An empirical study on the dynamic effect of regional industrial carbon transfer in China

An empirical study on the dynamic effect of regional industrial carbon transfer in China

Ecological Indicators 73 (2017) 1–10 Contents lists available at ScienceDirect Ecological Indicators journal homepage: www.elsevier.com/locate/ecoli...

2MB Sizes 0 Downloads 43 Views

Ecological Indicators 73 (2017) 1–10

Contents lists available at ScienceDirect

Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind

Original Articles

An empirical study on the dynamic effect of regional industrial carbon transfer in China Jing Xu a , Ming Zhang a , Min Zhou a,∗ , Hailong Li b a b

School of Management, China University of Mining and Technology, Xuzhou, Jiangsu 221116, China State Key Laboratory for Geomechanics and Deep Underground Engineering, China University of Mining and Technology, Xuzhou, Jiangsu 221116, China

a r t i c l e

i n f o

Article history: Received 3 March 2016 Received in revised form 31 August 2016 Accepted 2 September 2016 Keywords: Industrial carbon transfer Industrial transfer System-GMM Gravity model China

a b s t r a c t The dynamic trend of regional industrial carbon transfer along with industrial transfers in China is currently a hot topic. To explore this problem, the gravity model has been used to study the spatial distribution of industrial transfer and industrial carbon transfer. The results indicate that both barycenters are moving westward. Based on the STIRPAT model, a system-GMM model was then constructed that introduced the industrial transfer factor and its square to explore temporal changes in regional industrial carbon transfer in the course of industrial transfer. The estimated results revealed that the relationship between industrial transfer and industrial carbon transfer displays an inverted U-shaped pattern. For every 1% increase in industrial transfer, there was a 0.327% increase in industrial carbon transfer before the turning point, but industrial carbon transfer decreased by 0.07% when it passed the “peak”. Because no province had surpassed the turning point, carbon transfer during industrial transfer in China is currently in the growth phase. Moreover, real GDP per capita, industrial structure, and industrial carbon emission intensity promoted carbon emissions reduction in the course of industrial transfer. © 2016 Elsevier Ltd. All rights reserved.

1. Introduction Because regional economic development, resource endowment, industrial base, and industrial specialization vary across China, industrial transfer has been occurring (Wu et al., 2014; Liu et al., 2015). Quantitative studies of industrial transfer have claimed that the central and western regions have received a large amount of resource industry, such as metal smelting and manufacturing, electric power production and supply, and mining (Xiao et al., 2014). However, due to regional differences in environmental factors, inherent differences in environmental pollution externalities, and differences in environment awareness, environmental pollution and its transfer are happening in the process of industry transfer (Wei et al., 2012). In particular, the spatial reconstruction of raw material-intensive and labor-intensive industries will lead to industrial carbon transfer. China overtook the United States as the world’s largest carbon emitter, and its energy consumption has been growing rapidly. Notably, industrial energy consumption accounted for 70% of China’s energy consumption by the end of 2013 (China Statistical

∗ Corresponding author at: China University of Mining and Technology Daxue Rd., Quanshan District, Xuzhou, Jiangsu, China. E-mail address: [email protected] (M. Zhou). http://dx.doi.org/10.1016/j.ecolind.2016.09.002 1470-160X/© 2016 Elsevier Ltd. All rights reserved.

Yearbook, 2014). At present, China is in a unique historical period of industrialization and urbanization. Energy-intensive industries such as the vehicle, real estate, heavy chemical engineering, and electric power sectors, are rapidly advancing. By the end of the 20th Century, China had implemented a series of regional development strategies such as the Western Development, the revitalization of old industrial bases in northeastern China, and the rise of the Central Region. It was inevitable that the carbon emissions caused by industrial transfer from coastal to inland areas would become a hot topic. Therefore, it is crucial to investigate the spatial and temporal dynamic trends of industrial carbon transfer in the course of industrial transfer in China. Moreover, this study can also help to predict future carbon emissions in China. Studies of industrial transfer began in the 1930s, and related research fields have continued to emerge, such as global production networks (Coe et al., 2008; Yeung, 2009), knowledge transfer (Inkpen and Tsang, 2005; Dyer and Hatch, 2006), competitive advantage (Li, 2006; Rugman et al., 2011), gradient transfer (Wu et al., 2014; Liu et al., 2015), and others. Meanwhile, the negative effects of pollution transfer in the course of industry transfer began to attract corresponding attention. Walter and Ugelow (1979) first proposed the pollution haven hypothesis. Since then, the pollution transfer problem has begun to attract attention in the international trade field. Some researchers (Cole, 2004; Mongelli et al., 2006) verified that the pollution haven phenomenon did exist by study-

