University of Southern California Dissertations and Theses

The impact of agglomeration policy on CO₂ emissions: an empirical study using China’s manufacturing data

(USC Thesis Other)

The impact of agglomeration policy on CO₂ emissions: an empirical study using China’s manufacturing data

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

The Impact of Agglomeration Policy on CO
2
Emissions
--An Empirical Study using China’s manufacturing data

By

Wei Wei

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)
August, 2015

Copyright 2015 Wei Wei

1
ACKNOWLEDGMENTS

I would like to first thank my advisor, Professor Cheng Hsiao, for his constant
support, personal attention, essential advice and endless encouragement during my
graduation study and research. His guidance has made my pursuit for Ph.D in Economics
a thoughtful and rewarding journey.
This dissertation could not have been finished without the help from many
professors, staff, graduate students, and my family. Special thanks go to my advisory
committee members Professor Jeff Nugent, Professor Robert Dekle and Professor Marc
Schiler for their important remarks in this research and for their time and great effort in
service on my dissertation committee.
I also greatly appreciate the help from many Professors in and outside Economics
Department, including but not limited to Professor Mark Moore, Professor Guofu Tan,
Professor John Strauss, Professor Geert Ridder, Professor Juan Carrillo, and Professor
Caroline Betts.
I wish to thank Young Miller, Morgan Ponder and Shannon Durbin for their great
support and friendship. Many thanks to Bo Zhou, Yanyu Wu, Shuyang Sheng and many
other colleagues for sharing their experiences in research and life.
I must express my love and gratitude to my parents, my husband and my soon-to-be
born baby Andre. I would never be able to accomplish this work without your love.

2
TABLE OF CONTENTS

ACKNOWLEDGMENTS……………………………………………………………….1
LIST OF FIGURES……………………………………………………………………...3
LIST OF TABLES……………………………………………………………………......4
Abstract…………………………………………………………………………………...5
Chapter 1 INTRODUCTION……………………………………….……….………..6
Chapter 2 CHINA’S AGGLOMERATION POLICY……………………………....10
2.1 Concrete Steps to Promote Agglomeration…………………………….……..10
2.2 Agglomeration before and after 2001…………………………………………13
Chapter 3 LITERATURE REVIEW………………………………………………...17
Chapter 4 DATA………………………………………………………………………23
4.1 Sources, Availability and Horizon…………………………………………….23
4.2 Method of Constructing CO
2
emissions data at sub-sector level……………25
4.2.1 Methods in literature and Method adopted in this paper………………25
4.2.2 Fuel consumptions and missing data……………………………………..27
4.2.3 Conversion factors………………………………………………………...28
4.3 Other data………………………………………………………………………31
Chapter 5 MODELS AND FORECASTS…………………………………………...33
5.1 Motivation for various modeling and averaging……………………………..33
5.2 Time-series model……………………………………………………………...35
5.3 Economic model………………………………………………………………..53
5.4 Population averaged Model…………………………………………………...67
5.5 GVAR model……………………………………………………………………77
5.6 Average of forecasts……………………………………………………………92
Chapter 6 DISCUSSIONS OF RESULTS………………………………………….106
Reference..……………………………………………………………………………...116
Appendix A: Energy Consumption
(autocorrelation, partial autocorrelation and time series forecast)……………...124
Appendix B: Output
(autocorrelation, partial autocorrelation and time series forecast)……………...142
Appendix C: Employment
(autocorrelation, partial autocorrelation and time series forecast)…………..….160
Appendix D: Time series forecast of Purchasing price of fuel and raw materials...180
Appendix E: Weight matrix for GV AR model…………………………...……….….181

3
LIST OF FIGURES

Figure 1-1: CO
2
emissions of Manufacturing Industry as percentages of
China’s total CO
2
emissions…………………………………………………………..7
Figure 1-2: CO
2
emissions: China, China’s Manufacturing industry,
and the United States…………………………………………………………….……7
Figure 5-1: Autocorrelation graphs by sector_ CO
2
emissions…………………………..39
Figure 5-2: Partial autocorrelation graphs by sector_ CO
2
emissions…………………...45
Figure 5-3: Time series forecasts by sector __ CO
2
emission……………………………52
Figure 5-4: Time series forecast in comparison with the actual data
__ Total CO2 emission………………..……………………………………………...53
Figure 5-5: ECONOMIC MODEL_ forecast by sector_ CO
2
emissions………………...66
Figure 5-6: Economics model_ forecast in comparison with the actual data
__ CO2 emissions……………………………………………………………………66
Figure 5-7: POPULATION-AVERAGED model forecast by sector_ CO
2
emissions…..76
Figure 5-8: Population averaged model forecast in comparison with the actual data
__ Total CO
2
emission………………………………………………………………..77
Figure 5-9: GVAR model_ forecast by sector_ CO
2
emissions………………………….91
Figure 5-10: GV AR model_ forecast in comparison with the actual data
__ CO
2
emissions……..……………………………………………………………...92
Figure 5-11: Total CO
2
emissions_ Average of forecasts from 4 models
in comparison with the actual data…………………………………………….……..92
Figure 5-12: CO
2
per capita_ Average of forecasts from 4 models
in comparison with the actual data………..……………………………………….…93
Figure 5-13: CO
2
emissions_ by sector_ Average of forecasts from 4 models
in comparison with the actual data…………………………………………………...99
Figure 5-14: CO
2
per capita_ by sector_ Average of forecasts from 4 models
in comparison with the actual data……………………..…………………………...105
Figure 6-1: “CDM project distribution within host countries by region and type”…….107

4
LIST OF TABLES

Table 2-1: Calculation of data collected in section 4 for 21 manufacturing sector………16
Table 4-1: Names of 21 combined sectors……………………………………………….24
Table 4-2: IPCC Emission factors (for manufacturing industries)
to Conversion factors………………………………………………………………...30
Table 4-3: Summary of data……………………………………………………………...32
Table 5-1: Time series estimates by sector……………………………………………….46
Table 5-2: Selected results from Model Selection……………………………………….57
Table 5-3: Time series estimates of Output
it
Y
by sector…………………………………58
Table 5-4: Time series estimates of Energy consumption
it
E
by sector………………….59
Table 5-5: Time series estimates of Employment
it
N
by sector……………………………60
Table 5-6: Selected results for population-averaged model
(models with lagged dependent variables as covariates)…………………...………..69
Table 5-7: Selected results for population-averaged model
(models without lagged dependent variables as covariates)……………………...….70
Table 5-8: Summary of variables………………………………………………………...79
Table 5-9: Number of Cointegrating Relationships for the Individual VARX* Models...80
Table 5-10: VECMX* Estimates of the Individual Models……………………………...81
Table 6-1: CO2 growth rate_ by sector and the total…………………………………...108
Table 6-2: CO2 per capita_ by sector and the total……………………………………..111
Table 6-3: Output_ by sector and the total……………………………………………...113
Table 6-4: Energy intensity_ by sector and the total……………………………………114
Table A-1: Autocorrelation graphs for Energy consumption _ by sector……………….124
Table A-2: Partial Autocorrelation graphs for Energy consumption_by sector………...130
Table A-3: Time series forecast by sector __ Energy intensity…………………………136
Table B-1: Autocorrelation graphs for Output_by sector……………………………….142
Table B-2: Partial Autocorrelation graphs for Output_by sector……………………….148
Table B-3: Time series forecast by sector __ Output…………………………………...154
Table C-1: Autocorrelation graphs for Employment_by sector………………………...160
Table C-2: Partial Autocorrelation graphs for Employment_by sector…………………167
Table C-3: Time series forecast by sector __Employment……………………………..174
Table D-1: Time series forecast __Purchasing price of fuel and raw materials………...180
Table E-1: Weight matrix for GV AR model…………………………………………….181

5
Abstract
This paper conducts a program evaluation of the effects of China’s agglomeration policy
on the CO
2
emissions of its manufacturing sectors. By constructing industrial emission
data for China’s 21 manufacturing sectors from 1985 to 2011, this paper fills the gap in
data availability at China’s sub-industrial level. Results show that the national
agglomeration policy has generally aggravated manufacturing emissions since 2001, due
to a sharp expansion in heavy polluting sectors, and decreased emissions per capita
through urbanization, while energy efficiency declined.

Keywords: CO
2
Emission, Forecasting, Manufacturing industry, Agglomeration,
Program evaluation

6
Chapter 1 INTRODUCTION
Anthropogenic carbon dioxide (CO
2
) emissions are well known as a major source of
the green house gases that have been aggravating climate change and global warming.
China became the world’s number one emitter in 2006
1
, with a total of 6414.5 million
metric tons of CO
2
emissions
2
, and has contributed more than one fifth of the world’s
total emissions ever since
3
. Exempt from Kyoto Protocol’s emission reduction
requirements as a developing country, China finally put forward its emission reduction
pledge at the 2009 Climate Change Conference in Copenhagen: a 40 to 45 percent
reduction in carbon intensity (CO
2
emissions per unit of output) in 2020 compared to
2005
4
. Because the pledge allows the country’s total emissions to grow, albeit at a slower
speed, it has also been seen as too mild of a pledge for such a large global player.
Pressures have mounted on China’s coal-dominated energy consumption structure and in
particular the large primary materials industries, most of which belong to the
manufacturing sectors (Figure 1-1), which have shown significant increases in total CO
2

emissions since 2001 (Figure 1-2).

1
http://search.worldbank.org/quickview?name=%3Cem%3ECO2%3C%2Fem%3E+emissions+%28kt%29&id=EN.AT
M.CO2E.KT&type=Indicators&cube_no=2&qterm=china+co2+emission
2
http://www.pbl.nl/en/dossiers/Climatechange/moreinfo/Chinanowno1inCO2emissionsUSAinsecondposition
3
http://en.wikipedia.org/wiki/List_of_countries_by_carbon_dioxide_emissions
4
UNFCCC Copenhagen Accord: http://unfccc.int/resource/docs/2009/cop15/eng/l07.pdf

7

Figure 1-1: CO
2
emissions of Manufacturing Industry as percentages of China’s total CO
2

emissions

Figure 1-2: CO
2
emissions: China, China’s Manufacturing industry, and the United States
CO2 emissions of Manufacturing Industry as percentage of China’s total CO2 emissions
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
year
CO2 emissions (million metric ton)
0
1000
2000
3000
4000
5000
6000
7000
8000
1991 1993 1995 1997 1999 2001 2003 2005 2007
year
Chinese Manufacturing
industries
China
United States

8
As agglomeration5 is one of the major driving forces of China's economic growth,6
all levels of the Chinese government have been making substantial effort to foster
industrial agglomeration since 2001. In practice, the major national policies promoting
industrial agglomeration are China’s Western Development7 launched in March 2000,
and the Tenth, Eleventh and Twelfth “Five Year Plan”8 (effective since 2001, 2005 and
2010 respectively). Agglomeration policies at the provincial level were also introduced
through various local regulations in all 30 provinces from 2001,9 these mainly included
policies like tax incentives, large government-certified manufacturing projects,10 and
government sponsored industrial zones. Research11 indicates that there has been a surge
in the degree of agglomeration in the manufacturing industry since 2001: the degree of
agglomeration in 2004 was almost double what it was in 2001.12 Agglomeration has
been particularly highly concentrated in the coastal areas of East China.13 While the
literature overwhelmingly agrees upon the positive relationship between economic
growth and agglomeration, there is little research on the impact of agglomeration on
manufacturing emissions, largely because of the lack of industrial sub-sector-level data.

5
High concentration of the same industry in a specific geographic area
6
e.g. Marshall (1920), Martin and Ottaviano (2001) (theoretical) and Fan Jianyong (2008) (empirical)
7
Covered 12 provinces, aimed to alleviate uneven regional growth between East and West China through
agglomerating manufacturing industries in the West provinces and attracting industrial transfers by tax incentives.
8
China’s central government policy on the plan for development, formulated every five years.
9
In next section, this paper takes Guangdong Province as an example to illustrate the concrete steps to implement the
agglomeration policy at provincial level.
10
Majorly the projects on infrastructure construction
11
E.g Wen and Xi, 2012; Lu and Tao 2009.
12
Wen and Xi (2012)
13
He et al (2010)

9
An empirical evaluation of this policy is necessary because theory suggests
agglomeration can influence manufacturing emissions in either direction, which may
result in different strategies when considering the trade-offs between higher degrees of
agglomeration and greater emission reductions. On one hand, industrial agglomerations
may result in increasing local demand for energy inputs and lead to more CO2 emissions;
on the other hand, economies of scale from agglomerations may also bring down the
energy related cost in production (e.g. transportation) and thus a decrease in CO2
emissions. The aim of this paper is to evaluate the impact of China’s agglomeration
policy on its 21 manufacturing sectors’ CO2 emissions from 2001 to 2011.
This paper makes three major contributions: First, it constructs industrial emissions
data for China’s 21 two-digit manufacturing sectors from 1985 to 2011, filling a gap in
the availability of data at the sub-industrial level. Second, this paper explores various
econometric forecasting models with a comprehensive panel data set, and takes the
average across all forecasts in order to eliminate potential forecasting bias from a single
model. Third, the forecasts of CO2 emissions after 2000 are compared with actual data to
evaluate how agglomeration policy affected emissions. The remainder of the paper is
structured as follows: section 2 explores the concrete steps and influences of China’s
agglomeration policy. Section 3 performs a literature review, and section 4 summarizes
the construction of industrial emissions data using thermal energy conversion equations
with IPCC conversion factors. The fifth section discusses model construction and
forecasting; I conclude with a discussion of the program evaluation results.

10
Chapter 2 CHINA’S AGGLOMERATION POLICY
It has been suggested in both theoretical (e.g. Marshall, 1920; Martin and Ottaviano,
2001) and empirical literature (e.g. Fan Jianyong, 2008) that agglomeration has been one
of the major driving forces behind China's economic growth due to the economies of
scale, economies of scope and external economy resulting from having a high
concentration of the same industry in a specific geographic area. Accordingly, all levels
of the Chinese government have been making great efforts to foster industrial
agglomeration since 2001.
14

2.1 Concrete Steps to Promote Agglomeration
In practice, the first national action that promoted industrial agglomeration was
China’s Western Development
15
launched in March 2000, which covered 12 provinces
and aimed to alleviate the uneven regional development between eastern and western
China by agglomerating energy and mineral related manufacturing industries in the West
provinces, as well as attracting domestic and foreign industrial transfers through tax
incentives.
Later in 2001, the “Tenth Five Year Plan”
16
was implemented, for the first time
emphasizing agglomeration policy as a national policy. Aiming to balance urban and rural
development, form new urban agglomerations, and facilitate the formation of new

14
The starting year of “Tenth Five Year Plan”.
15
A policy from central Government that covered 12 provinces.
16
China’s central government policy on the plan for development, formulated every five years.
http://www.chinaemb.or.kr/chn/zgzt/zgjj/t81068.htm

11
industrial cities, the plan focused specifically on the concentration of populations in small
towns and villages, and corporate restructuring of township and village enterprises
towards more industrialized production. In the following “Eleventh
17
and Twelfth
18
Five
Year Plans (effective since 2005 and 2010 respectively), more advanced manufacturing
sectors were introduced, such as clean power generators and transmissions, advanced
transportation equipment, high-end CNC machinery, new materials and energy, etc.
Especially for the equipment manufacturing industry, both five year plans strengthened
support of the policy through government procurement of major construction projects.
19

Starting in 2001, agglomeration policies at the provincial level were gradually
launched in all 30 provinces. The major means by which provinces encouraged
agglomeration included tax incentives for foreign investment and government-certified
manufacturing projects, the establishment of government sponsored industrial zones, and
local investments in large scale infrastructure projects. To provide a concrete illustration
of the steps taken by provincial governments, this paper takes Guangdong province as an
example. As pointed out in Fan and Li (2011),
20
Guangdong province ranked the top
21

out of the seven
22
most agglomerated provinces from 2001 to 2007 in terms of
manufacturing share. In 2003, the Guangdong government issued a provincial policy

17
http://news.xinhuanet.com/ziliao/2006-01/16/content_4057926.htm
18
http://china.huanqiu.com/roll/2011-03/1567363_6.html
19
i.e. government investment in large-scale infrastructure constructions, e.g, the expansion of the eighth loop of Ring
Roads in Beijing.
20
The source used in their paper is “China Annual Survey of Industrial Firms”, issued by China’s National Bureau of
Statistics.
21
Accounts for 15-16% manufacturing shares from 2001 to 2007.
22
The rest of the six most agglomerated provinces are Jiangsu, Zhejiang, Shandong, Henan, Shanghai and Fujian, and
account for 12-14%, 8-10%, 10-14%, 3-6%,6-8% and 4% manufacturing shares respectively from 2001 to 2007.

