The Effect of Economic Growth, Urbanization, and ... - ACS Publications

Oct 6, 2016 - In this study, we combined panel data and econometric methods to ... Effects of land use and landscape pattern on PM 2.5 in Yangtze Rive...
0 downloads 0 Views 2MB Size
Subscriber access provided by CORNELL UNIVERSITY LIBRARY

Policy Analysis

The effect of economic growth, urbanization, and industrialization on fine particulate matter (PM2.5) concentrations in China Guangdong Li, Chuanglin Fang, Shaojian Wang, and Siao Sun Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.6b02562 • Publication Date (Web): 06 Oct 2016 Downloaded from http://pubs.acs.org on October 7, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 18

Environmental Science & Technology

The effect of economic growth, urbanization, and industrialization on fine particulate matter (PM2.5) concentrations in China Guangdong Li1, Chuanglin Fang1*, Shaojian Wang2*, Siao Sun1 1. Key Laboratory of Regional Sustainable Development Modeling, Institute of Geographic Sciences and Natural Resources Research (IGSNRR), Chinese Academy of Sciences (CAS), 11A Datun Road, Chaoyang District, Beijing 100101, China 2. Guangdong Provincial Key Laboratory of Urbanization and Geo-simulation, School of Geography and Planning, Sun Yat-sen University, Guangzhou, China

Corresponding author: Chuanglin Fang, Shaojian Wang E-mail: [email protected]; [email protected]. TEL: +86-01064889101

1

ACS Paragon Plus Environment

Environmental Science & Technology

1 2

The effect of economic growth, urbanization, and industrialization on fine particulate matter (PM2.5) concentrations in China

3 4

Abstract: Rapid economic growth, industrialization, and urbanization in China has led to extremely severe air

5

pollution that causes increasing negative effects on human health, visibility, and climate change. However, the

6

influence mechanisms of these anthropogenic factors on fine particulate matter (PM2.5) concentrations are poorly

7

understood. In this study, we combined panel data and econometric methods to investigate the main anthropogenic

8

factors that contribute to increasing PM2.5 concentrations in China at the prefecture level from 1999 to 2011. The

9

results showed that PM2.5 concentrations and three anthropogenic factors were cointegrated. The panel Fully

10

Modified Least Squares and panel Granger causality test results indicated economic growth, industrialization, and

11

urbanization increases PM2.5 concentrations in the long run. The results implied that if China persists in its current

12

development pattern, economic growth, industrialization and urbanization will inevitably lead to increased PM2.5

13

emissions in the long term. Industrialization was the principal factor that affected PM2.5 concentrations for the total

14

panel, the industry-oriented panel and the service-oriented panel. PM2.5 concentrations can be reduced at the cost of

15

short-term economic growth and industrialization. However, reducing the urbanization level is not an efficient way

16

to decrease PM2.5 pollutions in the short term. The findings also suggest that a rapid reduction of PM2.5

17

concentrations relying solely on adjusting these anthropogenic factors is difficult in a short term for the heavily

18

PM2.5-polluted panel. Moreover, the Chinese government will have to seek much broader policies that favor a

19

decoupling of these coupling relationships.

20 21

 INTRODUCTION

22

China’s rapid economic growth and urbanization since its reform and opening up has not only increased

23

comprehensive national power and residents’ living standards but also triggered severe ecological problems and

24

environmental pollution,1, 2 especially atmospheric pollution, which has recently attracted broad attention.3, 4 Fine

25

particulate matter (PM2.5 – particles with an aerodynamic diameter that is not larger than 2.5 µm) has been reported

26

as a major pollutant that threatens human health, decreases visibility, and affects the regional and global climate.5, 6

27

In response to severe and persistent PM2.5 pollution, the Chinese State Council declared a goal to reduce PM2.5

28

concentrations by 25% by 2017, relative to the 2012 level.7 To achieve this ambitious goal, an exploration of the 2

ACS Paragon Plus Environment

Page 2 of 18

Page 3 of 18

Environmental Science & Technology

29

impact of anthropogenic factors (particularly rapid economic growth, urbanization and industrialization) on PM2.5

30

is a prerequisite.8, 9 However, few studies have quantitatively investigated the influence mechanisms of economic

31

growth, urbanization, and industrialization on fine particulate matter (PM2.5) concentrations, which means that the

32

dynamic relationship between these variables in China are poorly understood.10 This information is particularly

33

important for China’s decision-makers to make policies to control air pollution and to make decisions on the

34

tradeoff between development and conservation.

35

A shortage of long-term and large-scale PM2.5 pollution data (monitoring networks have not been consistently

36

established in most developing countries) has also resulted in a lack of a full understanding of these complex

37

relationships. Estimating long-term PM2.5 concentrations through remote sensing is a feasible method to address the

38

problem.11 By combining satellite derived PM2.5 concentrations data and socioeconomic data, we developed a

39

long-term panel data from 1999 to 2011. To address heterogeneity, we divided this panel into five sub-panels based

40

on the industrial development level and PM2.5 pollution level of all Chinese prefecture-level cities. A set of

41

complete econometrics methods including the panel unit root test, the Pedroni cointegration test, the panel Fully

42

Modified Least Squares (FMOLS) test, the panel Granger causality test, variance decomposition and impulse

43

response were utilized to quantitatively examine the impact of economic growth, urbanization, and industrialization

44

on PM2.5 concentrations over the short and long term and trends and magnitudes for the five panels.