2

J. Xu et al. / Ecological Indicators 73 (2017) 1–10

ing different countries. However, other researchers (Eskeland and Harrison, 2003; Zhang and Zhou, 2016) saw no robust evidence of the so-called pollution haven hypothesis. All the literature mentioned above was related to carbon transfer. Carbon transfer was initially discussed in the context of international industrial transfer. Many researchers are exploring this issue from different angles. From a global perspective, some researchers (Copeland and Taylo, 2005; Sharif, 2011) have claimed that growth in carbon emissions transfer through international trade was a real problem and that pollution should be monitored by all countries. Furthermore, other studies have examined these issues in various regions. Steen-Olsen et al. (2012) found that the EU was displacing the carbon footprint of its environmental impact to the rest of the world by importing products with embodied impacts; the largest net exporter of impacts in the EU was Poland. Peters and Hertwich (2006) cited the Norwegian case and found that carbon emissions embodied in imports represented 67% of Norway’s domestic emissions and that about one-half of these originated in developing countries. Because China is the world’s second largest economy and the largest carbon-emitting country, carbon transfer in China has become an unavoidable topic. The main point of discussion has focused on whether China, as the world’s factory, has become the world’s “pollution haven”. Lin et al. (2014) claimed that about 21% of export-related Chinese emissions were attributable to Chinato-U.S. exports. Ren et al. (2014) stated that large foreign direct investment (FDI) inflows further aggravated China’s CO2 emissions. Su and Ang (2014) proposed that developed areas in China were a net importer of carbon emissions, whereas developing regions were net exporters of embodied emissions from interregional and international bilateral trade. Nevertheless, some researchers thought that the opposite was true. Li and Lu (2010) proposed that clean industry transferred from developed countries may decrease China’s CO2 emissions. Dietzenbacher et al. (2012) believed that China’s emissions as embodied in its exports were overestimated by more than 60% if processed goods were distinguished from normal exports. In the New Normal, cross-regional industrial transfer has become a significant driving force to achieve industrialization and urbanization. Meanwhile, the accompanying regional transfer of industrial carbon emissions has raised widespread concern among some researchers. However, related quantitative studies have been relatively few. Some researchers have claimed that industrial transfer could account for carbon transfer. In studies of rural regions, Xiao et al. (2014), Liu et al. (2015) and Zhang et al. (2015a) reported that the northwestern, northeastern, and central regions of China have taken on the load of high carbon emissions from other regions. Feng et al. (2012) suggested that not only advanced technologies but polluting sectors transfer result in the “greening” of the muchdeveloped Eastern-Coastal regions. Liang et al. (2016) found that the less developed Central and Western zones suffer from the carbon emission transfers from the Eastern-Coastal region though they can benefit from international exports. Jiang et al. (2015) suggested that carbon transfers in China result in more ‘efficiency losses’ instead of ‘efficiency gain’. Some researchers, however, have held the opposite view. Li and Cao (2013) found that industrial transfer can promote carbon reduction in both core and peripheral areas of the Pan-Yangtze River Delta region. Wei et al. (2012) believed that the western regions were more efficient in carbon reduction and had lower emission reduction costs. Liu et al. (2016) held that the provinces with low industrial carbon emissions can act as an exemplar for the rest due to the high correlation of carbon emissions among regions. The studies discussed above on carbon transfer in the course of industrial transfer relied mainly on static analysis. The inputoutput model (Zhang et al., 2015b; Su et al., 2013; Zhang et al.,

2013) was the most prevalent classical method, although some researchers have used the Logarithmic Mean Divisia Index (LMDI) model (Cheng and Wei, 2013; Rui et al., 2016; Dong et al., 2010), the CGE model (Gerlagh and Kuik, 2014; Lu et al., 2010; Caron, 2012), and spatial econometric measures (Kang et al., 2016; Zuo et al., 2013). After summarizing the relevant articles on carbon transfer, the authors discovered only a few papers that used the dynamic panel data model, especially in articles on carbon transfer in the course of regional industrial transfer in China. Sharma (2011) investigated the determinants of carbon dioxide emissions (CO2) for a global panel consisting of 69 countries using a dynamic panel data model. Ren et al. (2014) studied international trade, FDI, and embodied CO2 emissions in China’s industrial sectors using the GMM method. The purpose of this study is to examine the dynamic effect of industrial carbon transfer during industrial transfer. If the inputoutput or LMDI models are used, time inconsistency problems will result. Studying the dynamic trend of carbon transfer in China using the CGE model may be inefficient, and regional data may be lacking. The spatial econometric model focuses mainly on neighborhood spatial spillover effects. However, Feng et al. (2010) suggested that some technology-intensive industries across the central region have been directly transferred to the west. In addition, compared with the spatial panel model, the system-GMM model is efficient for predicting temporal trends. For these reasons, the authors have used the gravity model and the system-GMM model with provincial panel data for the period from 1995 to 2012 to investigate dynamic spatial and temporal effects respectively. Although some work has been done on this issue, to the best of the authors’ knowledge, there are no other published studies on dynamic industrial carbon transfer during industrial transfer in China using the gravity model and system-GMM model. In addition, carbon transfer in existing analyses is always indirectly measured by comparing regional differences (Feng et al., 2012; Liang et al., 2016; Liu et al., 2016). However, we construct industrial transfer factors to directly measure industrial carbon transfer in the course of industrial transfer, which is another innovation of this research. Moreover, steadystate industrial transfer factors have been verified in this research. Empirical results could also provide a reference and guidance for government in developing appropriate economic and environmental policies and energy strategies. The rest of this paper is organized as follows: Section 2 describes the data and methodology. Section 3 provides the empirical results and Section 4 presents analysis and discuss. Section 5 concludes.

2. Data and methodology 2.1. Data In this research, sample data for 30 provinces (excluding Tibet, Hong Kong, Macao, and Taiwan) for the period from 1995 to 2012 were used. The regional energy balance data used in this paper were provided by the China Statistical Yearbook and the China Energy Statistical Yearbook. The primary energy carbon emission coefficients used were recommended by the Energy Research Institute under the National Development and Reform Commission (2007). In this paper, a total of 17 types of primary energy were selected to calculate carbon emissions. These types included raw coal, cleaned coal, other washed coal, briquettes, coke, coke oven gas, other gas, other coking products, crude oil, gasoline, kerosene, diesel oil, fuel oil, other petroleum products, liquefied petroleum gas, refinery gas, and natural gas. All kinds of primary energy were converted into standard coal. To harmonize the dimensions to a common basis, real GDP per capita and industrial value added were calculated by deflation on the basis of their levels in 1978. Energy consump-

J. Xu et al. / Ecological Indicators 73 (2017) 1–10

tion data were missing for Ningxia Province from 2000 to 2002 and for Hainan Province in 2002. Population data were missing for Chongqing in 1996 and 1997. All the missing data were filled in by interpolation. 2.2. Estimation of industrial carbon emissions

17  

Ek × Fk ×

44 12

If the direction of gravity barycenter change is located along an axis, the latitudinal or longitudinal rate of barycenter change will be zero.   yt − yt−1 , (xt − xt−1 ) represent the latitudinal and longitudinal directions of barycenter change from the (t-1)-st year to the t-th y −y year respectively. Therefore, xt −xt−1 denotes the relative change in t

With reference to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories (2006), carbon emissions can be calculated as follows: CC =

3

 (1)

k

t−1

latitude and longitude from the (t-1)-st year to the t-th year. The ratio of the rate of change in latitude and longitude measures the equilibrium of certain properties of the spatial distribution. (2) The direction of movement of the gravity barycenter can be expressed as:  =  t+1 −  t =

y − yt−1 l + arctg t xt − xt−1 2

where CC denotes industrial carbon emissions in units of 104 t, Ek is the k-th kind of industrial primary energy in 104 t of standard coal, and Fk denotes the carbon emission coefficient of the k-th kind of is the ratio of the molecular weights primary energy. The factor 44 12 of CO2 and C. The primary-energy carbon emission coefficients used in this paper are presented in Table 1 (in tonnes C/tonne standard coal).

where ␪ denotes the direction of barycenter movement, due east is et to 0◦ , the counterclockwise direction is positive, (0◦ , 90◦ ) is the first quadrant and represents the northeast direction, (90◦ , 180◦ ) is the second quadrant and represents the northwest direction, (−180◦ , −90◦ ) is the third quadrant and represents the southwest direction, and (−90◦ , 0◦ ) is the fourth quadrant representing the southeast direction.