12
intended to accelerate the new industrialization of Guangdong.
23
The policy emphasized
the vertical integration of automobile manufacturing, distribution and control equipment
manufacturing, special equipment manufacturing, and electronics and communications
equipment manufacturing. This was done by creating government sponsored industrial
zones, and reducing taxes to attract international interest and thus stimulate the industrial
transfer of capital and technology. By 2004, government-certified large scale
manufacturing projects
24
included electronic product manufacturing, civil aerospace
manufacturing, urban rail transit manufacturing (including signaling and control system
manufacturing), and civil aircraft and satellite manufacturing. The 2007 “Guangdong
Province regulations to promote SME
25
development”
26
emphasized that it is the
responsibility of county-level governments to promote the construction of industrial
zones with a focus on the cluster of small to medium enterprises. Most recently, in 2014,
the province issued policies to accelerate the development of advanced equipment
manufacturing
27
and promote industrial restructuring and upgrading through new rounds
of technological innovation
28
, both of which strengthened government support in the
form of more powerful financial support.
29

23
“Accelerating the new industrialization of Guangdong”
http://www.gdei.gov.cn/zwgk/zcfg/gdsfg/200308/t20030801_42432.htm
24
“Catalogue of investment projects approved by the Guangdong Provincial Government”
http://www.gdei.gov.cn/zwgk/zcfg/szfgfxwj/200607/t20060731_51132.htm
25
Short for small to medium enterprise
26
http://www.gdei.gov.cn/zwgk/zcfg/dfsf/200711/t20071106_61722.htm
27
“Accelerating the development of advanced equipment manufacturing”
http://www.gdei.gov.cn/zwgk/zcfg/szfgfxwj/201411/t20141110_113485.htm
28
“Promoting industrial restructuring and upgrading through new rounds of technological innovation”
http://www.gdei.gov.cn/zwgk/zcfg/szfgfxwj/201411/t20141107_113481.htm
29
A total provincial budget of 51.6 billion RMB from 2015 to 2017, which was later confirmed in the 2015 “Industrial
transformation and upgrading three year plan.” For imported manufacturing projects with investment need over one

13
2.2 Agglomeration before and after 2001
Since the 1978 reform, China has made substantial efforts to build a fiscally
decentralized system
30
. While this strategy did boost local growth, the increased
interregional competition between duplicative industries led local governments towards
more local protectionism policies,
31
which thus, led to the fragmentation of the domestic
market and a decrease in the degree of industrial agglomeration.
32
Later in 1990s, the
degree of agglomeration started to increase slowly in some coastal cities, due to the
geographic advantage of being close to the international market.
33

Wen and Xi (2012) calculated the EG agglomeration index
34
for China’s
manufacturing industries from 1998 to 2009. Their data shows there has been a big
surge
35
in the degree of agglomeration in the overall manufacturing industry since 2001.

billion RMB, the local government could intervene to request financial institutions to make loans. The government also
supported manufacturing enterprises by developing State Laboratories for engineering, technology, and innovation
centers.
30
China has made great efforts to break down the previous highly centralized fiscal management system, in order to
promote local economic growth and local governments' responsibilities to carry out local fiscal functions.
31
E.g, impose a variety of interregional barriers to trade.
32
Alwyn Young (2000), “The Razor’s Edge: Distortions and Incremental Reform in the People’s Republic of China”
33
Catin et al (2005), “Openness, Industrialization and Geographic Concentration of Activities in China.” World Bank
working paper, 2005
34
Ellison and Glaeser (1997) presented the agglomeration index
i
γ for industry i: the higher the
i
γ is, the more
agglomerated is industry i.
2
2
[1 ]
[1 ](1 )
iri
r
i
ri
r
GxH
x H
γ
−−
=
−−
∑
∑
, where
i
G is the Space Gini coefficient for industry i,
with
2
()
i
i
rr
r
G x s =−
∑
,
r
x is the ratio of employment in area r out of the country’s total employment, and
i
r
s is the ratio of employment in industry i of area r out of the country’s total employment in industry i.
i
H is the
Herfindahl index for industry i, with
2
()
i
f
i
f
H z =
∑
,
f
i
z is the ratio of employment in firm f of industry i out
of the country’s total employment in industry i.
35
Similar findings are also showed in Lu and Tao 2009.

14
The concentration peaked in 2004 (when the index
36
was almost double the index before
2001
37
), after which the increase gradually slowed and then picked up again in 2008. The
ratio of highly concentrated
38
industries in the overall manufacturing industry was 2.79%
in 1998, 6.74% in 2004, and continued to rise to 10.51% in 2009.
The 2004 employment data shows that manufacturing agglomeration has been
highly concentrated in coastal areas
39
(eastern of China), especially (from south to north)
the Pearl River Delta (PRD
40
) Economic Zone, Yangtze River Delta (YRD
41
) Economic
Zone and the Bohai Bay Economic Rim (BER
42
). The provinces Guangdong, Jiangsu,
Zhejiang and Shandong are the most agglomerated manufacturing areas in the eastern
China; Beijing, Shanghai and Tianjin are the most agglomerated in northern China; Jilin
province is the most agglomerated in northeastern China; and Sichuan, Henan, Hubei and
Hunan provinces are the most agglomerated in the middle of China.
Data
43
collected in this paper also shows that manufacturing output increased at an
annual rate of 20.57%
44
from 2001 to 2011, far exceeding the annual rate of 8.36% from

36

i
γ , ranges from 0.046 to 0.05 in 2004
37

i
γ , which was around 0.025 before 2001
38
“Highly concentrated” industry with 0.05
i
γ > , “very concentrated industry” with 0.02 0.05
i
γ ≤ ≤ , and “not
very concentrated” industry with 0.02
i
γ <
39
He et al, 2010 “Industry characteristics, regional characteristics and Chinese manufacturing industries
agglomeration”
40
Including Guangzhou, Shenzhen, Dongguan, Foshan, Zhongshan, Zhuhai, Jiangmen, and parts of Huizhou and
Zhaoqing.
41
Encompasses Shanghai, Zhejiang and Jiangsu province, and accounts around 20% of China’s GDP.
42
Covers 24 cities and the surroundings, including Beijing, Tianjin,Tangshan, Qinhuangdao, Cangzhou, Langfang,
Chengde, Zhangjiakou, Qingdao, Jinan, Weihai, Weifang, Yantai, Dongying, Binzhou, and Zibo, Shenyang, Dalian,
Dandong, Huludao, Jinzhou, Yingkou, Anshan, and Panjin.
43
Collection and sources are explained in section 4
44
Table 2-1

15
1985 to 2000. Energy intensity (energy consumption per unit of output) also increased
rapidly since 2001 at a yearly rate of 8.58%
45
while the rate was only 4.20% during 1985
to 2000. CO
2
emissions since 2001 have also grown at a much faster annual speed of
10.01%
46
compared to the previous 3.41%. While the literature overwhelmingly agrees
upon the positive relationship between economic growth and agglomeration, there is little
empirical research on the impact of agglomeration policy on manufacturing CO
2

emissions, partly due to the lack of public data collected on the sub-sector emissions. This
paper aims to fill this gap.

45
Table 2-1
46
Table 2-1

16

CO
2
growth rate Output growth rate Energy intensity growth rate
sector 1985‐2000 2001‐2011 actual 1985‐2000 2001‐2011 actual 1985‐2000 2001‐2011 actual
1 1.13% 3.11% 7.19% 20.04% 1.98% 5.04%
2 ‐1.64% 4.20% 2.60% 15.82% 0.24% 8.22%
3 4.52% 5.09% 9.29% 14.57% 7.76% 8.10%
4 ‐1.46% ‐0.72% 12.43% 15.55% 2.56% 6.56%
5 ‐0.14% 7.35% 9.28% 24.68% 1.71% 12.28%
6 ‐1.55% 2.00% 6.14% 24.22% 0.82% 7.22%
7 2.08% 9.50% 8.84% 17.47% 2.33% 6.77%
8 2.23% ‐2.58% 5.62% 14.80% 6.63% 6.43%
9 2.34% ‐2.11% 11.25% 13.44% 7.22% 5.67%
10 5.63% 10.84% 12.97% 19.65% 10.54% 8.60%
11 2.77% 7.54% 8.66% 21.85% 3.04% 9.98%
12 1.87% 4.76% 9.90% 18.53% 3.52% 5.29%
13 5.75% ‐10.05% 9.63% 17.18% 8.23% ‐1.56%
14 ‐1.58% 5.28% 4.21% 19.88% 2.88% 8.39%
15 2.19% 6.40% 9.61% 18.48% 6.67% 11.51%
16 1.41% 8.35% 8.18% 22.26% 1.60% 10.13%
17 4.56% 12.67% 8.98% 23.70% 5.71% 12.55%
18 6.05% 14.53% 9.46% 27.47% 7.83% 13.20%
19 ‐0.02% 0.19% 8.90% 19.86% 2.80% 10.77%
20 ‐1.72% 7.52% 9.58% 21.17% 1.28% 11.58%
21 ‐3.52% 2.99% 10.15% 16.99% 1.86% 2.61%
total 3.41% 10.01% 8.36% 20.57% 4.20% 8.58%
Table 2-1: Calculation of data collected in section 4 for 21 manufacturing sector

17
Chapter 3 LITERATURE REVIEW
This section starts with a review of the relationship between agglomeration and
economic growth and the relationship between agglomeration and environmental issues,
and then summarizes the literature on the advantages and disadvantages of various
emission modeling techniques.
The earliest research on agglomeration can be traced back to Alfred Weber’s
industrial location theory proposed in the 19
th
century. Research after the 1950s has
focused on exploring the relationships between agglomeration and the growth of
economies (e.g. Ciccone（2002）, Brulhart et al.（2008）). Studies have also examined
reasons for the huge increases in the degree of agglomeration for the Chinese
manufacturing industry and its relationship with economic growth: Liu and Xu (2009)
showed that political incentives, geographical advantages and convenient transportation
are crucial factors that facilitated China's industrial accumulation. Liu, Jun and Xu,
Kangning (2010) used 1999-2007 provincial panel data to show that industrial
agglomeration promoted economic growth and at the same time led to significant
regional disparities.
While it is largely agreed in the literature that agglomeration has promoted
economic growth in China, there are disagreements on the relationships between China’s
agglomeration and its impact on environmental issues: Ho (2008) found out that the
regional agglomeration leads to fast urbanization and aggravated resource depletion in
and outside the cluster area. Using 2008-2010 data for Tianjin’s petroleum industry, Li

18
(2012) pointed out that industrial agglomeration changed the composition of energy use
in a way that should have led to emissions reductions. Similar thoughts are echoed in
Wang et al (2010). While some research has been done on the relationship between
agglomeration and the composition of energy use (e.g. Li (2012)), the relationship
between agglomeration and CO
2
emissions in China’s manufacturing industry has yet to
be determined. A major reason for this is likely the lack of data at the sub-industry level.
This paper constructs industrial emissions data for 21 two-digit manufacturing sectors
using IPCC emission factors in order to carry out further analysis at the sub-industry and
whole manufacturing level.
As this paper uses various forecasting models to conduct a program evaluation
analysis, I also summarize the literature that exploits the modeling of CO
2
emissions.
Decomposition methods, CGE modeling, and econometric models are the most widely
adopted.
The simplest decomposition model is the Kaya method (Yoichi Kaya, 1993), which
decomposes CO
2
emissions as the product of: carbon emission intensity (CO
2
emissions
over energy use), energy intensity (energy use over GDP), GDP per capita, and
population. Several papers use this method, including Pan, Shu and Xu (2011), who
concluded the carbon emission intensity of China’s manufacturing industry displayed an
overall downward trend during 1996-2007 which was caused by enhanced energy
efficiency. Structure, meanwhile, also had a dampening effect.

19
An extended form of the Kaya method developed by Ang, Zhang and Choi (1998)
and known as LMDI (Logarithmic Mean Divisia Index),
47
has been widely adopted.
Using this method, Ma and Stern (2007) found that the decrease of biomass consumption
has contributed significantly to China’s carbon emission reductions, and technology has
had a negative effect during the post-reform period in compared to the pre-reform time;
Sun (2011) concluded that compared to structural change, decreased energy intensity
resulting from technological progress is the key factor that reduced emissions.
Structural decomposition analysis (SDA)
48
is also widely adopted in energy and
emission studies. Xu and Xue (2012) used SDA input-output structural decomposition
and found that the scale effect contributes the most to industries’ emissions and cannot be
offset by the adoption of reduction technology, while the structural effect does not play an
important role. Pan, Shu and Xu (2011) found the structural change in China’s
manufacturing industry had a dampening effect on emissions during 1996 to 2007. While
these decomposition methods are useful for identifying and examining the impacts of
driving forces of emissions in the short run, it does not construct the regression
relationships between emissions and these components and thus lacks forecasting ability.
Another popular method is CGE modeling, which makes use of data from SAM
(social accounting matrix) and methods of mathematical programming to evaluate how an
economy might react to changes in exogenous factors (e.g. policies, technological

47
E.g. the energy consumption change is decomposed of three explanatory effects: activity effect (changes in
aggregates associated with changes in overall level of the activity), structure effect (changes in aggregates associated
with changes in mix of the activity by sub-category) and intensity effect (change in aggregates associated
with changes in sub-category energy intensities).
48
Analytical technique identifies and quantifies the components in emissions.

20
progress). It is a simulation system based on equations of cost-minimizing producers,
optimizing households, taxation and government balance, and international trade and
closure rules. Parameters from SAMs are used as inputs to the model, however, due to the
lack of recent data from China, most work has used input output tables from the late
1990’s and early 2000’s. Garbaccio et al. (1999) found that China’s increased energy
intensity from 1987 to 1992 was driven by structural change: industry shifted to more
heavy polluting sectors (e.g. large scale coal mining); Cao, Garbaccio and Ho (2009)
concluded that the aggregate economy benefited from shutting down small power plants
which offset the costs of desulfurization equipment. Although CGE modeling is useful in
evaluating the effectiveness of policy instruments, it is not a forecasting model due to the
lack of large number effect in neither space nor time, not to mention the tremendous data
needs and the sensitivity of equilibrium existence on assumptions of function forms and
parameter calibrations.
Various econometrics methods provide researchers many ways to
construct forecasting models, though a number of previous papers focus mainly on the
topic of the Environmental Kuznets Curve (EKC), which is a reduced form model of the
relationship between CO
2
emissions and outputs, mostly claimed to be an inverted U
shape. Shen and Hashimoto (2004) used China’s provincial panel data to examine the
relationship between per capita income and seven pollutants
49
, and concluded that the

49
Two air pollutants: SO2 and dust fall (data from 1990-2001); four water pollutants: chemical oxygen, arsenic,
cadmium and mercury (data from 1993-2001) and one solid pollutant: industrial waste stock (data from 1993-2001).

21
inverted U shape exists in five pollutants
50
, but the other two exhibit an N-shape. Jalil and
Mahmud (2009) investigated the time series from 1975 to 2005 and supported the EKC
hypothesis for CO
2
emissions and per capita income, where trade has had an insignificant
impact on emissions in the long run. A model selection process was introduced by
Auffhammer and Carson (2007), where they used 1985-2004 province level data and
tested the performance of different economic models. Their best model rejected the EKC
hypothesis and the projection suggested that the increase of China’s total emissions in
2015 will be several times what is set in the Kyoto Protocol. There are, however, three
major shortfalls in their paper: (1) First, the paper assumes that CO
2
emissions of each
province follow the same growth pattern. Due to the lack of emissions data at provincial
level, the paper resorts to waste gas emissions (WGE). They first aggregate the provincial
level WGE to a national level time series, and then regress national CO
2
emissions on
national WGE. The rest of the paper assumes that the regression relationship between
national CO
2
and national WGE holds for each province. By doing so, they neglect
possible cross-province heterogeneity which could lead to modeling bias. Second, when
making out-of-sample forecasts, the paper further assumes that the GDP growth path of
each province is homogeneous. They list three growth scenarios based on which future
emissions are projected: each province with a slow growth rate (3.04%), with a medium
growth rate (5.04%), or with a high growth rate (7.04%). Again, not allowing for
cross-province heterogeneity in GDP growth is a strong assumption. Third, while it might

50
SO2, chemical oxygen, arsenic, cadmium and mercury

22
be true that the chosen model best fits the sample data in their paper, it is not clear that
the chosen model will also outperform the other models for the whole population. This
paper will avoid these weaknesses by (1) adopting various modeling techniques, rather
than just focusing on economic models, and taking the simple average of forecasts from
different models to improve the forecasting results as suggested in Clemen (1989), and (2)
taking into account cross-industry heterogeneity and interdependence when applicable.

23
Chapter 4 DATA
4.1 Sources, Availability and Horizon
This paper uses a panel data set of China’s 21 two-digit manufacturing sectors
51

(categorized by National Bureau of Statistics of China
52
) from 1985 to 2011. The data are
mainly from five public sources: (1) China Statistical Yearbook (1985--2012), (2) China
Industrial Statistics Yearbook (1988--2012), (3) China Energy Statistical Yearbook
(1991—1993, 2008-2011), (4) “50 years of Industrial and Transportation Energy
Statistics Compilation 1949—1999”, and (5) “2006 IPCC Guidelines for National
Greenhouse Gas Inventories Vol.2 Energy”. The first four resources were published by
National Bureau of Statistics of China, and the last one was published by the
Intergovernmental Panel of Climate Change. The names of the 21 combined sectors are
listed in table 4.1.