45

 DATA AND METHODS

46

China has 337 relatively stable prefecture-level cities (a scale smaller than province but larger than county) and

47

was used in this paper as the unit of analysis to identify the impact of GDP (gross domestic product) per capita,

48

industrialization, and urbanization on PM2.5 concentrations. These cities can be used to characterize different

49

conditions and the regional variation of the PM2.5 concentrations in China at the prefecture level, and sufficient

50

socioeconomic data exists to support our study at the prefecture level (see Supporting Information Figure S1 for the

51

spatial units in our analysis; Taiwan, Hong Kong, and Macao are not included in the analyses). Based on this

52

spatial unit we empirically developed a panel data (panel data are data where multiple cases were observed at two

53

or more time periods) at the prefecture-level over the period of 1999 to 2011.

54

The ground-level PM2.5 concentrations used in our study were estimated by combining the data that were

55

retrieved from the Aerosol Optical Depth (AOD) of the Moderate Resolution Imaging Spectroradiometer (MODIS)

56

products of the National Aeronautics and Space Administration (NASA) and Multi-angle Imaging

57

SpectroRadiometer (MISR) instruments with an aerosol vertical profile and scattering properties that were

58

simulated by the GEOS-Chem chemical transport model.11, 12 The global PM2.5 concentrations dataset contains data

59

for the medium- to long-term average PM2.5 concentrations (with three-year moving averages from 1999 to 2011)

60

with an approximate 10 km resolution.11 The validation results indicated that the satellite-derived estimates were

61

generally consistent to the ground-based measurements in China (r = 0.798; slope = 0.849; n = 121, see Supporting

62

Information Figure S2).11 Thus, this dataset can be applied to the analysis of the dynamics of PM2.5 concentrations

63

in large regional studies.11 This dataset can be download from the Atmospheric Composition Analysis Group

64

(http://fizz.phys.dal.ca/~atmos/martin/?page_id=140). In this study, we utilized a subset of the global PM2.5 3

ACS Paragon Plus Environment

Environmental Science & Technology

Page 4 of 18

65

concentrations dataset that contained China’s prefecture-level cities from 1999 to 2011 (see Supporting Information

66

Figure S3).

67

The panel data of economic growth, urbanization and industrialization were derived from the China City

68

Statistical Yearbook and Chinese Regional Economy Statistical Yearbook for 337 prefecture-level cities.13, 14 We

69

used the change in GDP per capita as an indicator to quantify economic growth. All GDP per capita data were

70

converted to constant prices. Urbanization was measured by the proportion of the permanent population of an urban

71

area to the total population. We used the share of industry value-added of GDP to characterize industrialization.

72

The panel data were logarithmically transformed (natural logarithms) to reduce the heteroscedasticity to obtain a

73

stationary panel before applied to empirical tests.

74

Because the pace of development and the characteristics of various spatial units across China were different, the

75

impact of economic growth, urbanization, and industrialization on PM2.5 concentrations across China may be

76

heterogeneous. Therefore, it can be helpful to group the prefecture-level cities into several categories and analyze

77

them separately to provide greater insight. We therefore grouped our total panel (panel C) into four sub-panels

78

including agriculture-oriented (panel I1), industry-oriented (panel I2), service-oriented (panel I3) and heavily

79

PM2.5-polluted (panel H) panels (see Supporting Information Table S1-4 for more detailed information).

80

The main objective of this study was to investigate the effect of economic growth, urbanization, and

81

industrialization on PM2.5 concentrations, in China. To realize this goal, a panel data model was used. The

82

long-term relationship between PM2.5 concentrations, GDP per capita, industrialization and urbanization was

83

formulated as follows:

84

LNPM 2.5it =α + β1 LNGDPPCit + β 2 LN URBAN it + β 3 LN INDit + ε it

85

where LNPPM2.5 is the natural logarithm of PM2.5 concentrations, LNGDPPC is the natural logarithm of GDP per

86

capita, LNURBAN is the natural logarithm of urbanization and LNIND is the natural logarithm of industrialization,

87

i=1, …, N for each prefecture-level city in the panel, t=1, …, T refers to the time period, α is constant, β1, β2, and β3

88

are long-term elasticity estimates of GDP per capita, urbanization and industrialization, respectively, and εit is error

89

terms.

(1)

90

Here we used panel Granger causality to examine relationships between PM2.5 concentrations and other factors.

91

Variable X is said to be the Granger cause of Y if at time t, Yt+l is better predicted by using past values of X than by

92

not doing so (or X should help improving the prediction of Y).15 The estimation procedure included five steps (see

93

Supporting Information Text S1 for the detailed description):

94 95

(1) Two types of panel unit root tests from Levin, Lin & Chu t (LLC)16 and Im, Pesaran and Shin W-stat (IPS)17 were used to test whether our variables were stationary at the level or at the first difference.

96

(2) The Pedroni cointegration test was applied to explore whether there was any long-term relationship between

97

PM2.5 concentrations, economic growth, urbanization and industrialization. Pedroni conducted a series of tests for

98

panel cointegration and included two alternative hypotheses types, namely, the homogeneous alternative and the

99

heterogeneous alternative.18

4

ACS Paragon Plus Environment

Page 5 of 18

Environmental Science & Technology

100

(3) For the Panel Fully Modified Least Squares (FMOLS) estimates, if the four variables were cointegrated, the

101

Panel FMOLS regression19 can be utilized to estimate whether LNGDPPC, LNIND and LNURBAN (independent

102

variables) had a positive or negative long-term relationship with LNPM2.5 (dependent variable).