2.3. Barycenter of the gravity model

2.4. STIRPAT extension model

The gravity model was used to measure the migration paths year by year of industrial transfer and industrial carbon transfer. A migration path can be expressed as:

Ehrlich and Holdren (1971) constructed the earliest IPAT equation to specify the impact of population growth on environment change: I = PAT, where I denotes environmental impact, P is population, A is affluence, and T is technology level. However, the IPAT model can easily cause proportional change problems if a control variable is changed while keeping other factors stationary. To overcome this limitation, Dietz and Rosa (1994) constructed the STIRPAT model, which can be widely used to study the driving factors of non-unitary elastic environments. The advantages of the STIRPAT model are reflected not only in the respective coefficients provided for parameter estimation, but also in the fact that three drivers can be thoroughly decomposed to enable a more complete quantitative analysis. Due to its flexibility, the STIRPAT model is used in this study to examine industrial carbon transfer in the course of industrial transfer in China. The STIRPAT model can be summarized as:

Xt =

30 

30 

pi Mit /

i=1

Mit , Yt =

i=1

30  i=1

30 

qi Mit /

Mit

(2)

i=1

where (pi , qi ) are the geographical coordinates of the i-th provincial capital city in China, i is the province, t is the year, and Mit denotes industrial output or industrial carbon emissions, depending on context. According to the theory of the gravity barycenter, the center of industrial output value and the center of industrial carbon emissions for the whole of China are the points that will achieve a relatively balanced impact of their respective forces. Because provincial geographic coordinates remain unchanged, industrial output and industrial carbon emissions are the key variables that can be expressed as spatial movement. The spatial trajectory can be represented as direction and movement, which indicate the direction of spatial aggregation and the degree of equilibrium respectively. Because changes of centers of industrial output value and carbon emissions illustrate the spatial movement, we regard them as the movement of centers of industrial transfer and industrial carbon transfer in this study, respectively. (1) Movement of the gravity barycenter:



D = Dt − Dt−1 = w ×



(xt − xt−1 )2 + y t − y t−1

2

Constant coefficient: w = 40, 000 km/360◦ = 111.111 km/1◦ where D represents the distance that the gravity barycenter moves from the (t-1)-st year to the t-th year and w is a constant that converts spherical distance as measured by spatial geographic coordinates to a plane distance. Latitude/longitude rates of barycenter change: y −y z = | xt −xt−1 | t

t−1

If z < 1 the longitudinal rate of barycenter change is faster than the latitudinal rate of barycenter change. If z > 1 the latitudinal rate of barycenter change is faster than the longitudinal rate of barycenter change. If z = 1 the latitudinal and longitudinal rates of barycenter change are equal.

I = aPb Ac Td e

(3)

where I, P, A, T express the same meaning as in the IPAT model. The model varies the estimated parameters a, b, c, and d and the random disturbance e. In empirical studies, Eq. (3) can be converted to logarithmic form: lnI = lna + blnP + clnA + dlnT + lne

(4)

where b, c, and d can be thought of as elasticity coefficients of population, affluence, and technology. To calculate industrial carbon transfer in the course of regional industrial transfer in China, considering the specific situation and learning from past research experience (Fan et al., 2006; Wang et al., 2012; Li et al., 2015), the STIRPAT model was expanded by incorporating urbanization, industrial structure, industrial carbon emission intensity, the first-order industry transfer variable, and the square of this variable into the model. The improved model can be written as follows: lnCC = ␤0 + ␤2 lnPU + ␤3 lnA + ␤4 lnTIS + ␤5 lnTCI +␤6 lnIT + ␤7 lnIT2 + ␧

(5)

where CC represents industrial carbon emissions in units of 104 t; PU denotes urbanization, which is expressed by non-agricultural population in percent; A denotes affluence, which is expressed as real GDP per capita in yuan; and technology is decomposed into

4

J. Xu et al. / Ecological Indicators 73 (2017) 1–10

Table 1 Primary energy carbon emission coefficients. Primary energy type

Fk

Primary energy type

Fk

Raw coal Cleaned coal Other washed coal Briquettes Coke Coke oven gas Other gas Other coking products Crude oil

0.7559 0.7559 0.7476 0.7476 0.855 0.3548 0.3548 0.6449 0.5857

Gasoline Kerosene Diesel oil Fuel oil Other petroleum products Liquefied petroleum gas Refinery gas Natural gas

0.5538 0.5714 0.5921 0.6185 0.5857 0.5042 0.4602 0.4483

Fig. 1. Direction and movement of the gravity barycenters of industrial transfer and industrial carbon transfer.

TIS and TCI . TIS denotes industrial structure adjustment. Because the secondary and tertiary economic sectors account for a large proportion of the whole economy and the agricultural share is small, this paper uses the ratio of tertiary and secondary industry in percent as the measure of industrial structural adjustment. The ratio will increase when tertiary industry accounts for more proportion, which means the industrial structure adjustment is more reasonable and optimized. TCI represents industrial carbon emission intensity, expressed by energy consumption per unit of value added by industry, in units of 104 t of standard coal. Following Cheng and Wei (2013), the first-order industrial transfer factor was represented by the provincial share of gross industrial output value and its fluctuation trends. The national gross industrial output value (100) was taken as the base, the provincial share was calculated as a percentage, and the dynamic trends of the shares of provincial gross industrial output value were examined from 1995 to 2012. IT is the first-order industrial transfer factor, expressed as a percentage. In fact, the impacts of industrial transfer on industrial carbon emissions are uncertainty. On the one hand, industrial carbon efficiency may be improved in the less-developed economic zone though knowledge transfers. On the other hand, industrial carbon emission increase may occur when polluting sectors are relocated in the poorer regions. To accommodate the nonlinear effect of industry transfer on the time series, the model also introduced the squared factor IT2 to represent the

