Driving factors of carbon dioxide emissions in China: an empirical study using 2006--2010 provincial data

Abstract

The rapid urbanization of China has increased pressure on its environmental and ecological well being. In this study, the temporal and spatial profiles of China’s carbon dioxide emissions are analyzed by taking heterogeneities into account based on an integration of the extended stochastic impacts using a geographically and temporally weighted regression model on population, affluence, and technology. Population size, urbanization rate, GDP per capita, energy intensity, industrial structure, energy consumption pattern, energy prices, and economy openness are identified as the key driving factors of regional carbon dioxide emissions and examined through the empirical data for 30 provinces during 2006–2010. The results show the driving factors and their spillover effects have distinct spatial and temporal heterogeneities. Most of the estimated time and space coefficients are consistent with expectation. According to the results of this study, the heterogeneous spatial and temporal effects should be taken into account when designing policies to achieve the goals of carbon dioxide emissions reduction in different regions.

Keywords

carbon dioxide emission / heterogeneity / space spillover

Introduction

Along with the rapid progress of urbanization, China has experienced fast economic growth as well as environmental and ecological degradation during the past decades. One of the most concerning environmental and ecological issues is air pollution in urban areas, which started attracting broad attention after the 2013 Eastern China Smog event. In early 2013, Beijing, Tianjin, Shanghai, and most of the North and Northeast area of China were subject to serious fog haze weather; e.g., only five days were haze free for Beijing in January, 2013, impacting industrial production, transportation, and people’s health.

Since the use of fossil energy is considered a primary cause of air pollution, the transition to a low-carbon economy is deemed to be a promising solution ( Ji et al., 2014). Carbon dioxide emissions are a key indicator to measure how carbon-intensive an economy is; analyzing the profiles and driving factors of carbon dioxide emission has become a hot topic in recent years and multiple methodologies have been applied at national and regional levels (see e.g., Chen and Chen, 2010, 2011; Chen and Zhang, 2010; Chen et al., 2013; Liu et al., 2013; Li et al., 2014; Xia et al., 2014a, b, 2015; Zhang et al., 2014).

Compared with the other methodologies, the stochastic impacts by regression on population, affluence, and technology (STIRPAT) model developed by Dietz and Rosa (1994) allows a precise specification of environmental and ecological impacts to the forces driving them. This model can be applied to estimate the coefficient parameters and identify the various affecting factors ( Dietz and Rosa, 1994), based on which a variety of studies were conducted ( Dietz and Rosa, 1997; York et al, 2003). However, as a country experiencing rapid changes in various aspects including but not limited to population, wealth, and technology, many other factors such as urbanization, industrialization, energy consumption, energy affordability, and economy openness also impact the environmental and ecological changes of China. Therefore, it is necessary to extend the STIRPAT model to take into account the contributions of these other major factors as well as the fast-developing features when investigating the Chinese economy. Besides, there are notable differences between different regions within China because of its vast territory and unbalanced regional economy development mode. Therefore, spatial heterogeneity has to be considered when different cities, provinces, or regions of China are analyzed and compared. To address these problems, the geographically and temporally weighted regression (GTWR) model brought forward by Gelfand et al. (2003) and Huang et al. (2009, 2010) can be applied to reflect the regional differences and the dynamic features of China.

In this study, the temporal and spatial profiles of China’s carbon dioxide emission are analyzed by taking heterogeneity into account based on an integration of the extended STIRPAT model with the GTWR model. Population size, urbanization rate, GDP per capita, energy intensity, industrial structure, energy consumption pattern, energy prices, and economy openness are identified as the driving factors of regional carbon dioxide emission and examined through the empirical data for 30 provinces during 2006‒2010. The rest of this paper is organized as follows: Section 2 addresses the methodological and data issues; Section 3 presents results and discussion; and Section 4 concludes and provides policy recommendations.

Method and data

Extension of the STIRPAT Model

Dietz and Rosa (1994) constructed the STIRPAT model to estimate the random coefficients of carbon emissions by statistical regression on population, wealth, and technological condition. The advantage of this model is that it not only informs basic science of the anthropogenic impacts on environmental and ecological systems, but also points to the factors most responsive to policy, which is expressed as

(1) I=a{P}^b{A}^c{T}^{d}e

,

where I, P, A, and T stand for environmental impact, population size, affluence, and technology level, respectively; a, b, c, d, and e are the model coefficient, population index, affluence index, technology level index, and model error, respectively. Taking natural logarithms on both sides makes

(2) \ln I=\ln a+b(\ln P)+c(\ln A)+d(\ln T)+\ln e

.