103

(4) For the Panel Granger causality test, if the variables were cointegrated, the Vector Error-Correction Model

104

(VECM) was employed to investigate the short- and long-term bidirectional or unidirectional Granger causality

105

between the variables.20 According to the general VECM model, short-term Granger causality, also referred as

106

“weak Granger causality”, “short-term fluctuations” or “short-term disequilibrium”, usually concerns a period of

107

1-2 years according to relevant lags selected in the test, whereas long-term Granger causality, also referred as

108

“long-term equilibrium relationship”, concerns the whole study period from 1999-2011 in this test. However, if the

109

variables were not cointegrated, the Vector Autoregressive Model (VAR) was utilized to identify the short-term

110

Granger causality.

111

(5) As the panel Granger causality tests indicated only the causality among the variables, we employed the

112

variance decomposition method to capture the amount of information that each variable contributed to the other

113

variables in the autoregression.21 We also used an impulse response function to examine the response of one

114

variable to an impulse in the other variables.22

115

 RESULTS

116

Panel unit root test results. The panel unit root test was applied to the natural logarithm of the four variables,

117

PM2.5 concentrations, GDP per capita, industrialization and urbanization, in two models (intercept alone and

118

intercept and trend) to test the stationarity of the variables. Table 1 reports the Levin, Lin & Chu t (LLC) and Im,

119

Pesaran and Shin W-stat (IPS) test results, which indicate that at levels not all of the variables were stationary.

120

However, four variables were stationary at the first difference. Thus, the null hypothesis can be rejected at the 1%

121

level of significance, which indicates that the four variables contained a panel unit root. Supporting Information

122

Table S5 shows the results for the four sub-panels, all of which contain a panel unit root.

123

Table 1. Panel unit root test results of variables for panel C. Level First difference Intercept Intercept and trend Intercept Intercept and trend Levin, Lin & Chu t LNPM2.5 -36.8901*** -12.8776*** -40.1577*** -43.3925*** LNGDPPC 7.9962 -34.4917*** -34.2228*** -24.3503*** LNURBAN -2434.59*** -2566.27*** -2580.75*** -47.5252*** *** *** *** LNIND -24.6527 -20.835 -17.8525 -18.1678*** Im, Pesaran and Shin W-stat LNPM2.5 -18.9786*** 9.853 -19.3127*** -23.9887*** *** *** LNGDPPC 33.5389 -14.4999 -18.014 -3.4653*** LNURBAN -162.689*** -75.4164*** -195.06*** -16.289*** *** *** LNIND -0.1877 -5.6757 -10.743 -1.7616** Notes: The panel unit root tests with intercept and trend are carried independently; the optimal lag lengths are obtained automatically with the Schwarz information criteria (SIC). *** Indicates significance at 1% level. Variable

124 125 126 127

5

ACS Paragon Plus Environment

Environmental Science & Technology

Page 6 of 18

128

Pedroni cointegration test results. Because the variables were stationary at the first difference, the Pedroni

129

cointegration test was used. The results of the Pedroni cointegration test for the panel C data are given in Table 2.

130

The results reveal that ten statistics were significant for rejecting the null hypothesis of no cointegration, which

131

shows the existence of a long-term relationship between the independent variables, economic growth,

132

industrialization and urbanization, and the dependent variable PM2.5 concentrations, in China. Supporting

133

Information Table S6 also indicates the same result for the four sub-panels.

134

Table 2. Pedroni cointegration test results for panel C.

135 136 137

Alternative hypothesis: common AR coefs. (within-dimension) Statistic Panel v-Statistic -0.3908 Panel rho-Statistic 8.3586 Panel PP-Statistic -11.0098 Panel ADF-Statistic -20.6232 Panel v-Statistic (Weighted statistic) -6.4642 Panel rho-Statistic (Weighted statistic) 9.5352 Panel PP-Statistic (Weighted statistic) -16.6866 Panel ADF-Statistic (Weighted statistic) -28.8429 Alternative hypothesis: individual AR coefs. (between-dimension) Group rho-Statistic 16.9664 Group PP-Statistic -24.4278 Group ADF-Statistic -32.2108 Note: we use the automatic selection based on Schwarz to choose the optimal lag length.

Prob. 0.6520 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

The Panel Fully Modified Least Squares (FMOLS) test results. Giving that the variables were cointegrated,

138

the panel FMOLS was employed based on equation (1), and the results are presented in Table 3. All of the panels

139

showed that economic growth, industrialization and urbanization had a positive long-term relationship with PM2.5

140

concentrations. This result is the most important finding in this test. It suggests that an increase in economic growth,

141

industrialization and urbanization will increase PM2.5 concentrations in the long term in China for these

142

prefecture-level cities. A 1% increase in the GDP per capita, industrialization and urbanization will increase PM2.5

143

concentrations by 0.12%, 0.12% and 0.20% respectively, for panel C; 0.11%, 0.15% and 0.17% for panel I1; 0.16%,

144

0.26% and 0.21% for panel I2; 0.09%, 0.12% and 0.55% for panel I3; and 0.15%, 0.49% and 0.27% for panel H. A

145

trend was evident for the effect of LNIND on LNPM2.5; specifically, the heavily PM2.5-polluted panel had the

146

strongest effect, which was closely followed by panels I2, I1 and I3. For panels I2 and H, the effects of economic

147

growth on LNPM2.5 were stronger. Moreover, urbanization had the greatest impact on LNPM2.5 for panel I3.