uncertainty of industrial carbon emissions in the course of industrial transfer. Because IT and IT2 denotes the industrial transfer, we regard lnCC as industrial carbon transfer in this situation, which presents the industrial carbon emissions caused by industrial transfer. 2.5. System-GMM Because economic behavior exhibits some continuity, it is essential to introduce lagged dependent variables to ensure that the model coefficients are calculated consistently and effectively. In this paper, industrial carbon emissions are strongly related to past values of the dependent variable. However, the dynamic lag term of the explained variable is related to the individual effects of random disturbance in the model, resulting easily in an endogeneity problem. Because the generalized method of moments (GMM) uses a set of instrumental variables to compensate for deviations caused by endogenous effects, GMM was used in this research to estimate the dynamic panel data model, an approach which also accommodates the lagged level of carbon emissions (Omri and Nguyen, 2014). There are two types of GMM estimators: difference and system. The core concept of difference GMM is to eliminate individual fixed effects by taking the difference with the original model and considering the residual structure (Arellano and Bond, 1991). The core concept of system GMM is to add a hypothesis that the instrumental variable difference is endogenous as an individual effect

J. Xu et al. / Ecological Indicators 73 (2017) 1–10

5

Fig. 2. Movement tendencies of the gravity barycenters of industrial transfer from 1995 to 2012.

that can be used as an instrumental variable (Blundell and Bond, 1998). Compared with difference GMM, system GMM is advantageous because it helps solve the endogeneity problem caused by potential correlation between the independent variable and the error term in dynamic panel data models (Topcu, 2013). It can also overcome the effects of the omission of dynamics in static panel data models due to ignorance of the impacts of lagged values on the dependent variable (Bond, 2002). The system-GMM model estimation as adopted is following Sharma (2011), Ren et al. (2014) and Li et al. (2016). Accordingly, the empirical model can be constructed as follows: IS CI lnCCit = ␤0 + ␤1 lnCCit−1 +ˇ2 lnPU it + ␤3 lnAit + ␤4 lnTit + ␤5 lnTit

+␤6 lnITit + ␤7 lnIT2it + ␾t + εit

(6)

where the variables have the same meaning as in Eq. (5). The increase variable CCit−1 represents the carbon emission lag, t denotes country-specific effects, and εt−1 is an error term. 3. Empirical results 3.1. Results for gravity barycenter trajectories As shown in Fig. 1, from the perspective of the barycenter distribution, both barycenters moved mainly to the west. The range of longitudinal motion of the industrial transfer barycenter was (115.038, 115.848) East, and its range of latitudinal motion was (32.771, 32.972) North. The range of longitudinal motion of the industrial carbon transfer barycenter was (113.936, 114.955) East, and its range of latitudinal motion was (33.823, 34.179) North. The speed of industrial transfer barycenter movement in the east-west direction was relatively higher than in the north-south direction for 8 years (z < 1). The speed of industrial carbon transfer barycenter movement in the east-west direction was relatively higher than in the north-south direction for 12 years (z < 1). These results suggest that the change in both barycenters in the longitudinal direction was more obvious than in the latitudinal direction. As shown in Fig. 2, from the viewpoint of direction of variation, the industrial transfer barycenter, though initially moving to the south from 1995 to 2003, generally moved to the northwest after 2003. With a clockwise rotation of 47% and a counterclockwise rotation of 53%, there appeared to be a massive westward movement. This represents a gradual movement from Fuyang City in Henan Province to Zhumadian City in Anhui Province (1995–2012). Compared to the gravity trajectory of industrial transfer, the carbon

emissions trajectory was more uneven (Fig. 3). Except for movement to the northeast in 1999, 2003, 2006, and 2009 and to the south in 2003 and 2004, there was a clockwise rotation of 47% and a counterclockwise rotation of 53%. The barycenter of industrial carbon transfer still moved westward on the whole. During study period, the barycenter of industrial carbon transfer moves from Zhoukou City to Xuchang City in Henan Province (1995–2012). From the perspective of distance of movement, the total distance traversed by the industrial transfer barycenter was 153.061 km, and its linear distance was 79.827 km. The total distance traversed by the industrial carbon transfer barycenter was 361.215 km, and the linear distance was 113.247 km. The total distance and linear distance traversed by the industrial carbon transfer barycenter were 2.36 times and 1.419 times longer than the same distances for the industrial transfer barycenter. Compared to the industrial transfer barycenter, the industrial carbon transfer barycenter fluctuated significantly, and its spatial distribution was relatively unstable. The results indicate that the regional imbalance of industrial carbon transfer is more obvious than that of industrial transfer. As shown in Fig. 4, regions with rapid industrialization mainly exist in the much-developed Eastern-Coastal economic regions in 1995. Industrial output by top three provinces (Guangdong, Jiangsu, Shandong) produced 2112.55 × 108 yuan, which account for 30.46% of the national total. They were followed by Jiangsu, Shanghai, Henan, Ningxia, Hebei, Heilongjiang, and each province’s industrial output exceeded 250 × 108 yuan. The rest 19 provinces’ total industrial output is relatively small, accounting for 36.78% of the national total. In 2012, the national industrial output is developing rapidly, which is 6.13 times higher than that in 1995. Besides Xinjiang, Gansu, Qinghai, Ningxia, Guizhou and Hainan, industrial output by the other provinces exceeds 500 × 108 yuan. From the perspective of the annual growth rate of industrial output, Neimenggu was the fastest, followed by Jiangxi (15.00%), Shaanxi (14.80%), Qinghai (14.72%), Anhui (13.20%), Shanxi (13.18%), Hubei (13.15%), Ningxia (13.1%) and Xinjiang (13.07%). It is clearly that there is a rapid industrial development in Central and Western regions. The industrial transfer showed an obvious westward movement from EasternCoastal regions to Central and Western regions. As shown in Fig. 5, the major coal consumption regions were the much-developed Eastern-Coastal and Central regions in 1995. There were thirteen provinces, and coal consumption of each province exceeded more than 5000 × 104 t. The coal consumption in the thirteen provinces was 99,516.37 × 104 t, which account for 69.35% of the national total. There were third provinces and each province consumed more than 10,000 × 104 t in 1995: Guangdong,

6

J. Xu et al. / Ecological Indicators 73 (2017) 1–10

Fig. 3. Movement tendencies of the gravity barycenters of industrial carbon transfer from 1995 to 2012.