In order to simulate the fast-developing features of the Chinese economy, the STIRPAT model can be extended by introducing the factors of urbanization rate, industrial structure, energy consumption pattern, energy prices, and economy openness as

(3) \begin{array}{l}\ln I\\ \\ \\ \end{array}\begin{array}{l}=\ln a+{\beta }_{\mathrm{1}}(\ln P)+{\beta }_{\mathrm{2}}(\ln UR)+{\beta }_{\mathrm{3}}(\ln GDPPC)+{\beta }_{\mathrm{4}}(\ln EI)\\ +{\beta }_{\mathrm{5}}(\ln IS)+{\beta }_{\mathrm{6}}(\ln ECP)+{\beta }_{\mathrm{7}}(\ln EP)+{\beta }_{\mathrm{8}}(\ln OPEN)+\ln e\\ =\alpha +{\beta }_{\mathrm{1}}(\ln P)+{\beta }_{\mathrm{2}}(\ln UR)+{\beta }_{\mathrm{3}}(\ln GDPPC)+{\beta }_{\mathrm{4}}(\ln EI)\\ +{\beta }_{\mathrm{5}}(\ln IS)+{\beta }_{\mathrm{6}}(\ln ECP)+{\beta }_{\mathrm{7}}(\ln EP)+{\beta }_{\mathrm{8}}(\ln OPEN)+\epsilon \end{array}

,

where P is population measured by ten thousand people; UR is urbanization rate measured in percentage of urban population from total population, GDPPC is per capita GDP measured in CNY of GDP per person, EI is energy intensity measured in tons of standard coal of energy consumption per ten thousand CNY of GDP; IS is industrial structure measured by the percentage of GDP generated by secondary industry from total GDP; ECP is energy consumption pattern measured by the percentage of coal consumption from total energy consumption; EP is energy price measured by weighted PPI of all regions; OPEN is economy openness measured by billion US dollars of foreign investment registration situation. At the same time, \ln a=\alpha is the model coefficient, {\beta }_i (i=1 to 8) are the coefficients for corresponding factors, and \ln e=\epsilon is the model error. In this study, the volume of carbon dioxide emission (measured by ten thousand tons) will be examined as the environmental impact.

Geographically and temporally weighted regression model

The geographically weighted regression (GWR) model proposed by Brunsdon et al. (1996, 1998, 1999) and Fotheringham et al. (1996, 1998)is so far the most popular model to study spatial heterogeneity. It has been widely used, especially in the field of real estate, because of its flexibility in depicting different spatial location. By introducing space effects based on the GWR model, Song and Su (2010) studied the provincial carbon dioxide emission and economic development of China and drew some important conclusions. However, the GWR model has a weak point: as a cross section data model, the GWR model neglects the time effect and thus is unable to capture the dynamic factors of the concerned system. To address this problem, the GTWR model was brought forward by Gelfand et al. (2003) and Huang et al. (2009, 2010) by integrating the temporal effect into the GWR model, which captures parameter variation from both time and space dimensions. The GTWR Model is established as

(4) I_i=\alpha (u_i,v_i,t_i)+{\displaystyle \sum _{k=\mathrm{1}}^p{\beta }_k(u_i,v_i,t_i)x_{ik}}+{\epsilon }_i

,

where I_i is the carbon dioxide emission in region i, x_ik is the k^th driving factor for region i, u_i is the longitude of region i, v_i is the latitude of region i, t_i is the time variable for region i, and \alpha (u_i,v_i,t_i) , {\beta }_k(u_i,v_i,t_i) , and {\epsilon }_i are the model coefficient, factor coefficient, and model error, respectively. To separate the spillover effect from local impact, a spatial lag term can be added to the GTWR model as

(5) {{\displaystyle I}}_i=\alpha ({{\displaystyle u}}_i,{{\displaystyle v}}_{i,}{{\displaystyle t}}_{i,})+{\displaystyle \sum _{k=\mathrm{1}}^p{{\displaystyle \beta }}_k}({{\displaystyle u}}_i,{{\displaystyle v}}_{i,}{{\displaystyle t}}_{i,}){{\displaystyle x}}_{ik}+{\displaystyle \sum _{k=\mathrm{1}}^p{{\displaystyle \beta }}_k}({{\displaystyle u}}_i,{{\displaystyle v}}_{i,}{{\displaystyle t}}_{i,})W{{\displaystyle x}}_{ik}+{{\displaystyle \epsilon }}_i,

where {{\displaystyle \beta }}_k({{\displaystyle u}}_i,{{\displaystyle v}}_i,{{\displaystyle t}}_i) is the coefficient of the spatial lag term and W is the space-time weight matrix whose element is calculated as