148 149

Table 3. The Panel Fully Modified Least Squares (FMOLS) test results with LNPM2.5 as the dependent variable. Panel

150 151 152

LNGDPPC LNIND LNURBAN Coefficients t-statistics Coefficients t-statistics Coefficients t-statistics 0.1212*** 13.2792 0.1186*** 4.1601 0.1993*** 5.1120 C 0.1111*** 9.4880 0.1521*** 4.4084 0.1745*** 3.8037 I1 0.1551*** 7.7861 0.2550*** 2.4208 0.2068** 2.3129 I2 0.0913*** 4.2139 0.1249** 2.0429 0.5463*** 3.8644 I3 0.1482*** 6.3403 0.4947*** 8.6743 0.2714*** 3.0225 H ** Indicates significance at 5% level. *** Indicates significance at 1% level. Note that the coefficients value is the coefficient estimates of equation (1) for the log- transformed variables.

6

ACS Paragon Plus Environment

Page 7 of 18

Environmental Science & Technology

153

Panel Granger causality test results. The panel FMOLS tests provided a relatively reliable result that identified

154

the effect of four independent variables on LNPM2.5; however, the directions (bidirectional or unidirectional) of

155

possible causal relationships between variables remained untested. Because the variables were cointegrated, we

156

used the Vector Error-Correction Model (VECM) to test the panel Granger causality, and the results are presented

157

in Table 4. Two types of tests were considered for panel Granger causality. The first type was long-term causality,

158

which depended on the significance of the error-correction terms ECT (-1). The second type was short-term

159

causality, which was determined by the joint significance of the coefficients of the lagged terms of each

160

independent variable. If the variables, LNPM2.5, LNGDPPC, LNURBAN and LNIND are cointegrated, then it is

161

expected that at least one or all the ECTs have negative coefficient and should be significantly non-zero. Short run

162

Granger causality is determined by implementing a Wald test of the significance of the lags of each of the

163

explanatory variables.

164

Table 4 indicates that the coefficient of ECT (-1) is completely negative and significant at the 1% level for panels

165

C and I1, which indicates bi-directional long-term causality (i.e., a feedback effect) among PM2.5 concentrations,

166

economic growth, industrialization and urbanization for these two panels. Moreover, a bi-directional long-term

167

causality was found between PM2.5 concentrations, urbanization and industrialization for panels I2 and I3, and a

168

bi-directional long-term causality was found between PM2.5 concentrations, economic growth and industrialization

169

for panel H. In addition, a one-way causal relationship was found running from economic growth to the other

170

variables for panels I2 and I3 and running from urbanization to the other variables for panel H.

171

The short-causal relationship indicates a bi-directional positive causal relationship between PM2.5 concentrations

172

and other factors including economic growth, industrialization, and urbanization for panels C and I1. The

173

short-causality also identifies a bi-directional positive causal relationship between LNPM2.5, LNURBAN, and

174

LNIND for panel I2. Moreover, a unidirectional short-run causal relationship was concluded running from

175

economic growth to PM2.5 concentrations for panel I2, and from industrialization to PM2.5 concentrations for panel

176

I3. Figure 1 shows the long- and short-term causal relationships among the four variables for the five panels. It is

177

worth noting that all the links built between PM2.5 concentrations and social-economic factors are based on the

178

statistical test (i.e., the Panel Granger Causality test). These links represent Granger causality showing statistical

179

causal relationships between variables, which may not be “true causality”.

180

Table 4. Panel Granger causality test results. Panel C

I1

I2

Dependent variable ∆LNPM2.5 ∆LNGDPPC ∆LNIND ∆LNURBAN ∆LNPM2.5 ∆LNGDPPC ∆LNIND ∆LNURBAN ∆LNPM2.5 ∆LNGDPPC ∆LNIND

Independent variables Short run causality (χ2-Wald statistics) ∆LNPM2.5 102.2109*** 22.2378*** 5.9045** 148.7444*** 8.2102*** 9.1660*** 1.2085 4.7504*

∆LNGDPPC 130.0746*** 5.3372* 0.9662 105.5832*** 0.5290 7.1769** 11.8464*** 0.0480

∆LNIND 144.7306*** 16.6776*** 1.5789 49.9758*** 34.6976*** 4.9426* 54.6176*** 0.1202 -

Long run causality ∆LNURBAN 20.5743*** 20.1023*** 3.2116 29.1685*** 2.9033 1.6523 4.8906* 18.8542*** 11.5163***

7

ACS Paragon Plus Environment

ECT (-1) -0.7266*** -0.1325*** -0.1250*** -0.0022** -0.6648*** -0.2456*** -0.0462*** -0.0053*** -0.3865*** -0.0160 -0.7321***

t-statistics -32.2161 -12.4810 -11.4274 -2.2359 -26.0327 -16.0701 -4.5380 -3.0820 -9.7787 -1.4186 -13.6546