Fig. 4. Spatial pattern of industrial transfer in (a) 1995, (b) 2000, (c) 2005, and (d) 2012.

Jiangsu, and Hebei. In 2005, there are nineteen provinces and each province consumed more than 5000 × 104 t. In 2012, the national industrial carbon emission is 2.67 times higher than that in 1995. Besides Qinghai and Ningxia, all the provinces consumed more than 5000 × 104 t. There were five provinces and coal consumption of

each province exceeded more than 20,000 × 104 t: Hebei, Liaoning, Jiangsu, Shandong, and Hubei. The annual growth rate in coal consumption of Ningxia has been the fastest, followed by Hainan (20.71%), Qinghai (20.14%), Yunnan (12.72%), and Shaanxi (12.53%). In contrast, the annual growth rate in coal consumption of Beijing

J. Xu et al. / Ecological Indicators 73 (2017) 1–10

7

Fig. 5. Spatial pattern of industrial carbon transfer in (a) 1995, (b) 2000, (c) 2005, and (d) 2012.

is the slowest with a negative growth among the 30 provinces over the study period. In general, the coal consumption in Central and Western regions is obviously increasing, and the trend of industrial carbon transfer moves from east to west over time. 3.2. Results of system-GMM model The Hansen J-test is used to explore the validity of the full set of over-identification restrictions in GMM estimates. The validity of the instruments in all models is confirmed by the test. ArellanoBond AR (2) is a test for autocorrelation of differences and is used to explore the validity of model specifications. Because the p-value of Arellano-Bond AR (2) is insignificant, all the models can be considered reliable, and the estimated result is reasonable. The “steady-state” assumption made by Roodman (2009) requires a kind of steady state in the sense of deviations from long-term values that are not systematically correlated with the fixed effect. Because the coefficients of the lagged carbon emissions in all models lie between 0.259 and 0.280, the assumptions are robust and can be used to check the validity of instrumental variables in the system-GMM model. As shown in Table 2, all the regression results are statistically significant. Model 1 is the basic model used to investigate direct driving factors on industrial carbon emissions without considering industrial transfer factors. Based on Model 1, the first- and second-

order industrial transfer factors were progressively introduced to explore the influence of industrial transfer on industrial carbon transfer using Models 2 and 3. In the basic model, the estimated results were statistically significant at the 1% level. The findings indicate that lagged values of carbon emissions, urbanization, and industrial carbon emission intensity have a positive effect on carbon emissions. Typically, a 1% increase in the lagged values of carbon emissions and urbanization raises carbon emissions by 0.259% and 0.374%, respectively. Because large-scale energy consumption occurs in urban areas, urbanization will inevitably increase carbon emissions. Industrial carbon emissions intensity was one of the technical indicators used, and its coefficient was 0.749, which is consistent with theoretical expectations. Carbon emissions were affected negatively by real GDP per capita and industrial structure. Effectively, 1% increases in real GDP per capita and industrial structure decreased carbon emissions by about 0.74% and 0.826%, respectively. The estimator shows that without considering industrial transfer, economic development and industrial structure will produce carbon emissions reduction. The reason for this may be that economic development promotes energy efficiency and makes the industrial structure more rational, which is favorable to reducing industrial carbon emissions. Based on Model 1, Model 2 adds the first-order industrial transfer factor. Its results were similar to those from Model 1: except

8

J. Xu et al. / Ecological Indicators 73 (2017) 1–10

Table 2 System-GMM panel estimation regression results for China. Variable

Model 1

Model 2

Model 3

LNCC(-1) LNPGDP LNURB LNIS LNEI LNINT1 LNINT2 AR(1)(p-value) AR(2)(p-value) Hansen J-test Hansen J-test (p-value) Turning point Instrument Observation

0.259*** (0.019) −0.740*** (0.051) 0.374*** (0.145) −0.826*** (0.046) 0.749*** (0.021)

0.280*** (0.01) −0.811*** (0.023) 0.267** (0.035) −0.577*** (0.024) 0.754*** (0.023) 0.327*** (0.031)

(0.0014) (0.2684) 27.3469 (0.3388)

(0.0011) (0.0883) 29.5271 (0.2009)

30 480

30 480

0.273*** (0.017) −0.744*** (0.042) 0.365*** (0.117) −0.706*** (0.046) 0.772*** (0.023) 0.277*** (0.092) −0.070*** (0.023) (0.0001) (0.2033) 27.8538 (0.2214) 5.638 30 480

Note: Standard errors in parentheses: **p < 0.05, ***p < 0.01.

for urbanization (significant at the 5% level), all other factors were significant at the 1% level, and the direction never changed. The elasticity coefficients of the lagged value, urbanization, and industrial carbon emission intensity were 0.28, 0.267, and 0.754 respectively. Compared with Model 1, the effect of lagged value on carbon emissions increased; however, the effect of urbanization on carbon emissions declined. The results suggest that the lagged value will produce more carbon emissions than urbanization during industrial transfer. The coefficients of real per-capita GDP and industrial structure on carbon emissions were −0.811% and −0.577% respectively. Taking industrial transfer into account, the effect of real GDP per capita slightly increased, but the effect of industrial structure slightly decreased. The results revealed that industrial transfer promotes economic growth by means of the lowcarbon-technology spillover effect, which directly leads to carbon emissions reduction. Industrial transfer also increases the proportion of secondary industry, which limits the effect of carbon emissions reduction. Based on Model 2, the second-order industry transfer factor was first introduced into Model 3 to determine the indefinite influence of industrial transfer on carbon emissions. All the variables were significant at the 1% level, and the coefficient direction never changed. The elasticity coefficients of the lagged value and of urbanization were 0.273 and 0.365 respectively, which implies that urbanization will produce markedly more industrial carbon emissions in the long term. The elasticity coefficient of industrial carbon emission intensity was 0.772, an increase of 0.018 points compared with Model 2. This result suggests that technology transfer accompanying industrial transfer brings about a technology imitation effect, which significantly reduces carbon emissions. Real per-capita GDP and industrial structure still had a negative effect on carbon emissions. In particular, a 1% increase in real percapita GDP and industrial structure decreased carbon emissions by approximately −0.744 and −0.706 respectively. This shows that in the course of industrial transfer, economic development brought about obvious carbon emission reductions at the beginning. However, when industrial transfer is over the turning point, industrial structural optimization will produce even more carbon emission reductions. The strongest evidence supporting the hypothesis that industrial transfer has a dynamic impact on carbon emissions comes from Models 2 and 3. The industrial transfer variable and its square caused significant changes in carbon emissions at the 1% level. Effectively, a 1% increase in the industry transfer factor raised carbon emissions by about 0.327% in Model 2. This reveals that industrial transfer simultaneously brings about substantial carbon emissions transfer. What is more notable is that the amount of industrial carbon transfer was greater than the carbon emissions