51
Originally there are 30 sub-sectors under the manufacturing industry (manufacturing industry is one of the 20
one-digit industries categorized by National Bureau of Statistics of China). This paper combined those sectors into 21
sectors due to the changes of calibration in different years in the statistical yearbooks.
52
http://www.stats.gov.cn/tjbz/hyflbz/

24
Sector Name
1 (1) Processing of Food from Agricultural Products
(2) Manufacture of Foods
(3) Manufacture of Beverages
(4) Manufacture of Tobacco
2 Manufacture of Textile
3 Manufacture of Textile Wearing Apparel, Footwear and Caps
4 Manufacture of Leather, Fur, Feather and Related Products
5 Processing of Timber, Manufacture of Wood, Bamboo, Rattan, Palm and Straw Products
6 Manufacture of Furniture
7 Manufacture of Paper and Paper Products
8 Printing, Reproduction of Recording Media
9 Manufacture of Articles For Culture, Education and Sport Activity
10 Processing of Petroleum, Coking, Processing of Nuclear Fuel
11 Manufacture of Raw Chemical Materials and Chemical Products
12 Manufacture of Medicines
13 Manufacture of Chemical Fibers
14 Manufacture of Rubber
15 Manufacture of Plastics
16 Manufacture of Non-metallic Mineral Products
17 Smelting and Pressing of Ferrous Metals
18 Smelting and Pressing of Non-ferrous Metals
19 Manufacture of Metal Products
20 (1) Manufacture of General Purpose Machinery:
(2) Manufacture of Special Purpose Machinery
(3) Manufacture of Transport Equipment
(4) Manufacture of Electrical Machinery and Equipment
(5) Manufacture of Communication Equipment, Computers and Other Electronic Equipment
(6) Manufacture of Measuring Instruments and Machinery for Cultural Activity and Office Work
21 Manufacture of Artwork and Other Manufacturing
Table 4-1: Names of 21 combined sectors

25
4.2 Method of Constructing CO
2
emissions data at sub-sector level
4.2.1 Methods in literature and Method adopted in this paper
So far there is little public data collected for China’s CO
2
emissions at the two-digit
industry level. Traditionally, emissions are calculated using technology and production
based calculations, which are conducted by monitoring emissions from each production
step in a certain industry and considering possible emission reductions from adopting low
carbon technologies. The advantage of this type of calculation is perfect accuracy, but the
methods of calculation are different across industries and regions and even different
among companies in the same industry and in the same region, thus posing huge
difficulties and complexity for researchers. Most research that adopts this method usually
focuses on a specific sector or a firm, and the process of developing calculation
procedures usually involves interdisciplinary input from physics or engineering related
areas. Hasanbeigi, Price, Kong, and Arens (2013) compiled global data on energy
efficiency and low-carbon technologies for iron-making, pulp and paper industry; Wise,
Smith, Sinha and Lurz (2007) examined the U.S industrial sector response to climate
policies by modeling energy technology and fuel choices over a 100 year period; ICFPA
2005 developed industry-specific emission calculation tools for pulp and paper mills in
conjunction with various GHG accounting protocols; Vos and Newell (2009) compared
CO
2
emissions in coated paper production between China and U.S with a focus on the
supply chain, transportation and import structure. Other related resources are the EPA

26
2010 and IEA 2007, which have summaries of multiple sectors’ energy efficiency
conditions.
Due to the lack of complete emissions calculation methods for every manufacturing
sector in the literature, this paper uses the IPCC emission factors
53
and data on fuel
consumption to construct the emissions data for China’s 21 two-digit manufacturing
sectors.
54
There are two main advantages to using the IPCC factors. First, the factors are
available for all nine major types of energy
55
consumed in China’s manufacturing
process. We are thus able to construct the manufacturing emission data for all 21 sub
sectors. Second, the factors are highly reliable because they are computed for each type
of fuel based on physics laws of energy conversion during combustion. The disadvantage
is that the National Bureau of Statistics of China has not published energy consumption
data from renewable energy and clean energy, making the energy consumption data
incomplete. However, there are a number of reasons why this may not, in practice, be a
source of important inaccuracies. According to Li et al (2008), the process of adopting
renewable and clean energies in China’s manufacturing sectors has been very slow, and
by 2007, the energy use from three types of traditional fuels (coal, fuel oil and natural gas)
still accounted for 92.54% of the total energy use in manufacturing industry. Moreover,
renewable energy and clean energy will certainly exhibit lower than normal emissions
compared to the traditional energy sources. Thus, given the minimal amount of new types

53
Published in 2006 IPCC Guidelines for National Greenhouse Gas Inventories
54
Lu and Price (2012) constructed the data for the year 2008, yet only for nine provinces in China.
55
Coal, coke, crude oil, gasoline, kerosene, diesel oil, fuel oil, natural gas, electricity

27
of energy involved and its lower-than-traditional per unit emissions, the following
calculations should be close estimates of the actual industrial emissions.

4.2.2 Fuel consumptions and missing data
China Statistical Yearbook (1988, 1996--2012) and “50 years of Industrial and
Transportation Energy Statistics Compilation 1949—1999” provide energy consumption
data for nine fuels (coal, coke, crude oil, gasoline, kerosene diesel oil, fuel oil, natural gas,
electricity) for each sector in the years 1985, 1986 and 1994-2011. The China Energy
Statistical Yearbook (1991--1993) provides the rest data for the years 1991—1993. The
missing data from 1987 to 1990 are interpolated using the piecewise-cubic Hermite
method
56
in Matlab.
57
This method is used because the time horizon of data for this
paper is short (a panel of 27 year with 21 sectors) and cubic interpolations are widely
used in numerical analysis for constructing missing data points within the range of
a discrete set of known data points. Hermite cubics have the least stringent requirements
for the continuity of derivatives (though the derivatives do need to be specified) and are
thus more conveniently applied.

56
An interpolation method to construct new data points within a range of a discrete set of known data points, using an
interpolant based not only on equations for the function values, but also for the derivatives (Hermite conditions),
usually not required to be twice continuously differentiable.
57
See PCHIP in Matlab

28
4.2.3 Conversion factors
The 2006 IPCC Guidelines for National Greenhouse Gas Inventories published
emission factors
58
by fuel type based on the thermal unit in the measure of
Kg(CO
2
)/TJ(Heat). However, we need to translate the emission factors into applicable
conversion factors with a quantity based unit Kg(CO
2
)/Kg(Fuel) (or with a space measure
Kg(CO
2
)/m3(Fuel)) to match the measurements for fuel data. Therefore, specific calorific
values that vary across territories and across fuel types need to be applied. China Energy
Statistical Yearbook (2008) displays the calorific values for mainland China in the unit of
Kcal(Heat)/kg(Fuel) and Kcal(Heat)/m3(Fuel). Thus, emission factors by IPCC can be
transformed into applicable conversion factors using the formula below:

(KgCO2/KgFuel)
(KgCO2/TJheat)*4186.8*10 ^ ( 12)
* (KcalHeat/KgFuel)
fuel type
fuel type
fuel type
Conversion Factor
IPCC Emission Factor
Calorific Value
=
−

where TJ is short for terajoule, a thermal unit that equals 10^(12) joules; 4186.8 is the
thermal power unit conversion between Joule (J) and Kilocalories (Kcal), with
1Kcal=4186.8J.
Then, for each year, the emissions from a certain sector for each fuel type are
calculated by multiplying energy consumption with the conversion factor. Total emissions
for a given sector are calculated by adding up emissions from all types of fuel usage:

58
http://www.ipcc-nggip.iges.or.jp/public/2006gl/pdf/2_V olume2/V2_2_Ch2_Stationary_Combustion.pdf

29

sec ,
sec , , sec , ,
*
tori yeart
tor i fuel type year t tor i fuel type year t
fuel type
Emissions
Fuel Consumption Conversion Factor
=
∑

The only ‘missing’ IPCC factor (and thus missing conversion factor) is the one for
electricity. The reason why IPCC doesn’t report it is explained in its “detailed sector split
for stationary combustion,”
59
where IPCC indicates that its emission factors for fuel
consumptions have also taken into account the emissions from electricity generated
through fuel powers and consumed by the sectors. It does not, however, specify how
IPCC identifies the different sources of electricity consumed by manufacturing sectors
(i.e. whether it is from thermal power, hydroelectricity, nuclear power, wind power or
solar power), as different resources could have led to very different conversion factors to
CO
2
emissions.
60
Table 4-2 shows the results and calculation of the conversion factors:

59
http://www.ipcc-nggip.iges.or.jp/public/2006gl/pdf/2_V olume2/V2_2_Ch2_Stationary_Combustion.pdf Page 8.
60
E.g. the conversion factor should be huge for thermal power electricity by the use of coal, oil or natural gas
combustion, and should be nearly zero for hydroelectricity.

30
Fuel
Type
Emission Factors for Stationary Combustion in Manufacturing Industries
(from kgCO
2
/TJ to kgCO
2
/kg(or m3))

IPCC 2006 CO
2

Emission Factors
61

95% Confidence
Interval
62

Calorific Value:
Mainland China
Emission Factor (=IPCC
factor*4186.8*10^(-12)*
Calorific Value)
95%
Confidence
Interval
63

IPCC
Emission
Factors
Unit Lower Upper
Calorific
Value
Unit
Emission
Factor
Unit Lower Upper
Coal 94600 kgCO
2
/TJ 87300 101000 5000 Kcal/kg 1.98 kgCO
2
/kg 1.83 2.11
Coke 107000 kgCO
2
/TJ 95700 119000 6800 Kcal/kg 3.05 kgCO
2
/kg 2.72 3.39
Crude
Oil
73300 kgCO
2
/TJ 71000 75500 10000 Kcal/kg 3.07 kgCO
2
/kg 2.97 3.16
Gasoline 69300 kgCO
2
/TJ 67500 73000 10300 Kcal/kg 2.99 kgCO
2
/kg 2.91 3.15
Kerosene 71900 kgCO
2
/TJ 70800 73700 10300 Kcal/kg 3.10 kgCO
2
/kg 3.05 3.18
Diesel
Oil
74100 kgCO
2
/TJ 72600 74800 10200 Kcal/kg 3.16 kgCO
2
/kg 3.10 3.19
Fuel Oil 77400 kgCO
2
/TJ 75500 78800 10000 Kcal/kg 3.24 kgCO
2
/kg 3.16 3.30
Natural
Gas
56100 kgCO
2
/TJ 54300 58300 9310 Kcal/m3 2.19 kgCO
2
/m3 2.12 2.27
Table 4-2: IPCC Emission factors (for manufacturing industries) to Conversion factors

61
2006 IPCC Guidelines for National Greenhouse Gas Inventories Table 2.3
http://www.ipcc-nggip.iges.or.jp/public/2006gl/pdf/2_V olume2/V2_2_Ch2_Stationary_Combustion.pdf
62
The 95% Confidence Intervals are given in 2006 IPCC Guidelines for National Greenhouse Gas Inventories Table
2.3, description of which is on Page 40: “Data are given as upper and lower bounds of the 95 percent confidence
interval, and expressed as percent relative to the mean value.”
63
Same as foot note 25.

31
4.3 Other data
The rest of data are total output (by sector), employment (by sector), energy
consumptions (by fuel type and sector), and purchasing price of fuel and raw materials,
all of which range from year 1985 to 2011 and are adjusted for inflation wherever needed
using 1990 as the base year (when RMB=100) in accordance with the base year chosen in
most of the statistical year books. The data is summarized in Table 4-3:

32
Variable Mean Std.dev Min Max Observations
2
it
CO
CO
2
emissions
overall 111.70 258.20 0.61 1935.01 N=567
(million metric
ton)
between 208.11 1.01 719.83 n=21
within 159.20 -590.01 1326.88 T=27
it
E
Total energy
use
overall 4346.53 7766.75 37.44 58896.58 N=567
(10^4 ton
standard coal)
between 6321.15 115.04 23942.25 n=21
within 4711.80 -17863.40 39300.86 T=27
it
Y
Total output
overall 3448.051 8658.33 65.48 100623.4 N=567
(10^8 RMB: between 5231.99 437.12 24410.71 n=21
1990 as base
year)
within 6989.308 -17789.87 79660.74 T=27
it
N
Employment
overall 253.27 375.39 6.30 2949.23 N=567
(10^4 persons) between 353.93 38.07 1655.73 n=21
within 146.31 -514.46 1546.77 T=27
t
P
Purchasing
price
overall 206.10 97.46 25.70 379.3 N=567
of fuel and raw
materials
between 0.00 206.10 206.10 n=21
(1990=100) within 97.45 25.70 379.3 T=27
Table 4-3: Summary of data

33
Chapter 5 MODELS AND FORECASTS
5.1 Motivation for various modeling and averaging
As pointed out in the literature review, while a model could represent the best local
fit as long as it has the best in-sample and out-of-sample performance, it does not
necessarily represent the best fit for the whole population. In order to construct a better
forecast, Bates and Granger (1969) considered the combination of a pair of forecasts with
optimal weights. However, Clemen (1989) surveyed a large number of papers in the past
literature and pointed out that the simple averaged forecasts turned out to be more robust
procedures in practice than the optimal combination suggested by Bates and Granger
(1969). Therefore, rather than focusing on just one type of modeling solution or searching
for the optimal weights, this section will adopt various modeling methods using the data
from 1985-2000, and then take the simple average of the forecasts in order to circumvent
the disadvantages of using a single model. This average will serve as the basis of the
evaluation of the effects of agglomeration policy on emissions from 2001 to 2011.
This section starts with time-series modeling (5.2) for each industry. As indicated in
Li et al (2008), the process of adopting renewable and clean technologies in China’s
manufacturing sectors has been very slow and therefore, the time-series relationships
within each industry before 2001 should be relatively stable given no significant policy
changes by 2001. Time-series is also the most straightforward way of modeling. However,
in order to examine the contributing factors of CO
2
emissions, this paper needs to
construct economic models (5.3) which not only identify and quantify each influencing

34
factor, but can also take into account cross-industry heterogeneity. The disadvantage of an
economic model is that the cross-industry coefficients may contain information about the
average differences between industries (known as the population-averaged effect). For
example, the reason for high emissions from chemical materials manufacturing could be
because its energy structure uses more heavy polluting fuels (e.g. coal, coke and crude
oil). The relatively low emissions from manufacturing of printing and reproduction of
recording media could be due to of the use of electricity rather than dirty traditional fuels.
Thus, the third model in the paper is a population-averaged model (5.4). One of the
disadvantages from all previous models is that there is no consideration of cross-industry
dependence. For example, the output from manufacturing of leather, fur and feather may
greatly influence the manufacturing of furniture and wearing apparel. Therefore, this
paper then considers a fourth model of global vector autoregression approach or GVAR
(5.5) in order to capture the possible interdependencies across industries. As mentioned
before, the accuracy of the forecasts generated by a single model can vary dramatically
when applied to the whole population than in the sample data, so this paper then takes
average of forecasts from all the four models in order to eliminate disadvantages of using
a single model.

35
5.2 Time-series model
The most straightforward way of modeling as mentioned in 5.1 is to fit a simple
ARIMA model for each industry using the data from 1985 to 2000, where
() 2 ( )
ii
pq
iiti it
LCO L ρ θε = ,
with
2
12 ,
() 1 ...
ii
i
pp
iii ip
L LL L ρρρ ρ =− − − −
and
2
12 ,
() 1 ...
ii
i
qq
iii iq
L LL L θθθ θ =+ + + +
2
it
CO is the emission for industry i (i=1 to 21) in year t (t=1985 to 2000),
it
ε is a
white-noise disturbance;
i
p is the lag of autoregressive part of emissions for industry i,
and
i
q is the lag of moving average part for industry i.
The paper uses the Dickey-Fuller test for unit root to check if the emission series
(for each sector) is stationary. The tests show that, except for sector 4 and 5, emissions
from all the other sectors display non-stationarities. Therefore, the paper takes the first
difference for the non-stationary emission series, and tests for the existence of unit root
again. All first-differenced emission series are stationary. The next step is to find the
autocorrelation and partial autocorrelation for each sector in order to construct ARIMA
models respectively. Figure 5-1 and 5-2 display the autocorrelation and partial
autocorrelation graphs for all sectors. Given the estimates of each ARIMA model (shown
in table 5-1), the paper conducts the forecasts by sector from 2001 to 2011 (given in
Figure 5-3), and then aggregates them to obtain forecasts of total emissions (Figure 5-4)

36
-1.00 -0.50 0.00 0.50 1.00
A u toco rrelatio ns of D .em issio n7
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A u toco rrelatio ns of D .em issio n8
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A utoco rrela tio ns o f D .em issio n9
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rrelatio n s of D .em issio n1 0
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

37
-1.00 -0.50 0.00 0.50 1.00
A utoco rrelatio ns of D .em issio n 1 1
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rrelatio ns of D .em issio n 1 2
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rre lation s o f D .e m ission 13
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-1.00 -0.50 0.00 0.50 1.00
A utoco rre lation s o f D .e m ission 14
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

38
-0.50 0.00 0.50
A utoco rrelatio ns of D .em issio n 1 5
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-1.00 -0.50 0.00 0.50 1.00
A utoco rrelatio ns of D .em issio n 1 6
1 2 3 4 5
-1.00 -0.50 0.00 0.50 1.00
A utoco rre lation s o f D .e m ission 17
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rre lation s o f D .e m ission 18
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

39
-0.50 0.00 0.50
A utoco rrelatio ns of D .em issio n 1 9
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rrelatio ns of D .em issio n 2 0
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rre lation s o f D .e m ission 21
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

Figure 5-1: Autocorrelation graphs by sector
64
_ CO
2
emissions

64
Except for sector 4 and 5, all the other sectors’ graphs are done on the first difference of emissions.

40
-3.00 -2.00 -1.00 0.00 1.00
P artia l a u to co rre la tio ns of D .e m ission 1
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P artia l a u to co rre la tio ns of D .e m ission 2
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-1.00 0.00 1.00 2.00 3.00
P artia l a u to co rre la tio ns of D .e m ission 3
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
0.00 2.00 4.00 6.00 8.00
P a rtia l au to correlations o f em issio n 4
1 2 3 4 5 6
Lag
95% Confidence bands [se = 1/sqrt(n)]

41
-1.00 -0.50 0.00 0.50
P artia l a u to co rre la tio ns of D .e m ission 5
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P artia l a u to co rre la tio ns of D .e m ission 6
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-1.00 -0.50 0.00 0.50
P artia l a u to co rre la tio ns of D .e m ission 7
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50 1.00 1.50 2.00
P artia l a u to co rre la tio ns of D .e m ission 8
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

42
-1.50 -1.00 -0.50 0.00 0.50
P artia l a u to co rre la tio ns of D .e m ission 9
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P artia l au to corre la tions of D .e m ission 10
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.50 0.00 0.50
P artia l au to corre la tions of D .e m ission 11
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-8.00 -6.00 -4.00 -2.00 0.00
P artia l au to corre la tions of D .e m ission 12
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

43
-0.50 0.00 0.50
P artia l au to corre la tions of D .e m ission 13
1 2 3 4 5
-0.50 0.00 0.50 1.00 1.50
P artia l au to corre la tions of D .e m ission 14
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.50 0.00 0.50
P artia l au to corre la tions of D .e m ission 15
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-1.50 -1.00 -0.50 0.00 0.50 1.00
P artia l au to corre la tions of D .e m ission 16
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