Environmental Science & Technology

∆LNURBAN 9.6533*** 8.6323*** 35.1574*** -0.0078** -2.1015 ∆LNPM2.5 3.0907 27.9483*** 0.6320 -0.9489*** -12.4027 ∆LNGDPPC 2.3047 0.9927 2.8455 -0.0146 -1.1418 ∆LNIND 1.0513 3.3080 0.1709 -0.1456*** -4.4493 -1.9100 ∆LNURBAN 3.1847 0.4452 1.5146 -0.0122* ∆LNPM2.5 1.8454 0.2696 3.6301 -1.2241*** -13.2509 H ∆LNGDPPC 19.9780*** 4.9678* 2.2613 -0.0501*** -6.0574 ∆LNIND 13.3850*** 1.3965 0.2137 -0.0160*** -3.3911 ∆LNURBAN 0.4424 4.7330* 1.8989 -0.0111 -1.1034 Note: The null hypothesis is that there is no causal relationship between variables, and ∆ represents the difference operator. ECT (-1) is the error correction term lagged one period. * Indicates significance at 10% level. ** Indicates significance at 5% level. *** Indicates significance at 1% level. I3

181 182 183 184 185 186 187

One of the most important findings in the panel Granger causality results is that industrialization significantly

188

increased PM2.5 concentrations more than economic growth and urbanization for panels C, I2 and I3 in the

189

short-term. For panel I1, however, economic growth had a greater effect on PM2.5 concentrations than the other

190

variables. The lack of causal relationship from economic growth and urbanization to PM2.5 concentrations for panel

191

I3 indicates that slowing economy and reducing urbanization in short term will not necessarily decrease PM2.5

192

concentrations in these prefecture-level cities. In panel H, economic growth, industrialization and urbanization did

193

not have a short-term effect on PM2.5 concentrations. However, a bi-directional long-term causal relationship was

194

found among PM2.5 concentrations, industrialization and economic growth, which implies that the increasing PM2.5

195

concentrations was a long-term process driven by an increase in both industrialization and economic growth. This

196

also suggest that a rapid reduction of PM2.5 concentrations relying solely on adjusting these anthropogenic factors is

197

difficult in a short term for the heavily PM2.5-polluted panel.

198 199 200

Figure 1. Long-run causality and short-run causality between PM2.5 concentrations (PM2.5), economic growth (GDPPC), industrialization (IND) and urbanization (URBAN).

201 8

ACS Paragon Plus Environment

Page 8 of 18

Page 9 of 18

Environmental Science & Technology

202

The results of variance decomposition and impulse response. Figure 2 shows the results of the variance

203

decomposition analysis for panel C, which indicates that PM2.5 concentrations were mostly explained by its own

204

innovative shocks (60.12%), whereas the contributions of economic growth, industrialization and urbanization to

205

PM2.5 concentrations were equal to 16.13%, 23.67% and 0.08%, respectively (see Supporting Information Table

206

S7). Thus, the contribution of industrialization to PM2.5 concentrations was important and was closely followed by

207

economic growth and urbanization. Additionally, economic growth was principally self-explanatory (88.21%).

208

Moreover, 22.42% of industrialization was explained by one standard deviation shock in economic growth. Most of

209

the effects on industrialization were explained by its own standard shocks (69.01%), but the contribution of

210

urbanization was negligible (0.004%). The results also reveal that 99.72% of urbanization can be explained by its

211

own innovative shocks, whereas the contributions of economic growth and industrialization to urbanization were

212

equal to 0.02% and 0.008%, respectively. Supporting Information Table S7-11 and Figure S4-7 show detailed

213

results for the variance decomposition of the four sub-panels. In general, the variance decomposition results

214

indicate that the contribution of industrialization in explaining PM2.5 concentrations was the most important for

215

panels I2, I3 and H, whereas economic growth was crucial for panel I1, and urbanization had a negligible effect for

216

all panels.

217 218

Figure 2. Variance decomposition of four variables for panel C.

219 220

The impulse response analysis results for panel C are shown in Figure 3. This impulse response analysis revealed

221

the responses in one variable due to shocks that originate from other variables. PM2.5 concentrations first decreased

222

then stabilized because of shocks from decreasing economic growth and urbanization, whereas the response in

223

PM2.5 concentrations to shocks originated in industrialization increased in the earlier four years. The response of 9

ACS Paragon Plus Environment

Environmental Science & Technology

224

economic growth fluctuated because of shocks from increasing industrialization and urbanization and decreasing

225

PM2.5 concentrations in the previous four years. The response in industrialization first fluctuated then stagnated

226

because of shocks from increasing economic growth and PM2.5 concentrations in the earlier four years. The

227

response of urbanization increased in the previous two years because of shocks of increasing PM2.5 concentrations

228

and economic growth. Supporting Information Figure S8-11 provides the detailed results for the impulse response

229

analysis of the four sub-panels.

230 231

Figure 3. Responses of variables to one S.D. innovations for panel C.

232

 DISCUSSION

233

Previous research has shown that an increase in PM2.5 concentrations has been affected synthetically by natural

234

and socio-economic variables.8 However, to simplify these variables, we have chosen the three variables, economic

235

growth, industrialization and urbanization. We assumed that these three variables are the most important

236

socio-economic factors that affect PM2.5 concentrations and that these variables can indicate the overall trend and

237

condition of China's socio-economic development because socio-economic development is a complex interactive

238

system. The results of our analysis confirmed our hypothesis. These three variables had strong explanatory power

239

for the increase in PM2.5 concentrations (for example, the results of the FMOLS analysis indicated that all R2 of the

240

regressions were greater than 0.95).