caused by the lagged value and by urbanization. In Model 3, carbon emissions were still affected positively and significantly by the first-order industry transfer factor. Effectively, a 1% increase in the first-order industry transfer factor increased carbon emissions by approximately 0.277%. However, the second-order factor had a negative effect on carbon emissions and on the coefficient’s magnitude of −0.070. The empirical result indicates that the relationship between industrial transfer and industrial carbon transfer shows an inverted U-shaped pattern. This evidence clearly points out that industrial transfer brings about a substantial amount of industrial carbon transfer before the turning point; however, this industrial carbon transfer will decrease when industrial transfer passes the “peak”. To clarify the nonlinear relationship between industrial carbon transfer and industrial transfer, the turning point for industrial transfer1 was calculated to determine whether any province in the sample had passed the turning point. A comparison of the results showed that no province had passed the turning point. Carbon transfer during industrial transfer is currently in the growth phase in the entire country. 4. Analysis and discussion Several meaningful conclusions can be derived from the empirical results of the gravity barycenter model and the dynamic system-GMM panel model. Because the objective of this paper was to investigate the dynamic effects of industrial carbon transfer in the course of industrial transfer, this discussion will concentrate mainly on the results of system-GMM models 2 and 3 with industrial transfer factors. First, the results clearly show that environmental degradation is statistically significant in all models. In particular, industrial carbon transfer caused by industrial transfer creates pressure on individuals to improve environmental awareness and on government to make environmental regulations and policies to reduce or eliminate carbon emissions. Second, for the gravity model, the two trajectories are not consistent; in other words, the findings reveal that industrial transfer and industrial carbon transfer are not completely synchronized. The direction of variation of both barycenters is basically the same, and there appears to be a westward movement in both cases. The industrial transfer barycenter has shown accelerated movement westward, which indicates that China is gradually undertaking industrial transfer from coastal to inland regions. Judging from the gravity model, compared with the westward movement of industrial transfer, the path followed by industrial carbon transfer is less direct. However, on the whole, it still points towards the central and

1

Turning point for industrial transfer is calculated as 0.902/(2 × 0.08) = 5.638.

J. Xu et al. / Ecological Indicators 73 (2017) 1–10

western regions. Hence, there is some spatial correlation between industrial transfer and industrial carbon transfer. Third, a dynamic component of industrial carbon transfer is present during industrial transfer. Specifically, industrial transfer increases industrial carbon transfer before the turning point, but reduces carbon emissions over the turning point. The reason for this may be that industrial transfer leads to energy factor mobility and changes the regional industrial structure firstly. In particular, relocation of resource-based industries markedly changes regional energy consumption. Industrial carbon emissions are bound to increase at the beginning of an industrial transfer. However, when industrial transfer passes the turning point, the effects of industrial transfer can optimize resource allocation and produce technology spillovers, which extend the economic structure effect over time. Moreover, enterprising industries are more likely to make intensive use of capital and technology to reduce carbon emissions. Therefore, industrial carbon transfer is not an inevitable result of industrial transfer. More important factors are the selection of industries to be transferred, industrial structural adjustment, and development of environmentally friendly technology. Fourth, other factors could increase carbon emissions. Urbanization and the lagged value had a significant positive influence on carbon emissions in Models 1–3. The coefficient of urbanization ranged from 0.267 to 0.365 in Models 2 and 3. The results provide evidence that urbanization gradually increases carbon emissions during industrial transfer, which is consistent with that of Wang et al. (2012) for Beijing and Fan et al. (2006) for China. The reason may be that urbanization during industrial transfer not only increases energy consumption directly, but also changes production modes and lifestyles, which increase carbon emissions indirectly. The effect of the lagged value on carbon emissions decreased slightly in the process of industrial transfer. However, it still produced a certain amount of carbon emissions. This indicates that carbon emissions may have the lock-on effect, which means economic development highly depends on the energy usage. Fifth, other factors could reduce carbon emissions. Real percapita GDP and industrial structure were found to affect carbon emissions negatively in Models 1–3. In particularly, during industrial transfer, the effect of industrial structure gradually increased. The results confirm the finding of Li et al. (2015). Although the effect of real GDP per capita decreased slightly over time, it still had a strong negative influence. The results are different from the finding of Wang et al. (2012) regarding the positive impact of GDP on carbon emissions. The reason for this may be that industrial transfer narrows the relative disparity in GDP among the eastern, middle, and western regions (Wu et al., 2014). During industrial transfer, intensifying development accelerates structural transformation and optimizes resource allocation, which will gradually incline towards renewable energy instead of traditional energy sources. Industrial carbon emission intensity was found to affect carbon emissions positively in Models 1–3. A study by Li et al. (2015) has the same conclusions. Its effect increased gradually during the industrial transfer process. These results prove that lowcarbon technologies should be a prominent long-term solution to decreasing carbon emissions, rather than simply relying on controlling the scale of carbon emissions.