(6) w_{ij}^{ST}=\exp \left\{-\left({\displaystyle \frac{\alpha [{(u_i-u_j)}^{\mathrm{2}}+{(v_i-v_j)}^{\mathrm{2}}]+\beta {(t_i-t_j)}^{\mathrm{2}}}{b_{ST}^{\mathrm{2}}}}\right)\right\}

,

in which b_{ST} is the optimal bandwidth of the temporal weighting function chosen through a cross validation methodology. Since {\beta }_k(u_i,v_i,t_i) measures how a province’s carbon dioxide emission is impacted by the k^th driving factor of local and neighboring provinces, hereafter the value of the coefficient will be interpreted as the integrated effect. On the other hand, the coefficient of {{\displaystyle \beta }}_k({{\displaystyle u}}_i,{{\displaystyle v}}_i,{{\displaystyle t}}_i) , which measures the impact of the k^th local driving factor on local carbon dioxide emission, will be interpreted as the local effect.

Data sources

In order to study the spatial and temporal heterogeneity of the driving factors of carbon dioxide emission in China, the cross section data from 2006 to 2010 for 30 provincial regions (Tibet and Taiwan are excluded due to data availability) are used for an empirical analysis. The spatial and temporal changes of carbon dioxide emission from electric power and thermal generation in different regions are identified by distinguishing the dynamic changes of standard energy transformation coefficients calculated from China’s Energy Statistical Yearbook 2007‒2011. The other indicators, such as the urbanization rate and energy consumption pattern, are obtained from or calculated based on China’s Statistical Yearbook 2007‒2011 and China’s Energy Statistical Yearbook 2007‒2011.

Results and discussion

On the basis of the driving factors identified by the extended STIRPAT model, i.e., population, urbanization rate, GDP per capita, energy intensity, industrial structure, energy consumption pattern, energy price, and economy openness, the GTWR model is employed to examine the spatial and temporal heterogeneities of the driving factors of carbon dioxide emissions. The statistics of the estimated factors affecting carbon dioxide emission are listed in Table 1 (provincial results are provided in supplementary material).

According to the results, the concerned driving factors make different contributions to regional carbon dioxide emissions. The spatial spillover effects differ significantly for different driving factors. There exists large variation in the estimated coefficients of population size, urbanization rate, GDP per capita, energy intensity, industrial structure, energy consumption pattern, energy prices, and economy openness, indicating notable differences in the driving factors in different provinces. Therefore, it is necessary to consider the heterogeneity of carbon dioxide emissions based on the GTWR model. The results also confirm that many of the driving factors have significant temporal change in estimated coefficients, and the signs of the coefficients are consistent with expectations. The temporal and spatial heterogeneity of each driving factor impacting regional carbon emissions are discussed as follows:

Population: The integrated effects of population on carbon dioxide for all provinces are positive, implying a general trend that population growth leads to increasing carbon dioxide emission. The local effect of population is larger in the eastern developed provinces as well as in the western least-developed provinces than in other provinces. The results also suggest that the temporal factor has insignificant impact (less than 1% change over the research period) on the population effect for most provinces.

Urbanization rate: The integrated effect of urbanization shows that the increase of local and neighboring provinces’ urbanization rates increases local carbon dioxide emission for the eastern and central provinces, but decreases it for the western provinces. The different impacts are introduced by different industrial structures and economic development stages between eastern and western China. For the more developed eastern regions, the increase of urbanization rate is usually accompanied with the development of tertiary industry with low carbon intensity. In contrast, the urbanization progress of the western regions is mainly supported by the expansion of secondary industry which has high carbon intensity. Meanwhile, the increase of local urbanization rate decreases carbon dioxide emission of most provinces. The local effect of urbanization rate in some western provinces, such as Xinjiang and Shaanxi, changes significantly over time, confirming the notable heterogeneous temporal effect in these provinces.