241

All of the long-term causal relationships that we identified were consistent with our expectations and accorded to

242

the economic and environmental reality in China. We deduced that if the development mode in Chinese cities keep

243

unchanged, controlling PM2.5 concentrations has an adverse effect on economic growth, urbanization and

244

industrialization or at least on two of these variables for most of the panels based on the long-term causality results. 10

ACS Paragon Plus Environment

Page 10 of 18

Page 11 of 18

Environmental Science & Technology

245

Conversely, under current mode of development, increases in economic growth, urbanization and industrialization

246

will inevitably lead to increased PM2.5 emissions. This implies that the feedback effect exists between PM2.5

247

emissions and economic growth, urbanization and industrialization in the current setup. Thus, when establishing

248

policies to control PM2.5 emissions, policymakers should carefully make trade-offs between reduction of PM2.5

249

concentrations and increased economic growth, urbanization and industrialization under the current economic

250

development mode. Even more important, the Chinese government will have to seek much broader policies

251

(including transforming economic growth pattern, upgrading industrial structure and adjusting urbanization process)

252

that favor a decoupling of these coupling relationships. Our findings also indicate that the relationships between

253

PM2.5 concentrations and economic growth, urbanization, and industrialization across China were heterogeneous,

254

which was exemplified by the differences in the agriculture-oriented panel (I1), industry-oriented panel (I2),

255

service-oriented panel (I3), and the heavily PM2.5-polluted panel (H). We found that bi-directional long-term

256

causality exists between PM2.5 concentrations, economic growth, urbanization, and industrialization for panels C

257

and I1. Moreover, bidirectional long-term causal relationships between PM2.5 concentrations, urbanization, and

258

industrialization were found for panels I2 and I3, and bi-directional long-term causality was found among PM2.5

259

concentrations, economic growth and industrialization for panel H. Therefore, different regional policies are

260

necessary to establish corresponding strategies to reduce PM2.5 emissions.

261

The short-term Granger causality tests indicate that most of the results were consistent with the long-term results

262

for panels C, I1 and I2. Moreover, we found that industrialization was more significant than economic growth and

263

urbanization in increasing PM2.5 concentrations in panels C, I2 and I3 in short-term causality, which suggests that

264

reducing industrialization is probably one of the most effective countermeasures to reduce PM2.5 concentrations in

265

the short term for these panels. However, for panel I1, economic growth has a greater effect on PM2.5

266

concentrations followed by industrialization and urbanization, which implies that a stronger coupling relationship

267

exhibits between economic growth and PM2.5 concentrations in agriculture-oriented cities. In panel H, economic

268

growth, industrialization and urbanization did not have short-term effects on PM2.5 concentrations, which indicates

269

adjusting these three anthropogenic variables in short term will not significantly reduce PM2.5 concentrations. In

270

addition, compared with economic growth and industrialization, reducing the urbanization level will not effectively

271

reduce PM2.5 pollutions in the short term for all panels. This finding was further validated by the results of the

272

variance decomposition for LNPM2.5. For example, a one standard deviation shock in urbanization explained only

273

0.08% and 0.003% of the PM2.5 concentrations for panels C and H, respectively.

274

The impulse response analysis results do not show a decreasing linear relationship between PM2.5 concentrations

275

and economic growth. However, the variance decomposition results demonstrate that the contribution of economic

276

growth in explaining PM2.5 concentrations was higher than urbanization. Even for panel I1, economic growth has a

277

greater effect on PM2.5 concentrations than the other variables in the short term. This result again indicates that

278

there was a tight relationship between PM2.5 emissions and economic growth in China. In contemporary China,

279

economic growth is generally the most important target of the central and local governments. The 13th Five Year

280

Plan of China is a vital program document that sets goals and guidelines covering many social, economic and

281

environmental issues and informs Chinese policymaking from 2016 to 2020. The plan proposes that an annual 11

ACS Paragon Plus Environment

Environmental Science & Technology

282

growth rate of approximately 6.5% is still China's minimum requirement to maintain economic growth and improve

283

peoples’ livelihood.23 Thus, the tight relationship between PM2.5 emissions and economic growth indicates that

284

China needs to urgently explore a new pathway that uncouples the effects of economic growth on PM2.5 emissions

285

in order to avoid air pollution aggravation due to economic growth. In this context, for central and local

286

governments of China, the transformation of the economic growth pattern is necessary to mitigate PM2.5 emissions.

287

Hence, one of the priorities is to accelerate the transformation of the economic development pattern changing from

288

the extensive pattern relying on large amounts of resources and energy consumptions (with inefficient growth and

289

increased PM2.5 emissions) to the intensive pattern mainly based on technical progresses and efficiency

290

improvement (with efficient growth and reduced PM2.5 emissions). The transition from the extensive economic

291

growth pattern towards intensive pattern helps reduce PM2.5 emissions. Although this transformation is not likely to

292

happen in the short term, such a change is necessary and urgent, because it is an indispensable and fundamental

293

step in reducing the PM2.5 pollutions.

294

As expected, industrialization caused PM2.5 emissions in both the short and long terms as indicated by the

295

Granger causality test. The close linkage between PM2.5 emissions and industrialization indicates that the current

296

industrial development pattern was highly dependent on the energy use and resource consumption, leading to high

297

pollutions. In recent years, China has experienced a period when economic development has been driven

298

dominantly by heavy industry, especially the heavy-chemical industry contributing more than 70% of the total

299

industrial value since 2006.24 However, we conclude that China’s economic growth will not continue to rely on

300

heavy industries in the foreseeable future. On the one hand, China has severe overcapacity problem after rapid

301

development of heavy-industrialization in recent years, especially steel, aluminum and cement industries. This

302

problem will get worse if China still relies on heavy industries as an economic driver. On the other hand, both the

303

central and local governments in China have realized that economic development relying on heavy industries is

304

unsustainable and it needs to be altered. While progress has been made, there is still a long way to go before

305

industrial development transformation can be reached. The following measures may help speed up the

306

transformation which benefits the mitigation of PM2.5 emissions: 1) to transform the main industrial development

307

driver from traditional resources and energy consumptions to innovation and technical progresses; 2) to upgrade

308

China’s industrial structure by encouraging the tertiary industry development; 3) to limit and eliminate heavy

309

pollution industries. In addition, to raise industrial PM2.5 emissions standards and to improve energy use

310

efficiencies are among the more direct control measures to reduce PM2.5 pollutions.