5. Conclusions The objective of this study was to investigate the dynamic impact of industrial transfer on carbon emissions in China. First, from a spatial distribution perspective, the gravity model was used to reveal the gravity barycenters of industrial transfer and industrial carbon transfer and their westward movement. To explore further the dynamic temporal relationship between the

9

two barycenters, industrial transfer factors were introduced for the first time into a system-GMM for a set of thirty provinces in China from 1995 to 2012. Estimated results showed that there was an inverted U-shaped pattern between industrial transfer and industrial carbon transfer. For every 1% increase in industrial transfer, there was a 0.327% increase in industrial carbon transfer before the turning point. However, carbon transfer decreased by 0.07% when it surpassed the peak of its U-shaped pattern. This indicates that industrial transfer is not a prerequisite for carbon pollution. Estimated results also indicate that real per-capita GDP, industrial structure, and industrial carbon emission intensity are driving forces for carbon reduction. Conversely, the lagged value and urbanization are major forces causing carbon emissions. Finally, the industry transfer factors were tested against the steady state, which can be used as a dynamic index to describe economic phenomena. Although industrial transfer will reduce carbon emissions after the turning point, there is no province in this study that has surpassed the peak. Currently, China is faced with significant pressure to reduce “industrial carbon transfer”, and carbon emissions reduction has become a tough external constraint on industrial transfer. Based on the analysis described above, the following policy recommendations can be made. (1) Based on regional differentiation of carbon emissions, policy should aim to promote low-carbon industrialization and optimize the industrial structure. In particular, it is crucial to support energy substitution and renewable energy sources such as solar and wind. (2) Economic growth and environmentally friendly technology are the most effective measures to reduce carbon emissions. It is essential to encourage innovation capacity in all industrial sectors to obtain greater energy efficiency. Governments and enterprises involved in industry transfer should strengthen strategic investment in low- and zero-emissions technologies. (3) Environment regulations should be formulated to achieve effective control over regional industrial carbon emissions. Not only can such regulations prevent the entry of highly energyconsuming enterprises in less-developed areas, but they can also encourage the upgrade of industrial structure to achieve an economic virtuous cycle. Acknowledgements We gratefully acknowledge financial support from the Jiangsu Provincial Graduate Innovation Project (KYZZ15 0371) and the Humanities and Social Science Fund of the Ministry of Education (15YJA630106). References Arellano, M., Bond, S.R., 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev. Econ. Stud. 58, 277–297. Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. J. Econ. 87, 115–143. Bond, S., 2002. Dynamic Panel Models: A Guide to Micro Data Methods and Practice. Working Paper. CEMMAP (Centre for Microdata Methods and Practice) Institute for Fiscal Studies, Department of Economics, UCL. Caron, J., 2012. Estimating carbon leakage and the efficiency of border adjustments in general equilibrium- does sectoral aggregation matter? Energy Econ. 34 (12), S111–S126. Cheng, A.H., Wei, H.K., 2013. Target design on carbon reduction of promoting regional industrial transfer orderly and coordinated development. China Popul. Res. Environ. 23 (1), 55–62 (in Chinese). 2014. China Statistical Yearbook (CSY). National Bureau of Statistics of China, China. Coe, N.M., Dicken, P., Hess, M., 2008. Global production networks: realizing the potential. J. Econ. Geogr. 8 (3), 271–295. Cole, M.A., 2004. Trade, the pollution haven hypothesis and the environmental Kuznets curve: examining the linkages. Ecol. Econ. 48 (1), 71–81. Copeland, B.R., Taylo, M.S., 2005. Free trade and global warming: a trade theory view of the Kyoto protocol. J. Environ. Econ. Manage. 49 (2), 205–234. Dietz, T., Rosa, E.A., 1994. Rethinking the environmental impacts of population, affluence, and technology. Hum. Ecol. Rev. 1, 277–300.

10

J. Xu et al. / Ecological Indicators 73 (2017) 1–10

Dietzenbacher, E., Pei, J.S., Yang, C.H., 2012. Trade, production fragmentation, and China’s carbondioxide emissions. J. Environ. Econ. Manage. 64 (1), 88–101. Dong, Y.L., Ishikawa, M., Liu, X.B., Wang, C., 2010. An analysis of the driving forces of missions embodied in China-Japan trade. Energy Policy 38 (11), 6784–6792. Dyer, J.H., Hatch, N.W., 2006. Relation-specific capabilities and barriers to knowledge transfer: creating advantage through network relationships. Strat. Manage. J. 27 (8), 701–719. Ehrlich, P.R., Holdren, J.P., 1971. Impact of population growth. Science 171 (3977), 1212–1217. Energy Research Institute under the National Development and Reform Commission, 2007. National Greenhouse Gas Inventory of the People’s Republic of China. Chinese Environmental Science Press, Beijing. Eskeland, G.S., Harrison, A.E., 2003. Moving to greener pastures? Multinational and the pollution haven hypothesis. J. Dev. Econ. 70 (1), 1–23. Fan, Y., Liu, L.C., Wu, G., Wei, Y.M., 2006. Analyzing impact factors of CO2 emissions using the STIRPAT model. Environ. Impact Assess. Rev. 26, 377–395. Feng, G.F., Liu, Z.Y., Jang, W.D., 2010. An analysis on the trends features and causes of industrial transfer among China’s eastern, central and western regions. Mod. Econ. Sci. 32 (2), 1–10 (in Chinese). Feng, K.S., Siu, Y.L., Guan, D.B., Hubacek, K., 2012. Analyzing drivers of regional carbon dioxide emissions for China. J. Ind. Ecol. 16 (4), 600–611. Gerlagh, R., Kuik, O., 2014. Spill or leak? Carbon leakage with international technology spillovers: a CGE analysis. Energy Econ. 45 (9), 381–388. Intergovernmental Panel on Climate Change (IPCC), 2006. IPCC Guidelines for National Greenhouse Gas Inventories. Intergovernmental Panel on Climate Change, IPCC, Paris. Inkpen, A.C., Tsang, E.W.K., 2005. Social capital, networks, and knowledge transfer. Acad. Manage. Rev. 30 (1), 146–165. Jiang, Y.K., Cai, W.J., Wan, L.Y., Wang, C., 2015. An index decomposition analysis of China’s interregional embodied carbon flows. J. Clean. Prod. 88, 289–296. Kang, Y.Q., Zhao, T., Yang, Y.Y., 2016. Environmental Kuznets curve for CO2 emissions in China: a spatial panel data approach. Ecol. Indic. 63 (4), 231–239. Li, P.X., Cao, Y.H., 2013. Spatial the temporal changes of industrial carbon emissions under regional industrial transfer: the case of Pan-Yangtze river delta. Adv. Earth Sci. 28 (8), 939–947 (in Chinese). Li, X.P., Lu, X.X., 2010. International trade, pollution industry transfer and CO2 emissions in Chinese industries. China Econ. 3, 89–99. Li, B., Liu, X.J., Li, Z.H., 2015. Using the STIRPAT model to explore the factors driving regional CO2 emissions A case of Tianjin, China. Nat. Hazards 76, 1667–1685. Li, T.T., Wang, Y., Zhao, D.T., 2016. Environmental Kuznets curve in China: new evidence from dynamic panel analysis. Energy Policy 91, 138–147. Li, L., 2006. Relationship learning at trade shows: its antecedents and consequences. Ind. Mark. Manage. 35 (2), 166–177. Liang, H.W., Dong, L., Luo, X., Ren, J.Z., Zhang, N., Gao, Z.Q., Dou, Y., 2016. Balancing regional industrial development: analysis on regional disparity of China’s industrial emissions and policy implications. J. Clean. Prod. 126, 223–235. Lin, J.T., Pan, D., Davis, S.J., 2014. China’s international trade and air pollution in the United States. Proc. Natl. Acad. Sci. U. S. A. 111 (5), 1736–1741. Liu, L.C., Liang, Q.M., Wang, Q., 2015. Accounting for China’s regional carbon emissions in 2002 and 2007: production-based versus consumption-based principles. J. Clean. Prod. 103, 384–392. Liu, Y., Xiao, H.W., Zhang, N., 2016. Industrial carbon emissions of China’s regions: a spatial econometric analysis. Sustainability 8 (3). Lu, C.Y., Zhang, X.L., He, J.K., 2010. A CGE analysis to study the impacts of energy investment on economic growth and carbon dioxide emissions: a case of shaanxi province in western China. Energy 35 (11), 4319–4327. Mongelli, I., Tassielli, G., Notarnicola, B., 2006. Global warming agreements, international trade and energy/carbon embodiments: an input-output approach to the Italian case. Energy Policy 34 (1), 88–100. Omri, A., Nguyen, K.D., 2014. On the determinants of renewable energy consumption: international evidence. Energy 72, 554–560. Peters, G.P., Hertwich, E.G., 2006. Pollution embodied in trade: the Norwegian case. Glob. Environ. Chang.-Hum. Policy Dimens. 16 (4), 379–387.