Economic development: The improvement of local and neighboring provinces’ economic status increases carbon dioxide emission of all provinces, but the effects are larger in the western provinces, which might be attributed to the rapid industrialization process in the those provinces. The local effects of economic development are also positive for most provinces. Regarding the temporal factor, the local effects of several provinces change significantly in different years but the integrated effects do not shown significant heterogeneity in most provinces.

Energy intensity: The driving factor of energy intensity explains how the change of technological progress will contribute to carbon dioxide emission. The results show that technological progress (leading to lower energy intensity) in local and neighboring provinces can significantly decrease carbon dioxide emission in the central and western regions, while the local effects are larger in the eastern regions than in the other regions. At the same time, the local effects of energy intensity in some western and southern provinces are affected significantly by temporal factors.

Industrial structure: The integrated effect of industrial structure is positive in most provinces, which suggests that along with the industrialization process of the whole country, China is facing intrinsic pressure to control its carbon dioxide emissions. The increasing local industrial structure contributes to the decrease of carbon dioxide emission in 12 provinces, while the other 18 provinces experience opposite effects. The heterogeneous temporal effect of industrial structure is much larger in the western provinces than in the other provinces.

Energy consumption pattern: The integrated effect of energy consumption patterns is positive on carbon dioxide emission in most provinces, while the local effect is positive in eastern provinces but negative in most other provinces. The temporal heterogeneity is generally insignificant for the driving factor of energy consumption pattern.

Energy price: In most provinces, the integrated and local effects of energy price are both negative, which means that higher energy price leads to lower carbon dioxide emission. The heterogeneous temporal effect is significant in most provinces except for the northern and eastern ones.

Economy openness: In comparison to the other driving factors, the change of openness has little and inconsistent impact on provincial carbon dioxide emissions, which might be attributed to two reasons with opposite impacts on carbon dioxide emissions: (a) foreign investors usually invest in secondary industry with high carbon dioxide emission intensity; but (b) the companies with foreign capital usually have advantage high technology level and thus have lower carbon dioxide emission intensity. Therefore, the temporal heterogeneity is quite significant for the driving factor of economy openness.

Conclusions and policy recommendations

In this study, the driving factors of carbon dioxide emission in China, i.e., population size, urbanization rate, per capita GDP, energy intensity, industrial structure, energy consumption pattern, energy price, and economy openness, are first identified by an extended STIRPAT model. Then the effects of the driving factors are examined based on a GTWR model using the provincial panel data from 2006 to 2010. The empirical results confirm the spatial and temporal heterogeneities of carbon dioxide emission in China. The main conclusions are as follows:

1) The impacts of different driving factors vary according to their estimated coefficients calculated by the GTWR model. Generally speaking, the integrated effect of energy price is negative on regional carbon dioxide emissions (which means higher local and nearby energy prices will decrease carbon dioxide emissions of the local province), but the integrated effects of the other seven driving factors are all positive (which means larger population size, higher urbanization rate, higher economy development status, higher energy intensity, larger share of secondary industry, larger share of coal consumption, and higher level of economy openness in local and neighboring provinces all contribute to higher regional carbon dioxide emission in local province).

2) Both the local and integrated effects of all the eight driving factors have significant spatial heterogeneity. The carbon dioxide emissions in the eastern and western provinces are more sensitive to the change of driving factors.

3) Temporal heterogeneity is generally weaker than spatial heterogeneity, which might be attributed to the fact that this study includes only five year period of research. In spite of this, some of the western provinces still have notable temporal heterogeneity.

Results of this study confirm that the spatial and temporal heterogeneities should be addressed to realize the regional differentiation for carbon emission reduction. Policy recommendations drawn from this study include: a) it is necessary to narrow the unbalanced nature of development across China, especially by accelerating the development of the western regions since their poorer economic status contributes to less efficient energy use and thus leads to higher carbon dioxide emission; b) since the technological effect is larger in the eastern regions than in other regions, technological innovation concentrated in the eastern regions will help to decrease carbon dioxide emission of the whole country; c) since the western regions are more sensitive to energy price change, a moderate increase of energy price in the western provinces can be applied to decrease overall carbon dioxide emission; d) accelerating the industrial structure adjustment to increase the share of tertiary industry is a efficient way to reduce carbon dioxide emission in most provinces.