311

The results of the Granger causality test confirm that urbanization caused PM2.5 emissions for most of the panels

312

with the exception of panels I3 and H in the short term, although the effect was weaker than other variables.

313

Therefore, a gradual slowdown of urbanization may actually help reduce PM2.5 pollutions for these panels.

314

However, this deceleration of urbanization is not likely to occur, because a high target for increasing urbanization

315

to 60% in 2020 was established in the New-type Urbanization Plan (2014-2020)25 that was recently released by

316

China’s central government. Moreover, Chinese cities usually have more densely populated urban areas in

317

comparison to cities in developed countries. As the result shows, a city with a high urbanization level tends to have

12

ACS Paragon Plus Environment

Page 12 of 18

Page 13 of 18

Environmental Science & Technology

318

high PM2.5 concentrations in current China. Hence, limiting population influx into densely populated urban areas is

319

probably a practical strategy to reduce PM2.5 emissions.

320

It is worth noting that future scenarios may be totally different from the current state. Inevitably, China will, like

321

other developed countries, enter a different economic state. For example, people will no longer find it desirable to

322

further crowd into large cities. And heavy industry will no longer add enough value to the economy to fuel further

323

growth, which will further result in a move to lighter industries and finally services. A changing economic state

324

will certainly change causality between PM2.5 concentrations and other variables. Hence a simple extrapolation of

325

the future trend based on current relations may deviate from the reality. This, again, requires Chinese government

326

to seek much broader policies (including transforming economic growth pattern, upgrading industrial structure and

327

adjusting urbanization process) to decouple relationships between PM2.5 concentrations and other socio-economic

328

factors towards a more sustainable development mode.

329

One limitation of this study is that the spatial unit of this analysis is the prefecture-level city, a relative large

330

spatial scale in China. Although several different panels were utilized to address heterogeneity in this study, we

331

should investigate this heterogeneity and the differences more deeply. Moreover, different areas of a

332

prefecture-level city have different PM2.5 concentrations, especially in urban and rural areas. However, It would be

333

difficult to obtain long-term data of PM2.5 concentrations and social economic variables for a panel of a more

334

refined spatial unit. Therefore, if these data can be obtained (from a PM2.5 observation station, for example), this

335

research would be helpful for future study. Second, this study found that economic growth was more important in

336

explaining PM2.5 concentrations than urbanization. Future studies that employ GDP per capita as a proxy for

337

economic growth and PM2.5 concentrations as a proxy for environmental impact may provide a deeper insight into

338

the interactive relationship between environmental impact and economic growth and Kuznets’ environmental curve

339

can be tested. Third, exploring the causal relationships between PM2.5 emissions and other potentially related

340

variables, such as heavy industry development, automobile use, exports and dust pollution, may be an important

341

direction for future studies. Much evidence exists that the increased development of the construction, metal and

342

machine production sectors, exports and urban expansion has exacerbated PM2.5 emissions.8, 9

343 344

 ASSOCIATED CONTENT

345 346 347 348 349 350 351 352 353 354

Supporting Information Supporting Information Available: Prefecture-level divisions in China for period of 1999-2011 (Figure S1). The validation result of satellite-derived estimates and ground-based measurements in China for PM2.5 (Figure S2). Spatial distribution of mean PM2.5 concentrations in China from 1999 to 2011 (Figure S3). Descriptive statistics for five panels used in this study for the period 1999-2011 (Table S1). The mean changes of PM2.5 concentrations, per capita GDP, urbanization and industrialization for five panels in 1999 and 2011 (Table S2). The list of prefecture-level divisions for four sub-panels (Table S3). A list of baseline characteristics for the major prefecture-level divisions in China (Table S4). The detailed description for the estimation procedure (Text S1). The detailed estimation results of the four sub-panels (Tables S5-S11, Figures S4-S11). This material is available free of charge via the Internet at http://pubs.acs.org. 13

ACS Paragon Plus Environment

Environmental Science & Technology

355

 AUTHOR INFORMATION

356 357 358

*E-mail: [email protected]. Tel.: +86-01064889301. Fax: +86-01064889301 (C.F.).

359 360

Notes The authors declare no competing financial interest.

361

 ACKNOWLEDGEMENTS

362 363

This work was supported by the Natural Science Foundation of China (Grant nos. 41590842, 41501175 and 71433008).

364

 REFERENCES

365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402

(1) Liu, J.; Diamond, J., China's environment in a globalizing world. Nature 2005, 435, (7046), 1179-1186. (2) Bai, X.; Shi, P.; Liu, Y., Society: Realizing China's urban dream. Nature 2014, 509, (7499), 158-160. (3) Zhang, Y.-L.; Cao, F., Fine particulate matter (PM2.5) in China at a city level. Sci Rep 2015, 5, 14884-14884. (4) Han, L.; Zhou, W.; Li, W., Increasing impact of urban fine particles (PM2.5) on areas surrounding Chinese cities. Sci Rep 2015, 5, 12467. (5) Cao, J., Pollution status and control strategies of PM2. 5 in China. J. Earth Environ 2012, 3, 1030-1036. (6) Wang, Y.; Zhang, R.; Saravanan, R., Asian pollution climatically modulates mid-latitude cyclones following hierarchical modelling and observational analysis. Nat Commun 2014, 5, 3098. (7) Chinese State Council Atmospheric Pollution Prevention and Control Action Plan. (29 October 2015), (8) Guan, D.; Su, X.; Zhang, Q.; Peters, G. P.; Liu, Z.; Lei, Y.; He, K., The socioeconomic drivers of China's primary PM2.5 emissions. Environ Res Lett 2014, 9, (2), 024010. (9) Lin, G.; Fu, J.; Jiang, D.; Hu, W.; Dong, D.; Huang, Y.; Zhao, M., Spatio-Temporal Variation of PM2.5 Concentrations and Their Relationships with Geographic and Socioeconomic Factors in China. Int J Env Res Pub He 2014, 11, (1), 173-186. (10) Zhang, Q.; He, K.; Huo, H., Policy: Cleaning China's air. Nature 2012, 484, (7393), 161-162. (11) van Donkelaar, A.; Martin, R. V.; Brauer, M.; Boys, B. L., Use of Satellite Observations for Long-Term Exposure Assessment of Global Concentrations of Fine Particulate Matter. Environ Health Persp 2015, 123, (2), 135-143. (12) van Donkelaar, A.; Martin, R. V.; Brauer, M.; Kahn, R.; Levy, R.; Verduzco, C.; Villeneuve, P. J., Global Estimates of Ambient Fine Particulate Matter Concentrations from Satellite-Based Aerosol Optical Depth: Development and Application. Environ Health Persp 2010, 118, (6), 847-855. (13) National Bureau of Statistics of China. China City Statistical Yearbook 2000-2012 (in Chinese); China Statistic Press: Beijing, 2000-2012. (14) National Bureau of Statistics of China. China Statistical Yearbook for Regional Economy 2000-2012 (in Chinese); China Statistical Press: Beijing, 2000-2012. (15) Granger, C. W. J., Investigating causal relations by econometric models and cross-spectral methods. Econometrica 1969, 37, (3): 424–438 (16) Levin, A.; Lin, C. F.; Chu, C. S. J., Unit root tests in panel data: asymptotic and finite-sample properties. J Econometrics 2002, 108, (1), 1-24. (17) Im, K. S.; Pesaran, M. H.; Shin, Y., Testing for unit roots in heterogeneous panels. J Econometrics 2003, 115, (1), 53-74. (18) Pedroni, P., Panel cointegration: asymptotic and finite sample properties of pooled time series tests with an application to the PPP hypothesis. Economet Theor 2004, 20, (03), 597-625. (19) Pedroni, P., Fully modified OLS for heterogeneous cointegrated panels. Adv Ecoometrics, 2000, 15, 93-130. (20) Canning D, Pedroni P. Infrastructure, long-run economic growth and causality tests for cointegrated panels. Manch Sch 2008, 76, 504-527. (21) Alshehry, A. S.; Belloumi, M., Energy consumption, carbon dioxide emissions and economic growth: The case of Saudi Arabia. Renew Sust Energ Rev 2015, 41, 237-247.

Corresponding Author *E-mail: [email protected]. Tel.: +86-02064111475. Fax: +86-02064111475 (S.W.).

14

ACS Paragon Plus Environment

Page 14 of 18

Page 15 of 18

403 404 405 406 407 408 409 410 411 412

Environmental Science & Technology

(22) Yuan, J.-H.; Kang, J.-G.; Zhao, C.-H.; Hu, Z.-G., Energy consumption and economic growth: Evidence from China at both aggregated and disaggregated levels. Energ Econ 2008, 30, (6), 3077-3094. (23) CPC Central Committee's proposal on formulating the thirteenth five-year plan (2016-2020) on national economic and social development; http://news.xinhuanet.com/english/2015-11/03/c_134780200.htm (accessed Dec. 20, 2015). (24) Guan, D.; Klasen, S.; Hubacek, K.; Feng, K.; Liu, Z.; He, K.; Geng, Y.; Zhang, Q., Determinants of stagnating carbon intensity in China. Nature Clim Change 2014, 4, (11), 1017-1023. (25) The new-type of urbanization plan (2014-2020) (in Chinese), http://news.xinhuanet.com/house/bj/2014-03-17/c_126274610.htm (accessed Dec. 21, 2015).

15

ACS Paragon Plus Environment

Environmental Science & Technology

Figure 1. Long-run causality and short-run causality between PM2.5 concentrations (PM2.5), economic growth (GDPPC), industrialization (IND) and urbanization (URBAN). 339x222mm (300 x 300 DPI)

ACS Paragon Plus Environment

Page 16 of 18

Page 17 of 18

Environmental Science & Technology

Figure 2. Variance decomposition of four variables for panel C Figure 2. Variance decompositi 190x154mm (300 x 300 DPI)

ACS Paragon Plus Environment

Environmental Science & Technology

Figure 3. Responses of variables to one S.D. innovations for panel C Figure 3. Responses of variabl 190x162mm (300 x 300 DPI)

ACS Paragon Plus Environment

Page 18 of 18