Ren, S.G., Yuan, B.L., Ma, X., Chen, X.H., 2014. International trade, FDI (foreign direct investment) and embodied CO2 emissions: a case study of Chinas industrial sectors. China Econ. Rev. 28 (3), 123–134. Roodman, D.M., 2009. How to do xtabond2: an introduction to difference and system GMM in stata. Stata J. 9 (1), 86–136. Rugman, A.M., Verbeke, A., Nguyen, Q.T.K., 2011. Fifty years of international business theory and beyond. Manage. Int. Rev. 51 (6), 755–786. Rui, W., Yong, G., Dong, H.J., Tsuyoshi, F.J., Xu, T., 2016. Changes of CO2 emissions embodied in China-Japan trade: drivers and implications. J. Clean. Prod. 112 (1), 4151–4158. Sharif, H., 2011. Panel estimation for CO2 emissions, energy consumption, economic growth, trade openness and urbanization of newly industrialized countries. Energy Policy 39 (11), 6991–6999. Sharma, S.S., 2011. Determinants of carbon dioxide emissions: empirical evidence from 69 countries. Appl. Energy 88 (1), 376–382. Steen-Olsen, K., Weinzettel, J., Cranston, G., Ercin, A.E., Hertwich, E.G., 2012. Carbon, land, and water footprint accounts for the European Union: consumption, production, and displacements through international trade. Environ. Sci. Technol. 46 (20), 10883–10891. Su, B., Ang, B.W., 2014. Input-output analysis of CO2 emissions embodied in trade: a multi-region model for China. Appl. Energy 114 (SI), 377–384. Su, B., Ang, B.W., Low, M., 2013. Input-output analysis of CO2 emissions embodied in trade and the driving forces: processing and normal exports. Ecol. Econ. 88 (4), 119–125. Topcu, S.C.M., 2013. The nexus between financial development and energy consumption in the EU: a dynamic panel data analysis. Energy Econ. 39, 81–88. Walter, I., Ugelow, J., 1979. Environmental policies in developing countries. Ambio 8, 102–109. Wang, Z.H., Yin, F.C., Zhang, Y.X., Zhang, X., 2012. An empirical research on the influencing factors of regional CO2 emission: evidence from Beijing city, China. Appl. Energy 100 (4), 277–284. Wei, C., Ni, J.L., Du, L.M., 2012. Regional allocation of carbon dioxide abatement in China. China Econ. Rev. 23 (3), 552–565. Wu, A.Z., Li, G.P., Sun, T.S., Liang, Y.S., 2014. Effects of industrial relocation on Chinese regional economic growth disparities: based on system dynamics modeling. Chin. Geogr. Sci. 24 (6), 706–716. Xiao, Y.F., Wan, Z.J., Liu, H.G., 2014. An empirical study of carbon emission transfer and carbon leakage in regional industrial transfer in China: analysis based on inter-regional input-output model in 2002 and 2007. J. Finance Econ. 40 (2), 75–84 (in Chinese). Yeung, H.W.C., 2009. Regional development and the competitive dynamics of global production networks: an east Asian perspective. Reg. Stud. 43 (3), 325–351. Zhang, C.G., Zhou, X.X., 2016. Does foreign direct investment lead to lower CO2 emissions? Evidence from a regional analysis in China. Renew. Sustain. Energy Rev. 58, 943–951. Zhang, B., Chen, Z.M., Xia, X.H., Xu, X.Y., Chen, Y.B., 2013. The impact of domestic trade on China’s regional energy uses: a multi-regional Input-output modeling. Energy Policy 63 (12), 1169–1181. Zhang, B., Qiao, H., Chen, Z.M., Chen, B., 2015a. Growth in embodied energy transfers via China’s domestic trade: evidence from multi-regional input-output analysis. Appl. Energy (in press, corrected proof, available online 1 October). Zhang, Y., Zheng, H.M., Yang, Z.F., Su, M.R., Liu, G.Y., Li, Y.X., 2015b. Multi-regional input-output model and ecological network analysis for regional embodied energy accounting in China. Energy Policy 86, 651–663. Zuo, C.C., Birkin, M., Clarke, G., McEvoy, F., Bloodworth, A., 2013. Modelling the transportation of primary aggregates in England and Wales: exploring initiatives to reduce CO2 emissions. Land Use Policy 34 (12), 112–124.