References

[1] Brunsdon C, Fotheringham A S, Charlton M (1999). Some notes on parametric significance tests for geographically weighted regression. J Reg Sci, 39(3): 497–524

[2] Brunsdon C, Fotheringham A S, Charlton M E (1996). Geographically weighted regression: a method for exploring spatial nonstationarity. Geogr Anal, 28(4): 281–298

[3] Brunsdon C, Fotheringham S, Charlton M (1998). Geographically weighted regression-modelling spatial non-stationarity. Journal of the Royal Statistical Society : Series D (The Statistician), 47(3): 431–443

[4] Chen G Q, Chen Z M (2010). Carbon emissions and resources use by Chinese economy 2007: a 135-sector inventory and input–output embodiment. Commun Nonlinear Sci Numer Simul, 15(11): 3647–3732

[5] Chen G Q, Zhang B (2010). Greenhouse gas emissions in China 2007: inventory and input-output analysis. Energy Policy, 38(10): 6180–6193

[6] Chen Z M, Chen G Q (2011). Embodied carbon dioxide emission at supra-national scale: a coalition analysis for G7, BRIC, and the rest of the world. Energy Policy, 39(5): 2899–2909

[7] Chen Z M, Chen G Q, Chen B (2013). Embodied carbon dioxide emission by the globalized economy: a systems ecological input-output simulation. Journal of Environmental Informatics, 21(1): 35–44

[8] Dietz T, Rosa E A (1994). Rethinking the environmental impacts of population, affluence and technology. Hum Ecol Rev, 1: 277–300

[9] Dietz T, Rosa E A (1997). Effects of population and affluence on CO₂ emissions. Proc Natl Acad Sci USA, 94(1): 175–179

[10] Fotheringham A S, Charlton M, Brunsdon C (1996). The geography of parameter space: an investigation of spatial non-stationarity. International Journal of Geographical Information Systems, 10(5): 605–627

[11] Fotheringham S, Charlton M, Brunsdon C (1998). Geographically weighted regression: a natural evolution of the expansion method for spatial data analysis. Environ Plann A, 30(11): 1905–1927

[12] Gelfand A E, Kim H J, Sirmans C F, Banerjee S (2003). Spatial modeling with spatially varying coefficient processes. J Am Stat Assoc, 98(462): 387–396

[13] Huang B, Wu B, Barry M (2010). Geographically and temporally weighted regression for modeling spatio-temporal variation in house prices. Int J Geogr Inf Sci, 24(3): 383–401

[14] Huang B, Zhang L, Wu B (2009). Spatiotemporal analysis of rural–urban land conversion. Int J Geogr Inf Sci, 23(3): 379–398

[15] Ji X, Chen Z, Li J (2014). Embodied energy consumption and carbon emissions evaluation for urban industrial structure optimization. Front Earth Sci, 8(1): 32–43

[16] Li J S, Alsaed A, Hayat T, Chen G Q (2014). Energy and carbon emission review for Macao’s gaming industry. Renew Sustain Energy Rev, 29: 744–753

[17] Liu Y, Feng S, Cai S, Zhang Y X, Zhou X, Chen Y B, Chen Z M (2013). Carbon emission trading system of China: a linked market vs. separated markets. Front Earth Sci, 7(4): 465–479

[18] Song B Y, Su F L (2010). GWR empirical research of Chinese provincial carbon emissions and economic development. Finance & Ecnomics, 4: 41–49 (in Chinese)

[19] Xia X H, Chen Y B, Li J S, Tasawar H, Alsaedi A, Chen G Q (2014b). Energy regulation in China: objective selection, potential assessment and responsibility sharing by partial frontier analysis. Energy Policy, 66(0): 292–302

[20] Xia X H, Hu Y, Chen G Q, Alsaedi A, Hayat T, Wu X D (2014a). Vertical specialization, global trade and energy consumption for an urban economy: a value added export perspective for Beijing. Ecol Modell,

[21] Xia X H, Hu Y, Tasawar H, Alsaedi A, Wu X D, Chen G Q (2015). Structure decomposition analysis for energy-related GHG emission in Beijing: urban metabolism and hierarchical structure. Ecol Inform, 26: 60–69

[22] York R, Rosa E A, Dietz T (2003). STIRPAT, IPAT and ImPACT: analytic tools for unpacking the driving forces of environmental impacts. Ecol Econ, 46(3): 351–365

[23] Zhang B, Chen G Q, Li J S, Tao L (2014). Methane emissions of energy activities in China 1980–2007. Renew Sustain Energy Rev, 29: 11–21

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg