. Therefore, reducing the proportion of industrial sectors in GDP and changing the structure from a manufacturing-led high-emissions structure to service-led low-emissions would be more conducive to achieving a steady state of per capita carbon emissions in such cities.

Control the population of the two municipalities and reduce their population density. Compared with the cities in the other two groups, population density is an important factor that affects the convergence of carbon emissions for Shanghai and Chongqing. Therefore, controlling the population of these two cities with the unchanged amount of urban area would promote the realization of the convergence of per capita carbon emissions.

Supplemental Material

Supplemental material for Convergence of per capita carbon emissions in the Yangtze River Economic Belt, China

Supplemental Material for Convergence of per capita carbon emissions in the Yangtze River Economic Belt, China by Shiwei Yu, Xing Hu, Xuejiao Zhang and Zhenxi Li in Energy & Environment

Footnotes

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors gratefully acknowledge the financial support from the National Natural Science Foundation of China (grant no. 71573236).

Notes

Supplemental material

Supplemental material for this article is available online.

Shiwei Yu received his BE and PhD degrees in Business Management and Resources Management Engineering at China University of Geosciences, Wuhan, China, in 2003 and 2008, respectively. He is currently working as a Professor in School of Economic and Management, China University of Geosciences. His research interests includes economic environment and energy modelling, carbon dioxide emissions reduction policy, and energy system optimization.

Xing Hu is presently pursuing her Master's at School of Economics and Management, China University of Geosciences, Wuhan, China. Her research interests includes carbon emission and carbon intensity convergence of different sectors and regions, energy and climate change economics.

Xuejiao Zhang is currently pursuing her Master's at School of Economics and Management, China University of Geosciences, Wuhan, China. Her research interests includes energy economics and climate change.

Zhenxi Li is presently pursuing her Master's at School of Economics and Management, China University of Geosciences, Wuhan, China. Her research interests includes real option in geothermal investment and multi-objective optimization of renewable energy.

References

Xinhuanet. Available at: www.xinhuanet.com/energy/2016-10/31/c_1119816624.htm (2016, accessed 7 November 2018).

NBSC. National Bureau of Statistics. Available at: www.stats.gov.cn/ (2017, accessed 7 November 2018).

Karakaya

Alataş

Yılmaz

Replication of Strazicich and List (2003): are CO2 emission levels converging among industrial countries?

Energy Econ 2017, https://dx-doi-org.web.bisu.edu.cn/10.1016/j.eneco.2017.08.033

Huang

Yang

et al . Convergence of carbon intensity in the Yangtze River Delta, China. Habitat International 2017; 60: 58–68.

Lin

Global convergence in per capita CO₂ emissions. Renew Sustain Energ Rev 2013; 24: 357–363.

Solow

RM.

A contribution to the theory of economic growth. Q J Econ 1956; 70: 65–94.

Barro

RJ.

Convergence. J Polit Econ 1992; 100: 223–251.

K-W

Zhou

Pan

Cross-country output convergence and growth: evidence from varying coefficient nonparametric method. Econ Model 2016; 55: 32–41.

Quah

DT.

Empirics for economic growth and convergence. Eur Econ Rev 1996; 40: 1353–1375.

10.

Yang

Zhang

Sheng

et al . Carbon dioxide emissions and interregional economic convergence in China. Econ Model 2016; 52: 672–680.

11.

Ben-David

Trade and convergence among countries. J Int Econ 1996; 40: 279–298.

12.

Jobert

Karanfil

Tykhonenko

Convergence of per capita carbon dioxide emissions in the EU: Legend or reality?

Energy Econ 2010; 32: 1364–1373.

13.

Lee

C-C

Chang

C-P.

Stochastic convergence of per capita carbon dioxide emissions and multiple structural breaks in OECD countries. Econ Model 2009; 26: 1375–1381.

14.

Romero-Ávila

Convergence in carbon dioxide emissions among industrialised countries revisited. Energy Econ 2008; 30: 2265–2282.

15.

Strazicich

List

JA.

Are CO₂ emission levels converging among industrial countries?

Environ Resour Econ 2003; 24: 263–271.

16.

Hao

Liao

Wei

Y-M.

Is China’s carbon reduction target allocation reasonable? An analysis based on carbon intensity convergence. Appl Energy 2015; 142: 229–239.

17.

Wang

Zhang

Convergence of carbon dioxide emissions in different sectors in China. Energy 2014; 65: 605–611.

18.

Guo

et al . Convergence of carbon dioxide emissions in Chinese cities: a continuous dynamic distribution approach. Energy Pol 2016; 91: 207–219.

19.

Zhao

Wesley Burnett

Lacombe

DJ.

Province-level convergence of China’s carbon dioxide emissions. Appl Energy 2015; 150: 286–295.

20.

Brännlund

Lundgren

Söderholm

Convergence of carbon dioxide performance across Swedish industrial sectors: an environmental index approach. Energy Econ 2015; 51: 227–235.

21.

Moutinho

Robaina-Alves

Mota

Carbon dioxide emissions intensity of Portuguese industry and energy sectors: a convergence analysis and econometric approach. Renew Sustain Energ Rev 2014; 40: 438–449.

22.

Fan

et al . Convergence of carbon emissions intensity across Chinese industrial sectors. J Clean Prod 2018; 194: 179–192.

23.

Barassi

Cole

Elliott

RJR.

Stochastic divergence or convergence of per capita carbon dioxide emissions: re-examining the evidence. Environ Res Econ 2008; 40: 121–137.

24.

Camarero

Mendoza

Ordóñez

2011. Re-examining CO₂ emissions. Is the assessment of convergence meaningless? Working Papers, http://www.doctreballeco.uji.es/wpficheros/camareroetal2011.pdf

25.

Lee

C-C

Chang

C-P.

New evidence on the convergence of per capita carbon dioxide emissions from panel seemingly unrelated regressions augmented Dickey–Fuller tests. Energy 2008; 33: 1468–1475.

26.

Panopoulou

Pantelidis

Club convergence in carbon dioxide emissions. Environ Resource Econ 2009; 44: 47–70.

27.

Phillips

PCB

Sul

Transition modeling and econometric convergence tests. Econometrica 2007; 75: 1771–1855.

28.

Burnett

JW.

Club convergence and clustering of U.S. energy-related CO₂ emissions. Res Energy Econ 2016; 46: 62–84.

29.

Yavuz

Yilanci

Convergence in per capita carbon dioxide emissions among G7 countries: a TAR panel unit root approach. Environ Resource Econ 2013; 54: 283–291.

30.

Aldy

JE.

Per capita carbon dioxide emissions: convergence or divergence?

Environ Resource Econ 2006; 33: 533–555.

31.

Barassi

Cole

Elliott

RJR.

The stochastic convergence of CO₂ emissions: a long memory approach. Environ Res Econ 2011; 49: 367–385.

32.

Apergis

Payne

JE.

Per capita carbon dioxide emissions across U.S. states by sector and fossil fuel source: evidence from club convergence tests. Energy Econ 2017; 63: 365–372.

33.

Presno

Landajo

González

Stochastic convergence in per capita CO₂ emissions. An approach from nonlinear stationarity analysis. Energy Econ 2018; 70: 563–581.

34.

C-Q

Huang

Research on the performance and synergistic effect of green development of Yangtze River Economic Belt’s industry. J China Univ Geosci 2018. (In Chinese.)

35.

Wang

X-X

Peng

C-J

et al . Eco-environment health assessment of cities in the Yangtze River in economic zone based on urbanization approach. J Hunan Univ 2015; 42: 132–140. (In Chinese.)

36.

Jiang

A study on environmental justice evaluation of the Yangtze River Economic Belt. Rural Econ Sci Technol 2016; 27: 1–2. (In Chinese.)

37.

Chen

Environmental efficiency analysis of the Yangtze River Economic Zone using super efficiency data envelopment analysis (SEDEA) and Tobit models. Energy 2017; 134: 659–671.

38.

C-Q

Research on the effect of biased technical change on total factor energy efficiency of the Yangtze River Economic Belt. China Soft Sci 2018. (In Chinese.)

39.

Wang

Meng

Yang

et al . The industrial eco-efficiency of the Yangtze River Economic Zone based on environmental pressure. Resour Sci 2015; 37: 1491–1501. (In Chinese).

40.

Zhao

Sun

Tao

Energy eco-efficiency evaluation and convergence analysis of Yangtze River Economic Belt from the perspective capacity reduction. J Taiyuan Univ Technol 2016; 34: 45–50. (In Chinese.)

41.

Xing

Z-C

J.-G

Zhang

Research on spatial-temporal variation and evolution of total-factor ecological efficiency in the Yangtze River Economic Belt. Resource Environ Yangtze Basin. 2018; 27, (In Chinese).

42.

Huang

Analysis of influential factors of carbon emission based on spatial panel models: a case study for Yangtze River Economic Belt. Resource Environ Yangtze Basin. 2015; 24, 1665–1671. (In Chinese.)

43.

Jiang

Liu

et al . Regional disparities and influencing factors of average CO₂ emissions from transportation industry in Yangtze River Economic Belt. Transp Res D Transp Environ 2017; 57: 112–123.

44.

Lin

The influencing factors on the carbon emission of electric power industry: an empirical study based on Yangtze Economic Belt spatial dynamic panel. J Environ Econ 2016; 1: 91–105. (In Chinese.)

45.

Sun

Chen

Tian

Study on the urban state carrying capacity for unbalanced sustainable development regions: evidence from the Yangtze River Economic Belt. Ecol Indic 2018; 89: 150–158.

46.

Tian

Sun

Comprehensive carrying capacity, economic growth and the sustainable development of urban areas: a case study of the Yangtze River Economic Belt. J Clean Prod 2018; 195: 486–496.

47.

Tian

Sun

A spatial differentiation study on comprehensive carrying capacity of the urban agglomeration in the Yangtze River Economic Belt. Reg Sci Urban Econ 2018; 68: 11–22.

48.

Tan

An inquiry into water transfer network of the Yangtze River Economic Belt in China. J Clean Prod 2018; 176: 288–297.

49.

Anselin

Spatial econometrics: methods and models. Econ Geogr 1988; 65: 160–162.

50.

Anselin

Spatial effects in econometric practice in environmental and resource economics. Am J Agri Econ 2001; 83: 705–710.

51.

Lee

L-F

Estimation of spatial autoregressive panel data models with fixed effects. J Econom 2010; 154: 165–185.

52.

Fixed-effects dynamic spatial panel data models and impulse response analysis. J Econom 2017; 198: 102–121.

53.

Marbuah

Amuakwa-Mensah

Spatial analysis of emissions in Sweden. Energy Econ 2017; 68: 383–394.

54.

Shi

Lee

L-F.

Spatial dynamic panel data models with interactive fixed effects. J Econom 2017; 197: 323–347.

55.

Meng

Wang

Andrew

et al . Spatial spillover effects in determining China's regional CO₂ emissions growth: 2007–2010. Energy Econ 2017; 63: 161–173.

56.

Liu

Wang

et al . Regional inequality, spatial spillover effects, and the factors influencing city-level energy-related carbon emissions in China. J Geogr Sci 2018; 28: 495–513.

57.

Chuai

Huang

Wang

et al . Spatial econometric analysis of carbon emissions from energy consumption in China. J Geogr Sci 2012; 22: 630–642.

58.

Zhang

Yang

Sun

et al . Analyzing the impact factors of energy-related CO₂ emissions in China: what can spatial panel regressions tell us? J Clean Prod 2017; 161: 1085–1093.

59.

Zhang

Zhao

Identification of the driving factors’ influences on regional energy-related carbon emissions in China based on geographical detector method. Environ Sci Pollut Res 2018; 25: 9626–9635.

60.

Wang

Re-examining environmental Kuznets curve for China’s city-level carbon dioxide (CO₂) emissions. Spat Stat 2017; 21: 377–389.

61.

Meng

Huang

Shaping the relationship between economic development and carbon dioxide emissions at the local level: evidence from spatial econometric models. Environ Res Econ 2018; 71: 127–156.

62.

Tang

Liu

Gong

The measurement of the spatial effects of Chinese regional carbon emissions caused by exports. J Geogr Sci 2015; 25: 1328–1342.

63.

Zhang

Exploring the spatial spillover effects of industrialization and urbanization factors on pollutants emissions in China's Huang-Huai-Hai region. J Clean Prod 2018; 195: 154–162.

64.

Chen

et al . A gravity model and exploratory spatial data analysis of prefecture-scale pollutant and CO₂ emissions in China. Ecol Indic 2018; 90: 554–563.

65.

Huang

Meng

Convergence of per capita carbon dioxide emissions in urban China: a spatio-temporal perspective. Appl Geogr 2013; 40: 21–29.

66.

Adhikari

Chen

Energy productivity convergence in Asian countries: a spatial panel data approach. Int J Econ Financ 2014; 6: 94–107.

67.

Bernard

Durlauf

SN.

Interpreting tests of the convergence hypothesis. J Econom 1996; 71: 161–173.

68.

Harris

RDF

Tzavalis

Inference for unit roots in dynamic panels where the time dimension is fixed. J Econom 1999; 91: 201–226.

69.

Pesaran

Shin

Testing for unit roots in heterogeneous panels. J Econom 2003; 115: 53–74.

70.

Niu

Research on regional difference and convergence of China's carbon emission efficiency from the perspective of total factors. Chengdu, China: Southwestern University of Finance and Economics, 2016 (In Chinese).

71.

Liddle

Revisiting world energy intensity convergence for regional differences. Appl Energy 2010; 87: 3218–3225.

72.

S-C

Z-X

Long

R-Y.

Factors that influence carbon emissions due to energy consumption in China: decomposition analysis using LMDI. Appl Energy 2014; 127: 182–193.

73.

Wang

Zhang

et al . Examining the driving factors of energy related carbon emissions using the extended STIRPAT model based on IPAT identity in Xinjiang. Renew Sustain Energ Rev 2017; 67: 51–61.

74.

Zhang

Chen

How does urbanization affect carbon dioxide emissions? A cross-country panel data analysis. Energy Pol 2017; 107: 678–687.

75.

Zhang

Zheng

et al . Provincial carbon intensity abatement potential estimation in China: a PSO–GA-optimized multi-factor environmental learning curve method. Energy Pol 2015; 77: 46–55.

76.

Wang

Zhang

Huang

et al . Convergence behavior of carbon dioxide emissions in China. Econ Model 2014; 43: 75–80.

77.

Sharma

SS.

Determinants of carbon dioxide emissions: empirical evidence from 69 countries. Appl Energy 2011; 88: 376–382.

78.

Sheng

Guo

The long-run and short-run impacts of urbanization on carbon dioxide emissions. Econ Model 2016; 53: 208–215.

79.

Sadorsky

The effect of urbanization on CO₂ emissions in emerging economies. Energy Econ 2014; 41: 147–153.

80.

Rafiq

Salim

Nielsen

Urbanization, openness, emissions, and energy intensity: a study of increasingly urbanized emerging economies. Energy Econ 2016; 56: 20–28.

81.

Zhang

et al . Analysis on influence factors of China's CO₂ emissions based on Path–STIRPAT model. Energy Pol 2011; 39: 6906–6911.

82.

Liu

Kang

Changes in carbon intensity in China's industrial sector: decomposition and attribution analysis. Energy Pol 2015; 87: 28–38.

83.

Yang

Fan

Shao

et al . Does carbon intensity constraint policy improve industrial green production performance in China? A quasi-DID analysis. Energy Econ 2017; 68: 271–282.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.69 MB

Convergence of per capita carbon emissions in the Yangtze River Economic Belt,China

Abstract

Keywords

Introduction

Literature review

Literature on convergence of CO2 emissions

Literatures on green development model of the Yangtze River Economic Belt

Literatures on spatial effects of carbon emissions

Methodology

σ convergence

Stochastic convergence

β convergence

Data management

Cities in the Yangtze River Economic Belt

Carbon emissions estimation

Energy consumption structure

Other variables

Results and discussion

Test of σ convergence

Test of stochastic convergence

Test of β convergence

Spatial model specification

Absolute convergence and conditional convergence results

Analysis and discussion

Conclusions and policy recommendations

Conclusions

Policy recommendations

Supplemental Material

Supplemental material for Convergence of per capita carbon emissions in the Yangtze River Economic Belt, China

Footnotes

Declaration of conflicting interests

Funding

Notes

Supplemental material

References

Supplementary Material

Literature on convergence of CO₂ emissions

$σ$ convergence

$β$ convergence

Test of $σ$ convergence

Test of $β$ convergence