44
-0.50 0.00 0.50
P artia l au to corre la tions of D .e m ission 17
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P artia l au to corre la tions of D .e m ission 18
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-2.00 -1.50 -1.00 -0.50 0.00 0.50
P artia l au to corre la tions of D .e m ission 19
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-6.00 -4.00 -2.00 0.00
P artia l au to corre la tions of D .e m ission 20
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

45
-0.50 0.00 0.50
P artia l au to corre la tions of D .e m ission 21
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

Figure 5-2: Partial autocorrelation graphs by sector
65
_ CO
2
emissions

65
Except for sector 4 and 5, all the other sectors’ graphs are done on the first difference of emissions.

46

The ARIMA model and estimates by sector are given below:
ARIMA(p,d,q) Variables Coefficient Sigma ARIMA(p,d,q) Variables Coefficient Sigma
Sector 2 (1, 1, 0) AR(1) 0.71 4.14 Sector 13 (0, 1, 0) Constant 7.06 7.06
(0.202)*** (0.695)*** (0.512)*** (0.512)***
Sector 2 (0, 1, 0) Constant -0.58 3.9 Sector 14 (0, 1, 1) MA(1) 0.95 0.85
(1.049)* (0.811)*** (0.419)*** (0.192)***
Sector 3 (4, 1, 0) AR(4) -0.45 0.27 Sector 15 (1, 1, 0) AR(1) 0.51 0.51
(0.276)* (0.057)*** (0.262)** (0.095)***
Constant 0.127 Constant 0.04
(0.059)** (0.294)*
Sector 4 (0, 0, 0) Constant 2.87 0.68 Sector 16 (1/2, 1, 0) AR(1) 1.03 9.01
(0.298)*** (0.109)*** (0.379)*** (1.450)***
Sector 5 (3, 0, 1) AR(3) 0.4 0.51 AR(2) -0.45
(0.206)** (0.123)*** (0.341)*
Sector 6 (0, 1, 0) Constant 0.12 0.12 Constant 2.67
(0.022)*** (0.022)*** (5.724)
Sector 7 (0, 1, 0) Constant 2.95 2.95 Sector 17 (0, 1,1) MA(1) 0.46 43.68
(0.377)*** (0.377)*** (0.482)* (10.027)***
Sector 8 (0, 1, 0) Constant 0.19 0.19 Constant 15.32
(0.267)*** (0.267)*** (44.607)**
Sector 9 (0, 1, 1) MA(1) -0.94 0.21 Sector 18 (2, 1, 0) AR(2) 0.38 2.8
(1.366)* (0.034)*** (0.344)* (0.497)***
Constant 0.04 Constant 1.12
(0.009)*** (1.668)
Sector 10 (1, 1, 0) AR(1) 0.21 115.61 Sector 19 (0, 1, 0) Constant 1.05 1.05
(0.800)* (29.166)*** (0.164)*** (0.164)***
Constant 24.71 Sector 20 (0, 1, 1) MA(1) -0.46 14.72
(127.270) (0.190)*** (2.320)***
Sector 11 (3, 1, 0) AR(3) -0.43 24.15 Constant -0.685
(0.4.4)** (4.796)*** (2.450)*
Constant 7.61 Sector 21 (0, 1, 0) Constant 6.14 6.14
(8.848)* (0.978)*** (0.978)***
Sector 12 (1, 1, 0) AR(1) 0.08 1.92
(0.286)* (0.336)***
Constant 0.17
(0.802)*
Table 5-1: Time series estimates by sector

47

20 40 60 80 100
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission1
20 30 40 50 60
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission2

2 3 4 5 6 7
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission3
2 3 4 5
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission4

48
4 6 8 10
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission5
.8 1 1.2 1.4 1.6
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission6

20 40 60 80 100
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission7
1.2 1.4 1.6 1.8 2 2.2
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission8

49
.5 1 1.5 2
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission9
0 500 1000 1500 2000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission10

200 300 400 500 600
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission11
5 10 15 20
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission12

50
10 20 30 40
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission13
6 8 10 12
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2002) emission14

4 6 8 10 12
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission15
200 300 400 500 600
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission16

51
0 500 1000 1500
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission17
0 50 100 150
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission18

10 12 14 16
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission19
40 60 80 100 120
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) emission20

52
10 20 30 40
1985 1990 1995 2000 2005 2010
year

Figure 5-3: Time series forecasts by sector __ CO
2
emission

Except for sector 9 and 13, the forecasted emissions from all sectors are smaller
compared to the actual emissions since 2001. We aggregate the 21 sectors and plot the
trend for total emissions (Figure 5-4). In general, the forecasted emissions display an
increasing trend yet at a slower speed than the actual emissions.

53
Total CO2 emissions (million metric ton)
1000 2000 3000 4000 5000
1985 1990 1995 2000 2005 2010
year
ACTURAL DATA Time series model

Figure 5-4: Time series forecast in comparison with the actual data__ Total CO2 emission

5.3 Economic model
As discussed in 5.1, an economic model is useful for identifying and quantifying
each contributing factor of emissions. Based on previous literature and a careful
investigation of China’s manufacturing process, this section constructs the economic
model as below:

2
11 2 1
ln 2
ln2 ln ln ln lnlnln( )_
it
it it it it it t it
co
co y y y e P year con α ββ γ λ κ χ ε
−−
=
++ + + + + + +

54
where
2
it
co
is emission per worker;
it
e
is energy intensity (the ratio of energy use and
total output);
it
y
is real total output per worker and
t
P
is the real purchasing price of raw
materials and fuels, both in terms of the 1990 value, to be in line with statistical year
books.
The different ways of modeling the relationship between emissions per worker
2
it
co and output per worker
it
y poses a reexamination on the EKC hypothesis. So far
no consensus has been reached from the many EKC papers on the inverted U shape
(which would imply a quadratic term of
it
y involved, as in Jalil and Mahmud 2009).
Therefore, a polynomial and lagged structure of
it
y is considered here to capture the
many possibilities, and the model selection process shows that a linear term and a lagged
term of
it
y in this model better fits the data.

it
e represents energy efficiency (the ratio of energy consumption and total output),
which should therefore affect emissions in a positive way.
t
P may affect domestic
emissions in a positive way, as the high purchasing price reflects high demand for fuel
and raw materials, and thus high volume of fuel combustion and emissions. The time
trend ln( ) year captures the possible technological progress, so it should influence
emissions in a negative way.
The error term
it
ε is assumed to be independent identically distributed. Since the
paper allows serial correlations in both dependant and independent variables to capture
the possible time lag effects, Arellano-Bond estimators are applied in these regressions to
generate consistent estimators and the effectiveness of moments used by the GMM type

55
estimators are tested to guarantee consistency. Given the assumption of i.i.d. errors, the
first differenced errors at order 1 are serially correlated. However, no serial correlations
should be found in the differenced errors at orders higher than one if the model
specification is correct. Therefore, the Arellano-Bond test for zero autocorrelation in the
first differenced errors is implemented to check if the moment conditions used by GMM
estimators are valid.
The paper allows parameters to be sector variant in order to capture potential
cross-sector heterogeneity. However, after a careful model selection process (selected
results are listed in Table 5-2 below due to limited space) comparing performances from
different models, results show that the models with sector specific parameters do not
outperform the ones with homogeneous parameters across industries according to two
criteria. In general, models with cross-sector heterogeneity exhibit worse AIC and even
worse BIC (as BIC penalizes free regressors more than does AIC) than models with
homogeneous parameters. Secondly, these models also have a higher tendency not to
reject the Arellano-Bond test at order 1, meaning that the moments tend to be invalid for
these models.
Model 6 with homogeneous parameters across industries is chosen due to valid
moment conditions, decent model performance (low levels of AIC and BIC), and
significance of parameters. Signs of estimates are consistent with the discussion above.

56
Variable 1 2 3 4 5 6
66
7
1
ln 2
it
co
−

0.61
0.54 0.75 0.7 0.79 0.64
67
0.6

(0.106)*
**
(0.101)*
**
(0.021)*
**
(0.023)*
**
(0.020)*
**
(0.044)**
*
(0.046)*
**
ln
it
y 0.29 -0.21 0.74 0.32 0.79 0.83 0.36

(0.036)*
**
(0.119 )
**
(.033)**
*
(0.169)*
*
(0.0871)
***
(0.191)**
*
(0.169)*
*
1
ln
it
y
−

-0.68 -0.64 -0.63 -0.46 -0.43

(0.065)*
**
(0.069)*
**
(0.095)*
**
(0.090)**
*
(0.110)*
**
ln
it
e

0.48 0.41
(0.279)** (0.22)
2
ln
it
y

0.16

0.13

0.18

(.038 )*
**
(0.050)*
**
(0.05)
ln
t
P

0.29 0.19 0.43

(0.147)*
**
(0.152)** -0.174
ln( ) year

-83.45 -52.66
-141.75
08

(54.08)
(59.125 )
*
(63.18)
_ con
-1.24 -1.037 -0.63 -0.49 631.9 397.54
1073.38
1

(0.266)*
**
(0.343)*
**
(0.105)*
**
(0.120)*
**
(409.97)
(448.001
)**
(478.92
8)*
Without cross-sec hetero
Valid moment
conditions
68

no no yes** yes*** yes** yes** yes**
RSS 42.54 39.26 27.61 26.23 28.34 25.45 24.60
AIC -624.70 -647.97 -758.89 -772.96 -746.62 -778.56 -787.24
BIC -613.44 -632.95 -743.88 -754.20 -724.11 -752.29 -757.22

66
All models listed in the table exhibit joint significance.
67
*p<.10, **p<.05, ***p<0.01
68
Arellano-Bond test for zero autocorrelation in first-differenced errors. (H0: no autocorrelation)

57
# of Obs. 315 315 315 315 315 315 315
With cross-sec hetero
Valid moment
conditions
yes* yes* no no no no no
RSS 23.98 24.80 22.41 21.90 22.48 17.44 17.34
AICc
69
-711.43 -639.66 -671.62 -606.49 -726.25 -744.20 -672.53
BIC -563.93 -432.51 -464.47 -350.81 -572.66 -531.90 -412.93
# of Obs. 315 315 315 315 315 315 315
Table 5-2: Selected results from Model Selection
70

In order to obtain the forecasts of emissions that would have occurred since 2001
without agglomeration policy, we also need to know what would have been the paths for
the independent variables since 2001. Therefore, similar to the method in 5.2, this paper
fits time-series models for energy consumption, output, and employment by sector
respectively using the data from 1985 to 2000, and then makes forecasts for these
independent variables since 2001 based on its own ARIMA model, which are then used as
inputs into Model 6.
Table 5-3, 5-4 and 5-5 shows the estimates for output, energy consumption, and
employment, respectively (The autocorrelation and partial autocorrelation graphs and the
graphs of forecasts for the three variables are shown in the Appendix). Plugging the
forecasts into Model 6, we can obtain the forecasts of emissions by sector shown in
Figure 5-5, and the forecasts of total emissions by aggregating emissions from all sectors
(Figure 5-6).

69
Small sample AICc=AIC+2(k+1)(k+2)/(n-k) when n/k<40, where n is the sample size and k is number of regressors.
70
Estimates are not listed for models with cross-section heterogeneity due to limited space.

58
ARIMA(p,d,q) Variables Coefficient Sigma ARIMA(p,d,q) Variables Coefficient Sigma
Sector 1 (0, 1, 0) Constant 189.45 143.74 Sector 13 (0, 1, 1) MA(1) 0.42 33.8
(38.526)*** (42.618)*** (0.153)*** (7.939)***
Sector 2 (1, 1, 0) AR(1) -0.36 193.39 Constant 34.24
(0.168)** (38.681)*** (15.179)***
Constant 54.09 Sector 14 (1, 1, 0) AR(1) -0.28 16.039
(40.539)* (0.262)* (4.163)***
Sector 3 (0, 1, 0) Constant 59 63.405 Constant 13.33
(18.118)*** (12.514)*** (3.752)***
Sector 4 (0, 1, 1) MA(1) 0.43 30.11 Sector 15 (4, 1, 0) AR(4) 0.619 35.16
(0.326)* (5.913)*** (0.350)** (13.191)***
Constant 39.09 Constant 51.75
(13.265)*** (18.163)***
Sector 5 (0, 1, 1) MA(1) -0.39 28.78 Sector 16 (0, 1, 2) MA(2) -0.98 110.14
(0.232)** (7.295)*** (0.375)*** --
Constant 16.61 Constant 92.89
(4.819)*** (10.975)***
Sector 6 (0, 1, 0) Constant 7.65 11.81 Sector 17 (0, 1,1) MA(1) 0.45 307.05
(3.133)*** (03.442)*** (0.180)*** (42.789)***
Sector 7 (0, 1, 0) Constant 40.01 34.65 Constant 119.31
(10.980)*** (9.189)*** (132.27)
Sector 8 (0, 1, 0) Constant 12.06 18.31 Sector 18 (1, 1, 0) AR(1) 0.66 39.76
(4.749)*** (2.941)*** (0.283)*** (10.697)***
Sector 9 (1, 1, 0) AR(1) 0.34 11.37 Constant 65.27
(0.226)* (2.925)*** (36.403)*
Constant 17.12 Sector 19 (0, 1, 1) MA(1) 0.98 72.925
(4.641)*** (0.482)** --
Sector 10 (0, 1, 1) MA(1) 1.09 159.44 Constant 63.32
(0.204)* (35.116)*** (4.548)***
Constant 143.697 Sector 20 (1, 1, 0) AR(1) 0.49 542.91
(112.906)* (0.256)* (141.572)***
Sector 11 (0, 1, 0) Constant 143.21 99.32 Constant 655.83
(28.918)*** (24.999)*** (272.954)**
Sector 12 (0, 1, 0) Constant 47.17 41.93 Sector 21 (0, 1, 2) MA(2) -0.98 31.58
(10.944)*** (9.784)*** (0.370)*** --
Constant 31.49
(5.376)***
Table 5-3: Time series estimates of Output
it
Y
by sector

59

ARIMA(p,d,q) Variables Coefficient Sigma ARIMA(p,d,q) Variables Coefficient Sigma
Sector 1 (0, 1, 1) MA(1) 0.45 252.267 Sector 12 (0, 1, 1) MA(1) 0.79 79.986
(0.173)*** (66.949)*** (0.214)*** (13.464)***
Constant 49.01 Constant 21.55
(136.376)* (56.063)*
Sector 2 (1, 1, 0) AR(1) -0.45 183.88 Sector 13 (1, 1, 0) AR(1) -0.53 103.34
(0.241)* (34.241)*** (0.146)*** (23.804)***
Sector 3 (4, 1, 0) AR(4) -0.29 30.58 Constant 76.58
(0.370)* (6.569)*** (25.213)***
Constant 15.03 Sector 14 (1, 1, 0) AR(1) -0.28 41.86
(8.193)* (0.322)* (9.896)***
Sector 4 (0, 1, 0) Constant 37.15 37.15 Constant 17.66
(4.787)*** (4.787)*** (9.677)*
Sector 5 (2, 1, 0) AR(2) 0.39 24.94 Sector 15 (3, 1, 0) AR(3) -0.27 43.58
(0.334)* (3.638)*** (0.293)* (8.418)***
Sector 6 (1, 1, 0) AR(1) -0.07 26.61 Constant 28.3
(1.439)* (2.909)*** (9.194)***
Sector 7 (1, 1, 0) AR(1) 0.17 106.77 Sector 16 (1, 1, 0) AR(1) 0.47 628.89
(0.474)* (24.901)*** (0.461)* (108.344)***
Constant 36.93 Constant 114.56
(60.960)* (367.205)*
Sector 8 (1, 1, 0) AR(1) 0.598 11.23 Sector 17 (1, 1, 0) AR(1) -0.37 3692.85
(0.233)** (2.460)*** (0.155)** (509.022)***
Constant 8.9 Constant 641.04
(6.497)** (761.599)*
Sector 9 (0, 1, 1) MA(1) -0.98 13.39 Sector 18 (0, 1, 0) Constant 162.81 60.05
(0.324)*** -- (17.138)*** (14.133)***
Constant 5.29 Sector 19 (0, 1, 0) Constant 24.06 37.12
(0.852)*** (9.64)** (5.620)***
Sector 10 (1, 1, 0) AR(1) -0.56 952.99 Sector 20 (1, 1, 0) AR(1) 0.53 165
(0.185)*** (305.973)*** (0.197)*** (51.178)***
Constant 387.36 Constant 46.44
(240.609)* (97.820)*
Sector 11 (1, 1, 0) AR(1) 0.14 3616.67 Sector 21 (1, 1, 0) AR(1) -0.28 228.42
(0.695)* (772.080)*** (0.183)* (52.374)***
Constant 303.01 Constant 18.82
(2314.69) (48.621)*
Table 5-4: Time series estimates of Energy consumption
it
E
by sector

60
ARIMA(p,d,q) Variables Coefficient Sigma ARIMA(p,d,q) Variables Coefficient Sigma
Sector 1 (1, 1, 0) AR(1) -0.38 52.57 Sector 12 (1, 1, 0) AR(1) 0.34 7.23
(0.185)** (9.523)*** (0.229)* (1.336)***
Sector 2 (1, 0, 0) AR(1) 0.47 168.496 Constant 2.72
(0.260)** (50.188)*** (3.464)*
Constant 550.42 Sector 13 (1, 1, 0) AR(1) 0.28 4.57
(113.271)*** (0.165)** (0.772)***
Sector 3 (1, 0, 0) AR(1) 0.84 29.65 Sector 14 (2, 1, 0) AR(2) 0.47 9.72
(0.195)*** (5.746)*** (0.263)** (2.035)***
Constant 113.65 Constant 0.3
(46.749)** (4.337)*
Sector 4 (1, 0, 0) AR(1) 0.81 13.64 Sector 15 (1, 0, 0) AR(1) 0.8 17.81
(0.178)*** (3.198)*** (0.189)*** (4.254)***
Constant 61.41 Constant 70.08
(15.898)*** (20.290)***
Sector 5 (1, 1, 0) AR(1) -0.39 14.48 Sector 16 (1, 1, 0) AR(1) -0.42 53.584
(0.203)** (3.175)*** (0.176)** (11.250)***
Constant 0.11 Sector 17 (0, 1,2) MA(2) 0.42 23.69
(2.907)** (0.333)* (4.524)***
Sector 6 (0, 0, 0) Constant 32.19 11.17 Constant -1.32
(3.843)*** (2.959)*** (10.164)*
Sector 7 (1, 0, 0) AR(1) 0.78 17.66 Sector 18 (1, 0, 0) AR(1) 0.91 7.1
(0.33)** (7.806)** (0.114)*** (1.451)***
Constant 92.71 Constant 74.62
(18.336)*** (10.923)***
Sector 8 (2, 1, 0) AR(2) 0.43 12.99 Sector 19 (1, 0, 0) AR(1) 0.67 35.62
(0.256)* (2.714)*** (0.252)*** (11.621)***
Sector 9 (1, 0, 0) AR(1) 0.89 4.05 Constant 140.56
(0.119)*** (0.729)*** (31.491)***
Constant 23.75 Sector 20 (1, 1, 0) AR(1) -0.43 194.04
(6.566)*** (0.185)** (41.394)***
Sector 10 (1, 1, 0) AR(1) 0.17 7.54 Sector 21 (6, 0, 0) AR(6) -0.6 52.65
(0.410)* (1.285)*** (0.368)* (15.506)***
Constant 1.2 Constant 158.7
(3.073)*

(12.519)***
Sector 11 (3, 1, 0) AR(3) 0.29 24.15

(0.973)* (7.36)***

Constant 1.13

(15.438)*
Table 5-5: Time series estimates of Employment
it
N
by sector

61

20 40 60 80 100
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL
20 30 40 50 60
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL

2 4 6 8
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL
0 1 2 3 4 5
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL

62
2 4 6 8 10
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL
0 .5 1 1.5
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL

20 40 60 80 100
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL
.5 1 1.5 2
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL

63
0 .5 1 1.5 2
1985 1990 1995 2000 2005 2010
year
0 500 1000 1500 2000
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL

200 300 400 500 600
1985 1990 1995 2000 2005 2010
year
5 10 15 20
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL

64
10 20 30 40
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL
4 6 8 10 12
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL

2 4 6 8 10 12
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL
100 200 300 400 500 600
1985 1990 1995 2000 2005 2010
year

65
0 500 1000 1500
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL
0 50 100 150
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL

6 8 10 12 14 16
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL
20 40 60 80 100 120
1985 1990 1995 2000 2005 2010
year
ACTURAL ECONOMIC MODEL

66
10 20 30 40
1985 1990 1995 2000 2005 2010
year
ACTURAL

Figure 5-5: ECONOMIC MODEL_ forecast by sector_ CO
2
emissions

Total CO2 emissions (million metric ton)
1000 2000 3000 4000 5000
1985 1990 1995 2000 2005 2010
year
ACTURAL DATA Economic model

Figure 5-6: Economics model_ forecast in comparison with the actual data__ CO2
emissions

67
5.4 Population averaged Model
As pointed out in 5.1, the cross-industry coefficients may contain information about
average differences between industries (or the population-averaged effect). In order to
solve this problem, we fit a population-averaged model, a generalized linear model where
(ln 2 ) ,
it it
E co x β = ,
and
2
ln 2 ~ (), (ln ,(ln ) ,ln ,ln ,ln( ))
it it it it it t
co N x y y e P year =
The population averaged models can be estimated using general estimation
equations (GEE). The Akaike information criterion (AIC) is not applicable because no
distribution or likelihood was assumed or defined for GEE. Therefore, quasi-likelihood
under the independence model information criterion (QIC
71
) is applied as an extension of
the widely used AIC
72
. Similar to the selection of the economic model in section 5.3, a
model selection process is applied to tackle regressor issues. Selected results are shown in
Table 5-6 and 5-7. The estimates and corresponding QIC indicate that models with lagged
dependent variables (Table 5-6) generally perform better than other models (Table 5-7),
though Liang and Zeger (1986)
73
pointed out that “the GEE makes assumptions about
the interrelationship of yt and yt −1, and then uses those assumptions to perform

71
It allows for seven distributions (gaussian, inverse Gaussian, Bernoulli/binomial, Poisson, negative binomial and
gamma) and all link functions and working correlation structures (independent, exchangeable, autoregressive,
stationary, non-stationary and unstructured).
72
Pan, 2001:
1
ˆˆ
ˆ 2( ; ) 2 ( )
I R
QIC Q I trace V μ
−
=− + Ω , where
ˆ
I
Ω is the variance estimator under independence
correlation structure,
ˆ
R
V is the robust variance estimator under specified working correlation structure, and
ˆ (; ) QI μ is the quasi-likelihood function under specified distribution.
73
Liang and Zeger 1986, Longitudinal Data Analysis Using Generalized Linear Models, Biometrika, V ol. 73, No. 1.
pp.13-22, 1986.

68
“quasi-maximum likelihood.” While this is a well known and often used method, it is a
bit of a black box. While some political scientists, such as Zorn (2001), have found the
GEE to be useful, it clearly is not an attempt to model the dynamics. While the QIC is
calculated for selection of correlation structures in Table 5-7 (models without lagged
dependent variables as covariates), it is not applied to Table 5-6 (models with lagged
dependent variables as covariates), where the correlation structure is assumed to be
independent.
74
Semi-robust standard errors are applied
75
to all models.

74
Explained in Chapter 12 of Analysis of Longitudinal Data (2
nd
edition, Oxford Univ. Press, 2002) by Diggle,
Heagerty, Liang and Zeger: one should use an independence working correlation when lagged dependent variables are
used as covariates in a model fit by GEE, unless one can fulfill the "full covariate conditional mean assumption,"
described in the book.
75
Huber (1967) presented robust standard errors by treating each country as a cluster.

69
Variable 1 2 3 4 5 6 7
1
ln 2
it
co
−
0.95 0.95 0.98 0.98 0.98 0.81
76
0.79
(0.019)*** (0.023)*** (0.013)*** (0.017)*** (0.015)*** (0.054)***
77
(0.069)***
ln
it
y 0.16 0.08 0.81 0.74 0.79 0.86 0.64
(0.042)*** (0.102) (0.039)*** (0.117)** (0.049)*** (0.055*** (0.139)***
1
ln
it
y
−

-0.8 -0.81 -0.74 -0.59 -0.58
(0.059)*** (0.060)*** (0.068)*** (0.091)*** (0.100)***
ln
it
e

0.23 0.24
(0.070)** (0.079)***
2
ln
it
y

0.03

0.02

0.08
(0.042) (0.036) (0.044)*
ln
t
P

0.39 0.38 0.43
(0.081)*** (0.068)*** (0.096)***
ln( ) year
-98.67 -101.07 0.24
(19.652)*** (22.493 )*** (0.079)***
_ con
-0.36 -0.31 -0.08 -0.04 747.59 765.23 1073.381
(0.105)*** (.068 )*** (0.082) (0.029) (148.832)*** (170.334)*** (478.928)*
QIC
78
69.09 76.15 69.51 70.05 66.68 63.77 64.46
# of
Industries
21 21 21 21 21 21 21
# of Obs. 315 315 315 315 315 315 315
Table 5-6: Selected results for population-averaged model (models with lagged dependent
variables as covariates)
79

76
Semi-robust standard error is applied (adjusted for clustering on sector).
77
**p<.10, **p<.05, ***p<0.01
78
An independence working correlation is assumed, the reason was explained in Chapter 12 of Analysis of
Longitudinal Data (2
nd
edition, Oxford Univ. Press, 2002) by Diggle, Heagerty, Liang and Zeger, that one should use an
independence working correlation when lagged dependent variables are used as covariates in a model fit by GEE,
unless one can fulfill the "full covariate conditional mean assumption," described in the book.
79
All models listed in Table 4 exhibit joint significance.

70
Variable 1 2 3 4 5 6 7
ln
it
y 0.46 -0.15 0.32 0.48 1.75 1.28 0.73
(0.084)*** (0.167) (0.053)*** (0.163)*** (0.472)*** (0.141)*** (0.309)**
1
ln
it
y
−

0.09 0.01 -0.19 0.63 0.059
(0.109) (0.068) (0.205) (0.083) (0.079)
ln
it
e

1.17 1.15
(0.049)*** (0.043)***
2
ln
it
y

0.19

-0.02

0.19
(0.060)*** (0.047) (0.103)*
ln
t
P

0.61 0.21 0.41
(0.373)* (0.126)* (0.178)**
ln( ) year
-369.68 -143.79 -206.19
(169.189)** (59.852)** (75.567)**
_ con
-2.67 -2.25 -2.63 -2.63 2801.17 1086.693 1560.198
(0.279)*** (0.328)*** (0.254)*** (0.290)*** (1282.677)** (453.886)** (573.147)**
QIC 1375.77 1416.79 1347.10 1356.45 877.53 173.25 186.01
# of Industries 21 21 21 21 21 21 21
# of Obs. 315 315 315 315 315 315 315
QIC for selection of correlation structure
Independent 1536.27 1723.92 1524.49 1810.00 877.53 173.25 186.01
Exchangeable 1375.77 1416.79 1347.10 1387.73 1611.92 788.97 736.59
Autoregressive 1429.25 1431.89 1353.07 1356.45 1246.51 892.41 908.52
Stationary -- -- -- -- -- -- --
Non-stationary -- -- -- -- -- -- --
Unstructured -- -- -- -- -- -- --
Table 5-7: Selected results for population-averaged model (models without lagged
dependent variables as covariates)
80

Model 6 in Table 5-6 is chosen due to its significance of estimates and lowest QIC.

80
All models listed in Table 5-7 exhibit joint significance.

71
The signs of estimates are the same as in the economic model, but lagged emissions,
lagged output, purchasing price of fuel and raw materials, and the time trend have
stronger explanatory power than in the economic model.
As with the economic model, forecasts are conducted based on the
population-averaged model (Model 6) and the time-series forecasts of independent
variables by sector (the estimates of which are given in table 5-3, 5-4 and 5-5 in section
5.3). The forecasts are shown below in Figure 5-7:

50 60 70 80 90
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED
20 30 40 50 60
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED

72
2 4 6 8 10
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED
2 3 4 5
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED

4 6 8 10
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED
.8 1 1.2 1.4 1.6 1.8
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED

73
20 40 60 80 100
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED
1 2 3 4 5
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED

0 1 2 3
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED
0 500 1000 1500 2000
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED

74
100 200 300 400 500 600
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED
5 10 15 20
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED

10 20 30 40
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED
6 8 10 12
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED

75
0 5 10 15 20
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED
100 200 300 400 500 600
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED

0 500 1000 1500
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED
0 50 100 150
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED

76
10 12 14 16 18 20
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED
40 60 80 100 120
1985 1990 1995 2000 2005 2010
year

10 20 30 40
1985 1990 1995 2000 2005 2010
year
ACTURAL POPULATION-AVERAGED

Figure 5-7: POPULATION-A VERAGED model forecast by sector_ CO
2
emissions
The aggregate emissions from 21 sectors are given in Figure 5-8:

77
Total CO2 emissions (million metric ton)
1000 2000 3000 4000 5000
1985 1990 1995 2000 2005 2010
year
ACTURAL DATA Population averaged Model

Figure 5-8: Population averaged model forecast in comparison with the actual data__
Total CO
2
emission

5.5 GV AR model
As one of the disadvantages of all previous models is that no cross-industry
dependence is modeled, we adopt the global vector autoregression approach (GVAR)
advanced in Pesaran, Schuermann and Weiner (2004, PSW) to capture the possible
interdependence. Specifically, for industry i, the paper considers the VARX*(2,1)
81

structure given by

81
The actual lags for domestic and foreign variables are selected through Akaike criterion (AIC).

78
**
0 1 1, 1 2 , 2 0 , 1, 1 0 1 1 0 2 it i i i i t i it i i t i it i t i t i t it
x aat x x x x d d d u
−− − − −
= + + Φ + Φ +Λ +Λ +Ψ + Ψ + Ψ + (1)

Where
it
x is the vector of endogenous variables for industry i in year t,
with
,
(2 , , )'
it it it it it
x CO Y E N =
82
,
*
it
x contains the “foreign variables” (weighted averages of
variables from the rest industries) for industry i in year t,
with
***
0
(2, ) ( 2, )
N
it it it ij jt jt
j
x CO Y CO Y ω
=
==
∑
,
ij
ω
83
is the predetermined fix weighting matrix
constructed using cross-industry flows data
84
and generated by GVAR program;
t
d are
the “global variables”
85
with (,ln( ))
tt
dP year = , and
it
u is a serially uncorrelated and
cross-sectionally weakly dependent process.
In order to conquer the “curse of dimensionality” issue, GVAR imposes weak
exogeneity of the foreign variables and global variables. That is, it assumes that each
industry (with the exception of the sector Processing of Petroleum, Coking and Nuclear
Fuel) is a small sector with respect to the rest of the manufacturing world. I test this
empirically later.

82
2
it
CO ,
it
Y ,
it
E , and
it
N represent emissions, output, energy consumption and employment for industry i in year t.
83
See Appendix.
84
The flow data (impact of industry i on industry j) is constructed as the ratio of the total output of industry i and the
total output of industry j.
85
Variables that do not vary across industries

79
Variables Endogenous
86
Foreign
87

Globle
88

CO
2
emissions
,
2
it
CO
*
2
it
CO -
Energy use
it
E
- -
Total output
it
Y
*
it
Y -
Employment
it
N
- -
Purchasing price - -

t
P
Technology change - - ln( ) year
Table 5-8: Summary of variables

Estimation is performed on the error-correction form VECMX* of equation (1) to
take into account the integration properties:

'*
0,1 0 ,1
[(1)]
it i i i i t i i it i i t it
xcz t x z u αβ γ
−−
Δ= − − − +Λ Δ +ΓΔ + , where
*
(', ', ')'
it it it t
zx x d = (2)
.
The rank order of cointegration for each model is selected by the tests using the trace
statistic at the 5% level of significance:

86
Industry-specific variables
87
Weighted averages of variables from the rest industries
88
Variables that do not vary across industries

80
Industry ARTWORK CHEMICAL CULTURE FERROUS FIBERS FOOD FURNITURE
# Cointegrating
relations 2 0 2 2 0 2 2
Industry LEATHER MACHINERY MEDICINES METAL NONFERROUS NONMETALIC PAPER
# Cointegrating
relations 2 2 2 2 0 0 3
Industry PETROLEUM PLASTICS PRINTING RUBBER TEXTILE TIMBER WEARING
# Cointegrating
relations 2 2 2 2 3 2 2
Table 5-9: Number of Cointegrating Relationships for the Individual V ARX* Models

The estimates are given in Table 5-10:

81
ARTWORK Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo 11124081.90 365.41 0.23 0.03 -0.02 -0.11 -1.23 -0.01 42.98 -1465034.02 -1.58 -0.02 38.73 -736746.88 -0.42 -0.05 0.01 0.06
dgdp 111269975.25 3660.86 -0.43 0.07 0.00 -0.97 -7.92 -0.19 356.57 -14654170.33 -6.47 0.47 216.40 -7323180.82 -2.69 -0.28 0.00 0.56
de 179530529.83 5845.80 28.02 2.84 -1.71 -2.85 -58.83 0.51 1347.61 -23644135.93 -63.57 0.63 1120.97 -11946243.32 -20.75 -2.93 0.70 1.72
dem 111610597.98 3674.26 -1.46 -0.03 0.07 -0.93 -6.29 -0.22 329.93 -14699026.39 -3.36 -0.32 346.63 -7387215.13 -2.11 -0.19 -0.02 0.53
CHEMICAL Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo -0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 5.85 -0.11 17.05 3538.27 0.00 0.00 0.00 0.00
dgdp 0.04 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.87 1.86 -79.23 182320.09 0.00 0.00 0.00 0.00
de 3.16 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 220.63 56.02 9427.84 -6229660.44 0.01 -0.02 0.00 0.00
dem -0.05 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.52 -0.10 139.47 -35394.58 0.00 0.00 0.00 0.00
CULTURE Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo 610798.96 20.23 1.06 0.00 -0.02 0.02 -0.01 0.00 -0.15 -80441.03 -0.02 0.00 0.55 -40327.82 -1.33 0.00 0.02 0.00
dgdp -6249204.45 -202.68 38.20 -0.20 -0.97 -0.66 0.18 0.09 -26.92 823021.70 1.12 0.19 -59.58 437906.83 -41.44 -0.15 0.72 0.13
de 21612947.24 720.59 91.62 -0.42 -2.20 0.14 -0.13 0.20 -36.95 -2846367.72 -1.04 0.15 -12.73 -1426300.16 -107.93 -0.15 1.70 0.33
dem -15572813.06 -513.91 -5.59 0.00 0.09 -0.68 0.23 -0.01 -8.49 2050914.99 0.07 -0.04 7.40 1023512.65 9.91 -0.07 -0.09 -0.03
FERROUS Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo 15573204.54 516.24 -0.31 -0.02 0.00 0.14 5.34 -0.02 20.38 -2050973.71 15.75 -0.18 26.33 -1011001.66 0.05 0.01 0.00 -0.03
dgdp -1322160299.81 -43701.81 -3.28 -1.43 -0.08 -4.39 77.40 5.44 470.15 174125857.33 77.57 5.17 467.64 87118536.21 3.99 0.54 0.08 -0.78
de 8053707246.77 266966.37 -158.98 -11.65 -1.28 71.75 2726.61 -10.84 10385.79 -1060664201.57 1273.98 -44.13 10409.99 -528948408.00 23.83 6.02 0.93 -13.89
dem -56880580.05 -1881.64 0.22 -0.02 0.00 -0.28 -3.13 0.19 -6.53 7491072.59 1.40 0.08 100.37 3716643.24 0.07 0.00 0.00 0.00
FIBERS Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.68 0.03 -4.78 -957.97 0.00 0.00 0.00 0.00
dgdp 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 2.00 0.22 -77.99 68596.96 0.00 0.00 0.00 0.00
de 0.09 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.35 -1.69 -171.92 317791.71 0.00 0.00 0.00 0.00
dem 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.12 0.01 16.88 -4447.66 0.00 0.00 0.00 0.00
Table 5-10: VECMX* Estimates of the Individual Models

82
FOOD Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo -17022616.40 -562.74 -0.51 0.00 0.01 -0.07 0.49 -0.01 0.45 2241848.65 0.54 -0.08 8.43 1122014.55 -0.28 0.00 0.00 0.03
dgdp -623399772.65 -20605.64 0.58 -0.06 -0.01 -2.30 9.98 0.07 -35.57 82100667.15 7.97 3.12 -284.68 41246018.75 -12.93 -0.03 -0.02 0.94
de -1225413364.59 -40496.42 51.52 -0.09 -0.98 -4.20 -1.56 1.30 -206.51 161384845.63 7.39 -3.50 230.45 80910509.61 -32.43 0.01 0.37 0.97
dem -278246906.91 -9196.66 2.92 -0.03 -0.05 -1.01 3.34 0.09 -23.09 36644636.11 -0.35 0.36 82.83 18255099.52 -6.14 -0.01 0.01 0.37
FURNITURE Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo -316133.67 -10.45 -0.13 0.00 0.00 -0.01 0.00 0.00 0.02 41634.25 0.01 0.00 -0.07 20763.26 -0.11 0.00 0.00 0.00
dgdp -897022.10 -29.50 4.56 0.06 -0.05 -0.19 -0.06 -0.04 7.23 118133.94 0.60 0.31 -6.14 51041.46 -6.92 0.11 0.00 0.10
de 73756539.02 2441.52 140.11 -0.10 -0.68 -1.61 -1.85 -0.64 153.42 -9713657.04 -0.08 0.08 111.80 -4866275.96 -120.87 2.23 -0.29 1.44
dem -43589313.07 -1441.10 -25.51 0.57 -0.12 -0.99 0.30 -0.10 -7.49 5740639.20 0.07 0.02 9.84 2863795.95 -5.28 -0.11 0.10 0.21
LEATHER Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo 1649381.21 54.24 -2.69 0.02 0.05 0.05 -0.27 0.00 -2.54 -217218.53 -0.29 -0.01 -2.20 -109226.95 1.47 -0.03 -0.04 -0.04
dgdp 30485834.74 1008.72 -8.46 -0.08 0.22 0.91 0.35 0.05 -27.02 -4014920.13 3.43 0.78 34.65 -2014220.53 -6.73 -0.17 0.16 -0.31
de 53474324.64 1756.61 -100.23 0.76 1.96 1.60 -10.27 -0.03 -88.50 -7042398.54 -11.21 0.09 -41.30 -3567165.15 58.31 -0.99 -1.67 -1.42
dem -29735532.89 -983.10 13.52 0.02 -0.31 -0.89 0.33 -0.04 28.89 3916103.43 1.45 -0.09 62.38 1964901.38 2.23 0.21 -0.03 0.36
MACHINERY Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo -21298210.83 -702.82 -0.72 0.00 -0.01 0.02 1.25 -0.06 -16.84 2804948.25 1.35 0.12 -35.09 1401050.31 0.56 0.01 -0.02 -0.05
dgdp 320003154.33 10556.50 10.92 0.01 0.13 -0.63 -19.93 0.62 412.07 -42144046.40 -29.96 19.18 28.36 -21016368.07 -7.83 -0.11 0.36 0.86
de -297465725.68 -9802.69 -10.31 0.04 -0.22 1.58 22.06 0.35 -885.20 39175966.97 23.01 1.19 -481.83 19624194.44 5.64 0.06 -0.47 -1.03
dem -121939258.25 -4031.79 -4.02 -0.05 0.04 -0.64 4.45 -1.06 288.76 16059194.06 4.99 -0.76 458.34 7995972.13 4.44 0.08 -0.01 -0.12
MEDICINES Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo -4722321.58 -156.12 -1.65 0.02 0.02 0.01 0.09 -0.01 -3.33 621922.18 0.25 -0.03 1.43 311066.91 0.62 -0.02 -0.01 0.01
dgdp 33857348.49 1121.13 22.07 -0.10 -0.38 0.90 0.05 0.06 -3.18 -4458953.35 -1.64 0.72 -97.86 -2148742.94 -11.39 0.05 0.21 -0.01
de -104300499.13 -3444.17 -13.22 0.49 0.05 2.51 3.68 -0.41 -134.66 13736220.02 16.86 -1.50 140.18 6885246.29 -2.08 -0.59 0.06 0.21
dem 6417222.77 210.86 -5.20 -0.05 0.12 -0.74 -0.65 0.05 24.15 -845141.67 -0.18 0.06 26.06 -421625.22 4.21 0.08 -0.08 -0.03
Table 5-10: VECMX* Estimates of the Individual Models_ continued

83
METAL Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo 2286313.08 75.83 0.37 0.00 -0.02 0.01 0.09 0.00 2.00 -301102.24 0.14 -0.02 3.01 -149935.06 -0.42 0.01 0.01 0.00
dgdp -33971764.03 -1126.81 -5.47 0.06 0.37 -0.08 -1.39 0.07 -29.55 4474004.18 -1.38 2.59 89.77 2158799.53 6.16 -0.10 -0.16 0.02
de 121635069.80 4035.40 21.25 -0.28 -1.48 0.17 5.81 -0.25 128.48 -16019059.55 5.10 0.20 132.02 -7982916.08 -25.27 0.40 0.66 -0.04
dem -143204873.86 -4741.94 -8.23 -0.15 0.18 -1.39 1.62 0.35 76.74 18859782.77 0.57 0.12 96.42 9412986.13 -2.67 0.02 -0.03 0.31
NONFERROUS Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo -0.05 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.51 0.02 2.76 -106.39 0.00 0.00 0.00 0.00
dgdp 0.42 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -2.34 0.86 -40.76 70660.96 -0.01 0.00 0.00 0.00
de -0.26 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.09 -0.83 -148.77 421668.15 0.01 0.00 0.00 0.00
dem -0.06 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.29 0.02 34.20 -7969.89 0.00 0.00 0.00 0.00
NONMETALIC Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.76 -0.24 36.81 7784.88 0.00 0.00 0.00 0.00
dgdp -0.04 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -0.98 5.03 544.12 -313036.09 0.00 0.00 0.00 0.00
de -0.49 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 79.85 -5.19 1991.82 -99067.43 0.02 0.00 0.00 0.00
dem -0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.29 -0.13 193.69 -49319.86 0.00 0.00 0.00 0.00
PAPER Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo 10106082.44 334.37 -0.32 -0.02 0.00 0.13 0.03 -0.01 3.03 -1330952.33 0.50 0.00 10.00 -667142.69 0.13 0.02 0.00 -0.05
dgdp -281184150.99 -9296.04 29.78 -0.14 -0.63 -0.37 -6.90 1.35 -145.62 37031477.37 -4.08 1.54 -36.67 18498047.89 -21.72 -0.39 0.50 0.47
de 50766579.32 1725.81 65.91 -1.31 -1.57 6.10 -30.99 0.56 -213.14 -6685686.91 -4.99 0.99 310.14 -3388108.12 -16.92 1.78 0.41 -2.40
dem -82359797.77 -2722.06 5.26 0.18 -0.01 -1.05 -2.02 -0.06 -35.14 10846641.17 0.31 -0.02 31.52 5410660.14 0.06 -0.03 -0.05 0.36
Table 5-10: VECMX* Estimates of the Individual Models_ continued

84
PETROLEUM Intercept Trend cotwo_1 gdp_1 e_1 em_1 rpp_1 tech_1 cotwos_1 gdps_1 dcotwos_0 dgdps_0 dcotwo_1 dgdp_1 de_1 dem_1 drpp_1 dtech_1
dcotwo 25288.25 2.98 -0.10 -0.05 0.02 -0.03 13.22 -3358.56 3.41 0.09 60.12 0.54 -0.05 0.01 -0.01 2.65 254.93 -40574.03
dgdp -174911.93 25.87 0.69 0.31 -0.11 0.22 -87.62 23105.92 -22.85 -0.63 -12.25 4.56 -0.34 -0.51 0.06 0.42 207.61 51082.60
de -1986647.81 275.02 7.86 3.56 -1.29 2.53 -997.12 262600.39 -259.93 -7.16 101.63 -14.38 -3.52 -5.68 0.66 -2.81 1564.57 -1032067.12
dem -133.79 -0.36 0.00 0.00 0.00 0.00 -0.10 18.86 -0.02 0.00 -0.28 -0.04 0.01 0.00 0.00 -0.15 -16.13 -454.31
drpp -81.97 0.00 0.00 0.00 0.00 0.00 -0.04 10.88 -0.01 0.00 0.02 0.00 0.00 0.00 0.00 -0.01 -0.49 -65.07
dtech 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -1.00
PLASTICS Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo -3526040.36 -116.32 -0.21 0.00 -0.01 0.00 0.11 0.00 -2.19 464373.79 0.10 -0.01 -0.96 232533.59 -0.63 0.00 0.00 0.00
dgdp -52518932.06 -1734.60 -2.03 0.00 -0.03 -1.47 0.41 -0.23 54.84 6916633.78 0.01 0.77 -119.96 3544420.09 -9.29 0.04 0.11 0.45
de -289584081.35 -9551.90 -17.67 0.10 -0.61 0.69 9.53 -0.11 -235.84 38137768.59 6.11 0.73 -116.97 19049785.81 -51.47 -0.06 0.26 -0.17
dem -42808168.42 -1414.01 -1.58 0.00 -0.02 -1.29 0.26 -0.20 50.67 5637744.81 -0.32 -0.08 65.84 2815713.75 -7.57 0.04 0.09 0.39
PRINTING Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo -1135108.60 -37.48 -1.05 0.01 0.01 -0.02 -0.04 0.00 -0.06 149491.95 -0.02 0.00 0.36 74724.21 -0.22 -0.01 -0.01 0.01
dgdp 32675014.54 1077.80 56.51 -0.22 -0.64 0.66 1.52 0.17 0.85 -4303248.72 -0.90 0.45 -28.56 -2141376.36 -2.75 0.30 0.06 -0.38
de -46707880.45 -1544.59 19.51 0.26 0.08 -1.08 -0.71 -0.04 -4.54 6151345.75 0.14 -0.01 25.75 3067374.55 -30.98 -0.31 -0.47 0.55
dem -21156861.48 -700.56 32.38 0.11 -0.16 -0.52 0.02 0.03 -2.84 2786319.20 -0.77 -0.05 16.92 1386936.05 -22.22 -0.11 -0.30 0.25
RUBBER Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo -3487879.84 -115.11 0.09 0.00 -0.01 -0.03 0.04 0.00 -1.63 459348.41 0.10 -0.02 -0.72 230826.18 -0.21 0.00 0.00 0.00
dgdp -69160683.91 -2277.92 -7.86 -0.90 0.15 -1.15 -0.14 -0.04 -40.86 9108383.89 -1.23 0.17 -89.88 4588826.31 0.19 0.51 -0.13 0.13
de -433685980.80 -14324.50 37.59 2.02 -1.92 -2.96 6.66 -0.02 -179.52 57115699.38 5.10 -0.07 -198.97 28607382.78 -37.96 -1.11 0.91 -0.90
dem -30344066.77 -999.64 -3.01 -0.35 0.05 -0.48 -0.02 -0.02 -17.54 3996278.00 -0.25 0.01 18.05 1990408.23 -0.11 0.20 -0.05 0.05
Table 5-10: VECMX* Estimates of the Individual Models_ continued

85
TEXTILE Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo -51164033.50 -1689.82 -0.47 0.01 -0.02 0.00 0.99 0.00 -25.76 6738221.46 0.70 0.00 -8.10 3367661.63 -1.02 -0.01 0.02 0.00
dgdp -2731441822.96 -90250.20 -29.81 -0.77 -0.70 -0.62 34.63 2.67 -795.52 359726434.48 14.41 4.45 138.03 179648006.73 -21.87 0.04 1.24 0.51
de -2519861967.71 -83214.23 -20.62 0.65 -1.10 0.00 54.08 -0.65 -1427.31 331861816.82 25.47 1.88 -475.16 165808783.84 -60.96 -0.35 1.17 -0.06
dem -811899804.65 -26838.35 17.33 -0.03 -0.54 -1.15 7.78 0.29 -11.95 106925790.92 1.06 0.42 293.98 53385507.89 -28.97 0.06 0.63 c
TIMBER Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo -1812988.76 -59.91 -0.36 -0.01 -0.01 0.04 0.10 0.01 -2.70 238768.07 0.10 0.00 -0.98 119401.27 -0.39 0.00 0.01 -0.02
dgdp -23482016.15 -779.39 12.54 -0.21 -0.20 -0.52 -0.47 0.23 31.95 3092517.70 0.04 0.76 52.04 1519959.90 -10.21 0.04 -0.07 0.33
de -63447994.14 -2102.32 16.16 -0.44 -0.41 -0.36 0.57 0.52 17.68 8355960.03 -0.91 0.10 78.73 4173909.42 -22.22 0.03 0.01 0.24
dem 22760997.63 750.92 10.27 0.03 0.02 -0.83 -1.87 -0.09 55.96 -2997601.60 -0.79 -0.11 82.04 -1507813.47 3.11 0.07 -0.19 0.51
WEARING Intercept Trend cotwo_1 gdp_1 e_1 em_1 cotwos_1 gdps_1 rpp_1 tech_1 dcotwos_0 dgdps_0 drpp_0 dtech_0 dcotwo_1 dgdp_1 de_1 dem_1
dcotwo 116035.09 3.82 -0.22 0.00 0.00 0.00 -0.01 0.00 0.59 -15281.69 0.01 0.00 -0.50 -7099.95 -0.11 0.00 0.00 0.00
dgdp -231963744.12 -7646.10 -15.15 -0.56 -0.81 -1.57 5.61 0.86 -109.76 30549231.97 7.15 1.34 -190.83 15347258.75 54.17 0.02 0.52 0.77
de -79859055.92 -2632.31 33.61 -0.31 -0.58 0.35 3.03 0.84 -128.77 10517312.92 4.11 -0.37 51.11 5304477.69 32.60 -0.06 0.59 0.24
dem -67074652.85 -2210.98 -38.78 -0.05 0.03 -1.24 0.65 -0.24 48.86 8833608.72 -0.45 -0.03 74.66 4398527.00 3.30 0.07 -0.22 0.24
Table 5-10: VECMX* Estimates of the Individual Models_ continued

86
The point forecasts are computed recursively for each sector by

where 2001, 2002...2011 h =
with initial values of
0 ,2000 ii
x μ = ,
, 1 ,1999 ii
x μ
−
= ,
**
0 ,2000 ii
x μ =
**
, 1 ,1999 ii
x μ
−
= ,
00
d κ = and
**
00
d κ =

The forecasts are shown below with respect to each sector (Figure 5-9) and the aggregate
of the 21 sectors (Figure 5-10):

0 200 400 600
1985 1990 1995 2000 2005 2010
year
ACTURAL GVAR
0 50 100 150 200
1985 1990 1995 2000 2005 2010
year
ACTURAL GVAR

87
0 10 20 30 40 50
1985 1990 1995 2000 2005 2010
year
ACTURAL GVAR
0 1 2 3 4 5
1985 1990 1995 2000 2005 2010
year
ACTURAL GVAR

5 10 15 20 25
1985 1990 1995 2000 2005 2010
year
ACTURAL GVAR
.5 1 1.5
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2

88
20 40 60 80 100
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2
0 5 10 15
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2

0 1 2 3 4
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2
0 500 1000 1500 2000
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2

89
200 400 600 800 1000 1200
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2
0 50 100 150 200 250
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2

0 20 40 60
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2
0 50 100 150 200 250
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2

90
0 20 40 60 80 100
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2
0 500 1000 1500 2000 2500
1985 1990 1995 2000 2005 2010

0 500 1000 1500
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2
0 50 100 150
1985 1990 1995 2000 2005 2010
year

91
0 50 100 150 200
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2
0 50 100 150
1985 1990 1995 2000 2005 2010
year

0 10 20 30 40
1985 1990 1995 2000 2005 2010
year
CO2 GVAR CO2

Figure 5-9: GV AR model_ forecast by sector_ CO
2
emissions

92
Total CO2 emissions (million metric ton)
0 2000 4000 6000 8000
1985 1990 1995 2000 2005 2010
year
ACTURAL DATA GVAR

Figure 5-10: GV AR model_ forecast in comparison with the actual data__ CO
2
emissions

5.6 Average of forecasts
As discussed in 5.1, the model that best fits the local data does not necessarily best
fit the whole population, and the simple averaged forecasts turned out to be more robust
procedures in practice (Clemen (1989)) than the optimal combination suggested by Bates
and Granger (1969). Therefore, based on the forecasts of all four models developed above,
we can take the simple average and conduct the program evaluation exercise in the next
section.

93
1000 2000 3000 4000 5000
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE

Figure 5-11: Total CO
2
emissions_ Average of forecasts from 4 models in comparison
with the actual data

94
.2 .4 .6 .8 1
1985 1990 1995 2000 2005 2010
year
CO2/N actural CO2/N AVERAGE

Figure 5-12: CO
2
per capita_ Average of forecasts from 4 models in comparison with the
actual data

50 100 150 200
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE
20 30 40 50 60
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE

95
0 5 10 15
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE
2 3 4 5
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE

4 6 8 10
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE
.8 1 1.2 1.4 1.6
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE

96
20 40 60 80 100
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE
1 2 3 4 5
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE

.5 1 1.5 2 2.5
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE
0 500 1000 1500 2000
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE

97
200 300 400 500 600
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE
0 20 40 60 80
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE

10 20 30 40
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE
0 20 40 60 80
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE

98
5 10 15 20 25 30
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE
200 300 400 500 600 700
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE

0 500 1000 1500
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE
0 50 100 150
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE

99
10 20 30 40 50 60
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE
40 60 80 100 120
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE

10 20 30 40
1985 1990 1995 2000 2005 2010
year
CO2 ACTURAL CO2 AVERAGE

Figure 5-13: CO
2
emissions_ by sector_ Average of forecasts from 4 models in
comparison with the actual data

100
.1 .1 5 .2 .2 5
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE
0 .1 .2 .3
1985 1990 1995 2000 2005 2010
year

0 .0 2 .0 4 .0 6 .0 8 .1
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE
0 .0 5 .1 .1 5
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE

101
.0 5 .1 .1 5 .2
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE
0 .0 5 .1 .1 5 .2
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE

.2 .3 .4 .5 .6
1985 1990 1995 2000 2005 2010
year
0 .0 2 .0 4 .0 6 .0 8
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE

102
0 .0 2 .0 4 .0 6 .0 8
1985 1990 1995 2000 2005 2010
year
0 5 10 15 20
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE

0 1 2 3 4 5
1985 1990 1995 2000 2005 2010
year
.1 .1 2 .1 4 .1 6 .1 8 .2
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE

103
0 .5 1 1.5
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE
.0 5 .1 .1 5 .2 .2 5
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE

0 .0 5 .1 .1 5 .2
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE
.4 .6 .8 1 1.2
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE

104
1 2 3 4 5
1985 1990 1995 2000 2005 2010
year
.2 .4 .6 .8
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE

.0 5 .1 .1 5 .2
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE
.0 4 .0 6 .0 8 .1 .1 2
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE

105
0 .1 .2 .3 .4
1985 1990 1995 2000 2005 2010
year
CO2/N ACTURAL CO2/N AVERAGE

Figure 5-14: CO
2
per capita_ by sector_ Average of forecasts from 4 models in
comparison with the actual data

106
Chapter 6 DISCUSSIONS OF RESULTS
The first finding is that the forecasted total emissions (average of the forecasts) from
2001 to 2011 grew at an annual rate of 6.21%, which is far less than the actual growth of
emissions with an annual rate of 10.01% (Table 6-1). This indicates that the national
agglomeration policy since 2001 generally aggravated the growth of carbon emissions in
manufacturing industries. Out of the four models, the GVAR model contributes the most
to the increase in forecasted emissions. The differences between the actual and forecasted
emissions (in terms of percentage change) first increase from 2001 to 2008 and then
slightly decrease until 2011. The reason for this reduction after 2008 is likely the
substantial increase in the number of emission reduction projects in China,
89
starting in
the first quarter of 2007.
90
By October 2012, these projects had generated 59.9%
91
of the
over 1 billion CERs (Certified Emission Reduction units by UNFCCC) issued through
CDM,
92
most of which are wind power projects (43%) and hydroelectric projects
(38.3%).
93
The revenues from CDM projects provided the largest source of emission
reduction financing to developing countries to date
94
.

89
The total number of emission reduction projects in China by 2006 is 36, and the number sharply increased to 150 in
2007 (which means the number of projects filed in just 2007 is 114). The data are from
http://cdm.unfccc.int/Statistics/Public/CDMinsights/index.html in the “Distribution of registered projects by Host
Party”.
90
http://cdmpipeline.org/cdm-projects-region.htm
91
http://cdm.unfccc.int/Statistics/Public/CDMinsights/index.html
92
Clean Development Mechanism-- the mechanisms under the Kyoto Protocol that makes CERs from emissions
reduction projects in developing countries tradable, or http://unfccc.int/resource/docs/publications/mechanisms.pdf
93
Figure 6-1
94
World Bank (2010). "World Development Report 2010: Development and Climate Change". Page 261-262
http://siteresources.worldbank.org/INTWDR2010/Resources/5287678-1226014527953/WDR10-Full-Text.pdf

107

Figure 6-1: “CDM project distribution within host countries by region and type”, Source
from http://cdmpipeline.org/cdm-projects-region.htm

108
CO
2
growth ate
sector 1985‐2000
2001‐2011
actual
2001‐2011
forecasted
1 1.13% 3.11% 14.01%
2 ‐1.64% 4.20% 6.88%
3 4.52% 5.09% 13.62%
4 ‐1.46% ‐0.72% ‐4.07%
5 ‐0.14% 7.35% 7.54%
6 ‐1.55% 2.00% ‐2.40%
7 2.08% 9.50% 1.59%
8 2.23% ‐2.58% 9.45%
9 2.34% ‐2.11% 5.30%
10 5.63% 10.84% ‐1.67%
11 2.77% 7.54% 7.02%
12 1.87% 4.76% 20.21%
13 5.75% ‐10.05% ‐4.85%
14 ‐1.58% 5.28% 26.99%
15 2.19% 6.40% 16.60%
16 1.41% 8.35% 13.60%
17 4.56% 12.67% 6.04%
18 6.05% 14.53% ‐0.04%
19 ‐0.02% 0.19% 17.51%
20 ‐1.72% 7.52% ‐2.47%
21 ‐3.52% 2.99% ‐2.29%
total 3.41% 10.01% 6.21%
Table 6-1: CO
2
growth rate_ by sector and the total

109
The reason why China’s agglomeration policy contributed to the growth of carbon
emissions is because the highest degree of agglomeration is concentrated in the heavy
polluting sectors.
95
Empirically, the breakdown of 21 sectors (Table 6-1) shows that there
are nine sectors
96
with an actual annual growth rate exceeding the forecasted rate, most
of which are extremely heavy polluting sectors (e.g. the Processing of Petroleum, Coking
and Nuclear (sector 10), the Manufacture of Raw Chemical Materials and Chemical
Products (sector 11) and the Smelting and Pressing of Ferrous Metals (sector 17)). There
is also a huge jump in the direct and indirect embodied carbon emissions of China’s
manufacturing export industry since 2002.
97
The concentration in heavy polluting
industries can be explained by specific agglomeration policies: one of the key ways of
promoting industrial agglomeration in China is to attract foreign investment.
98
In 2003,
FDI (foreign direct investment) attracted by China ranked first in the world
99
and
continued to grow at a steady rate. While the tax incentives brought a significant amount
of FDI to China, the lack of penalties for violating the legal environmental protection
standards and the low environmental protection fees pulled a large amount of FDI to the
heavy polluting manufacturing sectors, investment in which would have caused a
significant amount of legally imposed cost if made in the foreigners’ native countries.

95
Li, 2010. Agglomeration, environmental pollution and Regional Development
96
They are sector 4 (Manufacture of Leather, Fur, Feather and Related Products), sector 6 (Manufacture of Furniture),
sector 7 (Manufacture of Paper and Paper Products), sector 10 (Processing of Petroleum, Coking, Processing of Nuclear
Fuel), sector 11 (Manufacture of Raw Chemical Materials and Chemical Products),sector 17 (Smelting and Pressing of
Ferrous Metals), sector 18 (Smelting and Pressing of Non-ferrous Metals), sector 20 (Manufacture of Machinery) and
sector 21 (Manufacture of Artwork and Other Manufacturing).
97
Li and Luo, 2012
98
Indicated in “Relationship between Chinese FDI and the Industrial Agglomeration”, Economic Science Press, 2011,
ISBN: 9787514105421
99
http://stats.unctad.org

110
Evidence of this is found in Wu et al (2011),
100
where they calculate the average FDI
ratio
101
for manufacturing sub-sectors from 1999 to 2008. Out of the top five sectors
102

with the highest FDI ratios, four
103
of them are categorized as high polluting
104
sectors
according to the classification by Smarzynska and Wei (2001).
105

The second major finding of this paper is that while agglomeration policy increased
industrial emissions, it decreased emissions per capita (Table 6-2). Compared to the
forecasted 0.90% annual increase, the actual rate of emissions per capita decreased at a
yearly rate of 0.40% since 2001. One of the contributing factors to the faster increase in
population concentration could be the urbanization from the “Tenth five year plan”, in
which the agglomeration policy emphasized the concentration of populations in small
towns to form new urban agglomerations and to facilitate the formation of industrial
cities. From 2001 to 2011, the urbanization ratio increased from 37.66% to 51.27%.
106
In
2011, 36.5% of new migrants of previously rural household entered the manufacturing
industry,
107
ranked 2
nd
after to the services sectors which had a proportion of 39.8%.

100
Wu et al, 2011. “FDI, Industrial Character and Geographic Concentration of Manufacturing Industry—Empirical
Study from Chinese data”.
101
Calculated as the ratio of sales revenue from FDI enterprises and the total sales revenue of all enterprises.
102
Manufacture of Machinery (sector 20), Processing and Manufacture of Food (sector 1), Smelting and Pressing of
Ferrous Metals (sector 17), Manufacture of Paper and Paper Products (sector 7), Manufacture of Raw Chemical
Materials and Chemical Products (sector 11)
103
Processing and Manufacture of Food (sector 1), Smelting and Pressing of Ferrous Metals (sector 17), Manufacture
of Paper and Paper Products (sector 7), Manufacture of Raw Chemical Materials and Chemical Products (sector 11)
104
Smarzynska and Wei, 2001 classify the industrial sectors as High Pollution, Medium Pollution and Low Pollution
by their pollution intensities.
105
Smarzynska and Wei, 2001, “Pollution Havens and Foreign Direct Investment: Dirty Secret of Popular Myth?”
106
Data are from the “Sixth National Census” and “2011 China Statistical Bulletin”
107
Wang 2011, “Report on China's Migrant Population Development”.

111
CO
2
per capita growth rate
sector 1985‐2000
2001‐2011
actural
2001‐2011
forecasted
1 1.63% ‐6.54% ‐2.23%
2 ‐7.85% ‐2.57% ‐5.72%
3 ‐8.35% ‐6.31% 1.98%
4 ‐7.85% ‐15.03% ‐1.86%
5 ‐1.36% ‐7.90% ‐3.13%
6 ‐7.48% ‐15.92% ‐2.20%
7 1.25% 0.31% ‐2.12%
8 1.19% ‐5.67% 3.83%
9 ‐4.58% ‐15.22% 2.41%
10 3.31% 5.07% ‐1.93%
11 2.91% 0.43% 15.41%
12 ‐2.13% ‐3.09% 1.02%
13 3.21% ‐13.92% ‐1.39%
14 ‐2.62% ‐3.34% 0.42%
15 ‐5.72% ‐8.49% ‐0.02%
16 1.68% ‐0.58% 0.43%
17 4.75% 7.08% 3.78%
18 3.04% 4.71% 2.87%
19 ‐4.20% ‐11.57% ‐3.02%
20 ‐1.38% ‐4.44% ‐2.62%
21 ‐8.47% ‐3.15% 2.62%
total 2.18% ‐0.40% 0.90%
Table 6-2: CO
2
per capita_ by sector and the total

112
The third finding is that there was a general decrease of energy efficiency in the
manufacturing industry overall. Although the actual annual growth rate of manufacturing
output greatly exceeded the forecasted rate (20.57% versus 3.94%
108
), energy
consumption grew at an even faster rate, driven by faster actual growth in energy
intensity (at an annual rate of 8.58% compared to a forecasted rate of -1.60% (Table 6-4)).
There are 20 sectors
109
with an actual growth rate of energy intensity exceeding the
forecasted growth rate since 2001, and the energy inefficiency problem is particularly
serious in three sectors: sector 5 (Processing of Timber, Manufacture of Wood, Bamboo,
Rattan, Palm and Straw Products), sector 17 (Smelting and Pressing of Ferrous Metals)
and sector 18 (Smelting and Pressing of Non-ferrous Metals).

108
Table 6-3
109
All sectors with the exception of the Manufacture of Chemical Fibers (sector 13)

113

Output growth rate

sector 1985‐2000
2001‐2011
actural
2001‐2011
forecasted
1 7.19% 20.04% 3.52%
2 2.60% 15.82% 1.92%
3 9.29% 14.57% 3.91%
4 12.43% 15.55% 4.25%
5 9.28% 24.68% 3.96%
6 6.14% 24.22% 3.27%
7 8.84% 17.47% 3.84%
8 5.62% 14.80% 3.12%
9 11.25% 13.44% 4.15%
10 12.97% 19.65% 4.16%
11 8.66% 21.85% 3.81%
12 9.90% 18.53% 4.00%
13 9.63% 17.18% 4.01%
14 4.21% 19.88% 3.12%
15 9.61% 18.48% 4.89%
16 8.18% 22.26% 4.30%
17 8.98% 23.70% 3.77%
18 9.46% 27.47% 4.93%
19 8.90% 19.86% 3.99%
20 9.58% 21.17% 4.25%
21 10.15% 16.99% 4.27%
total 8.36% 20.57% 3.94%
Table 6-3: Output_ by sector and the total

114
Energy intensity growth rate
sector 1985‐2000
2001‐2011
actural
2001‐2011
forecasted
1 1.98% 5.04% ‐1.96%
2 0.24% 8.22% ‐7.71%
3 7.76% 8.10% 0.01%
4 2.56% 6.56% ‐4.08%
5 1.71% 12.28% ‐4.29%
6 0.82% 7.22% ‐3.17%
7 2.33% 6.77% ‐1.94%
8 6.63% 6.43% 0.68%
9 7.22% 5.67% ‐0.29%
10 10.54% 8.60% ‐0.06%
11 3.04% 9.98% ‐1.63%
12 3.52% 5.29% ‐1.47%
13 8.23% ‐1.56% ‐0.20%
14 2.88% 8.39% 0.22%
15 6.67% 11.51% ‐1.13%
16 1.60% 10.13% ‐3.42%
17 5.71% 12.55% ‐0.65%
18 7.83% 13.20% ‐1.21%
19 2.80% 10.77% ‐1.89%
20 1.28% 11.58% ‐3.31%
21 1.86% 2.61% ‐2.73%
total 4.20% 8.58% ‐1.60%
Table 6-4: Energy intensity_ by sector and the total

115
In summary, this paper constructed industrial emissions data for China’s 21
manufacturing sectors by calculating forecasts that take the average across four types of
econometric models. It conducts a program evaluation of China’s agglomeration policy
and its effect on manufacturing CO
2
emissions. Results indicate that the agglomeration
policy generally aggravated manufacturing emissions since 2001, due to the rapid
expansion of heavy polluting sectors, and slowed down the emissions per capita through
urbanization, while energy efficiency also declined. As the government plays an
important role in the formation and development of industrial agglomeration process,
improving the degree of concentration should not be the only goal. It is of central
importance to balance the need for economic growth
110
with the importance of the
environmental carrying capacity.

110
In 2006, in "The notice from State Council on accelerating the restructuring of industry overcapacity" (Issued in the
No. 11 document by NDRC), the central government officially confirmed, for the first time, the existence of
overcapacity problem in the manufacturing industries. In 2013, China's manufacturing capacity utilization rate dropped
to 78%, which was lower than the international standard of 79% (the cutoff level to determine whether there is an
overcapacity issue).

116
Reference

Albino, V., Dangelico, R.M., Natalicchio A., and Yazan, D.M. (2011): “Alternative
energy sources in cement manufacturing: A Systematic Review of the Body of
Knowledge”, Department of Mechanical and Management Engineering, Politecnico di
Bari Italy.

Auffhammer, M., and Carson, R.T. (2008): “Forecasting the path of China's
CO2 emissions using province-level information”, Journal of Environmental Economics
and Management, 55 (3), 229–247.

Brown, T., Gambhir, A., Florin N., and Fennell, P. (2012): “Reducing CO2 emissions
from heavy industry: a review of technologies and considerations for policy makers”,
Grantham Institute for Climate Change, Briefing paper No.7.
Cao, J., Garbaccio, R., and Ho, M. (2009): “Benefits and Costs of SO_2 Abatement
Policies in ChinA”, Review of Environmental Economics and Policy, 3(2), 231-250.

Casler, S.D., and Rose. A. (1998): “Carbon Dioxide Emissions in the U.S. Economy, A
Structural Decomposition Analysis”, Environmental and Resource Economics, 11(3-4),
349-363.

117

Doda, B. (2012): “Evidence on CO2 emissions and business cycles”, Grantham Research
Institute on Climate Change and the Environment London School of Economics.

Fan, J.Y. and Li. F.W. (2011): “Spatial agglomeration in Chinese manufacturing
industry”, South China Journal of Economics, 6.

Garbaccio, R.F., Ho, M.S., and Jorgenson, D.W. (1999): “Why has the nergy-output ratio
fallen in China”, Energy journal, 20(3), 63-91.

Guo, J., Liu, C.X., and Sun. P. (2013): “Carbon Emissions from Industrial Sectors in
China: Driving Factors and the Potential for Emission Reduction”, Journal of Resources
and Ecology, 4(2), 132-140.

Guo, L.P. (2011): “Relationship between Chinese FDI and the Industrial Agglomeration”
Economic Science Press.

Hao, Z.Z., and Jian, L. (2013): “The adjustment and optimized path of regional industrial
structure in low-carbon economy society in China: A case study of Tianjin”, African
Journal of Business, 7(30), 3022-3034.

118
Hasanbeigi, A., Price, L., Kong, L.B., and Arens, M. (2013): “Emerging
Energy-Efficiency and CO2 Emissions-Reduction Technologies for Industry: A Review
of Technologies for Alternative Ironmaking and Pulp and Paper Industry”, Lawrence
Berkeley National Laboratory and South China University of Technology and Fraunhofer
Institute for Systems and Innovation Research, working paper.

He, C.F., Zhu, Y.G, and Zhu, M.J. (2010): “Industry characteristics, regional
characteristics and Chinese manufacturing industries agglomeration”, ACTA
GEOGRAPHICA SINICA, 65 (10).

Hou, F.Q. (2008): “Environmental Effects of China's regional economic agglomeration”,
Northwest Agriculture and Forestry University (Social Sciences), 5, 20-25.

IEA. (2007): “Tracking Industrial Energy Efficiency and CO2 Emissions”.

IPCC. (2006): “2006 IPCC Guidelines for National Greenhouse Gas Inventories”

Jalil, A., and Mahmud, S.F. (2009): “Environment Kuznets curve for CO2 emissions: A
cointegration analysis for China”, Energy Policy, 37 (12), 5167–5172.

Jin, X.R., and Zhu, X,W. (2002): “The origin and development of specialized industrial

119
zone”, Economic Research, 08, 74-82.

Li, S.S, and Luo L.W. (2012): “The impact of FDI structure on the implied carbon
emissions during the foreign trade”, Resources Science, 34 (5).

Li, J., Wang, S., and Zhang, J.H. (2012): “Elasticity analysis and path of Carbon
Emissions in the Regional Industry Clusters”, Soft Science, 7, 70-74.

Li, W.N. (2010): “Agglomeration, environmental pollution and Regional Development”,
Modern Management Science, 3, 47-48.

Liu, J., and Xu, K.N. (2010): “Industry agglomeration, Economic Growth and Regional
Disparity - An Empirical Study Based on Chinese Provincial Panel Data”, China Soft
Science, 07, 91-102.

Liu, J., and Xu, K.N. (2009): “History of Chinese industry agglomeration”, Southeast
University (Philosophy and Social Sciences), 05, 54-58.

Lu, H.Y., and Price, L. (2012): “China’s Industrial Carbon DioxideEmissions in
Manufacturing
Subsectors and in Selected Provinces”, Lawrence Berkeley National Laboratory, working

120
paper.

Ma. C.B., and Stern, D.I. (2007): “China's Carbon Emissions1971-2003”, Rensselaer
Polytechnic Institute, Working Papers.

Ma, W., Smith, S.J., Sinha, P., and Lurz, J.P. (2007 ): “Long-Term US Industrial Energy
Use and CO2 Emissions”, Pacific Northwest National Laboratory, working paper.

Office of Air and Radiation. (2010): “Available and Emerging Technologies for
Reduceing Greenhouse gas Emissions from the Pulp and Paper Manufacturing Industry”,
EPA.

Pan, X.F., Shu, T., and Xu, D.W. (2011): “On the Changes in the Carbon Emission
Intensity of China’s Manufacturing Industry and Its Factors Decomposition”, China
population, resources and environment, 21 (5).

Shaoa, S., and Yanga, L., Yub, M., and Yu, M. (2011): “Estimation, characteristics, and
determinants of energy-related industrial CO2 emissions in Shanghai (China),
1994–2009”, Energy Policy, 39 (10).

Shen, J.Y., and Hashimoto. Y. (2004): “Environmental Kuznets Curve on Country Level:

121
Evidence from China”, Osaka School of International Public Policy (OSIPP), working
paper.

Smarzynska, B.K., and Wei, S.G. (2001): “Pollution Havens and Foreign Direct
Investment: Dirty Secret of Popular Myth?”, NBER working paper.

Sun, N. (2011): “Carbon Emission Reduction of Manufacturing Rely on Technical
Progres”, Forum on Science and Technology in China, 4.

The Climate Change Working Group. (2005): “Calculation Tools for Estimating
Greenhouse Gas Emissions from Pulp and Paper Mills”, ICFPA.

UNFCCC (2007): “The Kyoto Protocol Mechanisms: International Emissions Trading
Clean Development Mechanism Joint Implementation”.

Vos, R.O., and Newell, J. (2009): “A Comparative Analysis of Carbon Dioxide
Emissions in Coated Paper Production: Key Differences between China and the U.S.”,
Center for Sustainable Cities, University of Southern California, working paper.

Wang, P.A. (2011): “Report on China's Migrant Population Development”, China
Population Publishing House.

122

Wang, H.N., and Chen, Y.Y. (2010): “Industrial cluster and Energy Efficiency”, Journal
of Finance and Economics, 9, 69-79.

World Bank. (2010): “World Development Report 2010: Development and Climate
Change”.

Wu, Y.H., Bo, W.G., and Yin G.W. (2011): “FDI, Industrial Character and Geographic
Concentration of Manufacturing Industry—Empirical Study from Chinese data”, Nankai
Economic Studies, 1.

Xu, T.S., and Xue, Z.Y. (2012): “Carbon Emission in Manufacturing Export: Total
Quantity, Structure and Factor Decomposition”, Finance and Trade Research, 2.

Xu, Y.Z., Xu K.N, and Hu Y.S. (2011): “Driving Factors and Decoupling Effect of
Carbon Emissions: Evidence from China’s Manufacturing Sector”, Statistical Research,
28 (7).

Xu, K.N. (2006): “The source of the formation of industrial agglomeration”,
People's Press.

123
Young, A. (2000): “The Razor’s Edge: Distortions and Incremental Reform in the
People’s Republic of China”, The Quarterly Journal of Economics, 115 (4), 1091-1135.

124
Appendix A: Energy Consumption
(autocorrelation, partial autocorrelation and time series
forecast)
Table A-1: Autocorrelation graphs for Energy consumption _ by sector
-0.50 0.00 0.50
A utoco rrelatio ns of D .e1
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-1.00 -0.50 0.00 0.50 1.00
A utoco rrelatio ns of D .e2
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A utoco rrelatio ns of D .e3
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rrelatio ns of D .e4
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

125
Table A-1: Autocorrelation graphs for Energy consumption_by sector_ continued
-0.50 0.00 0.50
A utoco rrelatio ns of D .e5
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rrelatio ns of D .e6
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A utoco rrelatio ns of D .e7
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-1.00 -0.50 0.00 0.50 1.00
A utoco rrelatio ns of D .e8
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

126
Table A-1: Autocorrelation graphs for Energy consumption_by sector_ continued
-1.00 -0.50 0.00 0.50 1.00
A utoco rrelatio ns of D .e9
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-1.00 -0.50 0.00 0.50 1.00
A utoco rrela tio ns of D .e1 0
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A utoco rrela tio ns of D .e1 1
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rrela tio ns of D .e1 2
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

127
Table A-1: Autocorrelation graphs for Energy consumption_by sector_ continued
-1.00 -0.50 0.00 0.50 1.00
A utoco rrela tio ns of D .e1 3
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-1.00 -0.50 0.00 0.50 1.00
A utoco rrela tio ns of D .e1 4
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A utoco rrela tio ns of D .e1 5
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-1.00 -0.50 0.00 0.50 1.00
A utoco rrela tio ns of D .e1 6
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

128
Table A-1: Autocorrelation graphs for Energy consumption_by sector_ continued
-0.50 0.00 0.50
A utoco rrela tio ns of D .e1 7
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rrela tio ns of D .e1 8
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A utoco rrela tio ns of D .e1 9
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-1.00 -0.50 0.00 0.50 1.00
A utoco rrela tio ns of D .e2 0
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

129
Table A-1: Autocorrelation graphs for Energy consumption_by sector_ continued
-0.50 0.00 0.50
A utoco rrela tio ns of D .e2 1
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

130
Table A-2: Partial Autocorrelation graphs for Energy consumption_by sector
-6.00 -4.00 -2.00 0.00 2.00
P a rtia l au to corre la tions of D .e 1
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P a rtia l au to corre la tions of D .e 2
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.50 0.00 0.50
P a rtia l au to corre la tions of D .e 3
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P a rtia l au to corre la tions of D .e 4
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

131
Table A-2: Partial Autocorrelation graphs for Energy consumption_by sector_ continued
-0.50 0.00 0.50
P a rtia l au to corre la tions of D .e 5
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-6.00 -4.00 -2.00 0.00
P a rtia l au to corre la tions of D .e 6
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-3.00 -2.00 -1.00 0.00 1.00
P a rtia l au to corre la tions of D .e 7
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P a rtia l au to corre la tions of D .e 8
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

132
Table A-2: Partial Autocorrelation graphs for Energy consumption_by sector_ continued
-2.00 -1.50 -1.00 -0.50 0.00 0.50
P a rtia l au to corre la tions of D .e 9
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-1.00 -0.50 0.00 0.50 1.00
P artia l au to corre la tio ns of D .e 10
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.50 0.00 0.50
P artia l au to corre la tio ns of D .e 11
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-1.00 -0.50 0.00 0.50
P artia l au to corre la tio ns of D .e 12
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

133
Table A-2: Partial Autocorrelation graphs for Energy consumption_by sector_ continued
-1.00 0.00 1.00 2.00 3.00
P artia l au to corre la tio ns of D .e 13
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50 1.00 1.50 2.00
P artia l au to corre la tio ns of D .e 14
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.50 0.00 0.50 1.00
P artia l au to corre la tio ns of D .e 15
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-1.00 -0.50 0.00 0.50
P artia l au to corre la tio ns of D .e 16
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

134
Table A-2: Partial Autocorrelation graphs for Energy consumption_by sector_ continued
-0.50 0.00 0.50
P artia l au to corre la tio ns of D .e 17
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P artia l au to corre la tio ns of D .e 18
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-1.00 -0.50 0.00 0.50
P artia l au to corre la tio ns of D .e 19
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P artia l au to corre la tio ns of D .e 20
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

135
Table A-2: Partial Autocorrelation graphs for Energy consumption_by sector_ continued
-0.50 0.00 0.50
P artia l au to corre la tio ns of D .e 21
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

136
Table A-3: Time series forecast by sector __ Energy intensity
2000 3000 4000 5000 6000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E1
1000 2000 3000 4000 5000 6000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E2

0 200 400 600 800
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E3
100 200 300 400
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E4

137
Table A-3: Time series forecast by sector __ Energy intensity_ continued
200 400 600 800 1000 1200
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E5
50 100 150 200
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E6

1000 2000 3000 4000
1985 1990 1995 2000 2005 2010
year
0 100 200 300 400
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E8

138
Table A-3: Time series forecast by sector __ Energy intensity_ continued
50 100 150 200 250
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E9
0 5000 10000 15000 20000
1985 1990 1995 2000 2005 2010
year

0 10000 20000 30000 40000
1985 1990 1995 2000 2005 2010
year
500 1000 1500
1985 1990 1995 2000 2005 2010
year

139
Table A-3: Time series forecast by sector __ Energy intensity_ continued
500 1000 1500 2000 2500
1985 1990 1995 2000 2005 2010
year
500 1000 1500
1985 1990 1995 2000 2005 2010
year

0 500 1000 1500 2000
1985 1990 1995 2000 2005 2010
year
10000 15000 20000 25000 30000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E16

140
Table A-3: Time series forecast by sector __ Energy intensity_ continued
0 20000 40000 60000
1985 1990 1995 2000 2005 2010
year
0 5000 10000 15000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E18

0 1000 2000 3000 4000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E19
4000 6000 8000 10000 12000 14000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E20

141

Table A-3: Time series forecast by sector __ Energy intensity_ continued
1000 1500 2000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) E21

142
Appendix B: Output
(autocorrelation, partial autocorrelation and time series
forecast)

Table B-1: Autocorrelation graphs for Output_by sector
-0.50 0.00 0.50
A u toco rrelatio ns of D .y1
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A u toco rrelatio ns of D .y2
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-1.00 -0.50 0.00 0.50 1.00
A u toco rrelatio ns of D .y3
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A u toco rrelatio ns of D .y4
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

143
Table B-1: Autocorrelation graphs for Output_by sector_ continued
-0.50 0.00 0.50
A u toco rrelatio ns of D .y5
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A u toco rrelatio ns of D .y6
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A u toco rrelatio ns of D .y7
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A u toco rrelatio ns of D .y8
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

144
Table B-1: Autocorrelation graphs for Output_by sector_ continued
-0.50 0.00 0.50
A u toco rrelatio ns of D .y9
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A uto co rrela tio ns of D .y10
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A uto co rrela tio ns of D .y11
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A uto co rrela tio ns of D .y12
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

145
Table B-1: Autocorrelation graphs for Output_by sector_ continued
-0.50 0.00 0.50
A uto co rrela tio ns of D .y13
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A uto co rrela tio ns of D .y14
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A uto co rrela tio ns of D .y15
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A uto co rrela tio ns of D .y16
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

146
Table B-1: Autocorrelation graphs for Output_by sector_ continued
-0.50 0.00 0.50
A uto co rrela tio ns of D .y17
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-1.00 -0.50 0.00 0.50 1.00
A uto co rrela tio ns of D .y18
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A uto co rrela tio ns of D .y19
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A uto co rrela tio ns of D .y20
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

147
Table B-1: Autocorrelation graphs for Output_by sector_ continued
-0.50 0.00 0.50
A uto co rrela tio ns of D .y21
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

148
Table B-2: Partial Autocorrelation graphs for Output_by sector
-0.50 0.00 0.50
P artia l au to corre la tions of D .y1
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P artia l au to corre la tions of D .y2
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.50 0.00 0.50
P artia l au to corre la tions of D .y3
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P artia l au to corre la tions of D .y4
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

149
Table B-2: Partial Autocorrelation graphs for Output_by sector_ continued
-0.50 0.00 0.50
P artia l au to corre la tions of D .y5
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P artia l au to corre la tions of D .y6
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-1.00 -0.50 0.00 0.50
P artia l au to corre la tions of D .y7
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-1.00 -0.50 0.00 0.50
P artia l au to corre la tions of D .y8
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

150
Table B-2: Partial Autocorrelation graphs for Output_by sector_ continued
-0.50 0.00 0.50
P artia l au to corre la tions of D .y9
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-1.50 -1.00 -0.50 0.00 0.50 1.00
P a rtia l a u to corre la tion s o f D .y1 0
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.50 0.00 0.50
P a rtia l au to corre la tions of D .y1 1
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P a rtia l au to corre la tions of D .y1 2
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

151
Table B-2: Partial Autocorrelation graphs for Output_by sector_ continued
-2.00 -1.50 -1.00 -0.50 0.00 0.50
P a rtia l a u to corre la tion s o f D .y1 3
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P a rtia l a u to corre la tion s o f D .y1 4
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.50 0.00 0.50
P a rtia l au to corre la tions of D .y1 5
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P a rtia l au to corre la tions of D .y1 6
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

152
Table B-2: Partial Autocorrelation graphs for Output_by sector_ continued
-0.50 0.00 0.50
P a rtia l a u to corre la tion s o f D .y1 7
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-1.00 -0.50 0.00 0.50 1.00
P a rtia l a u to corre la tion s o f D .y1 8
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.50 0.00 0.50
P a rtia l au to corre la tions of D .y1 9
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
P a rtia l au to corre la tions of D .y2 0
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

153
Table B-2: Partial Autocorrelation graphs for Output_by sector_ continued
-0.50 0.00 0.50
P a rtia l a u to corre la tion s o f D .y2 1
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

154
Table B-3: Time series forecast by sector __ Output
0 10000 20000 30000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y1
0 5000 10000 15000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y2

0 2000 4000 6000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y3
0 1000 2000 3000 4000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y4

155
Table B-3: Time series forecast by sector __ Output_ continued
0 1000 2000 3000 4000
1985 1990 1995 2000 2005 2010
year
0 500 1000 1500 2000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y6

0 1000 2000 3000 4000 5000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y7
0 500 1000 1500
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y8

156
Table B-3: Time series forecast by sector __ Output_ continued
0 500 1000 1500
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y9
0 5000 10000 15000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y10

0 5000 10000 15000 20000 25000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y11
0 2000 4000 6000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y12

157
Table B-3: Time series forecast by sector __ Output_ continued
0 500 1000 1500 2000 2500
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y13
0 1000 2000 3000
1985 1990 1995 2000 2005 2010
year

0 2000 4000 6000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y15
0 5000 10000 15000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y16

158
Table B-3: Time series forecast by sector __ Output_ continued
0 5000 10000 15000 20000 25000
1985 1990 1995 2000 2005 2010
0 5000 10000 15000
1985 1990 1995 2000 2005 2010
year

0 2000 4000 6000 8000 10000
1985 1990 1995 2000 2005 2010
year
0 20000 40000 60000 80000 100000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y20

159
Table B-3: Time series forecast by sector __ Output_ continued
0 1000 2000 3000
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) Y21

160
Appendix C: Employment
(autocorrelation, partial autocorrelation and time series
forecast)

Table C-1: Autocorrelation graphs for Employment_by sector
-0.50 0.00 0.50
A utoco rrelatio ns of D .n1
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A u toco rrelatio ns of n 2
1 2 3 4 5 6
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A u toco rrelatio ns of n 3
1 2 3 4 5 6
Lag
Bartlett's formula for MA(q) 95% confidence bands
-1.00 -0.50 0.00 0.50 1.00
A u toco rrelatio ns of n 4
1 2 3 4 5 6
Lag
Bartlett's formula for MA(q) 95% confidence bands

161
Table C-1: Autocorrelation graphs for Employment_by sector_ continued
-0.50 0.00 0.50 1.00
P a rtia l au to corre la tions of D .n 5
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
A utoco rrelatio ns of D .n5
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.60 -0.40 -0.20 0.00 0.20 0.40
P artia l a u to co rre la tio ns of n 6
1 2 3 4 5 6
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
A u toco rrelatio ns of n 6
1 2 3 4 5 6
Lag
Bartlett's formula for MA(q) 95% confidence bands

162
Table C-1: Autocorrelation graphs for Employment_by sector_ continued
-1.00 -0.50 0.00 0.50 1.00
A u toco rrelatio ns of n 7
1 2 3 4 5 6
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rrelatio ns of D .n8
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-1.00 -0.50 0.00 0.50 1.00
A u toco rrelatio ns of n 9
1 2 3 4 5 6
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rrela tio ns of D .n1 0
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

163
Table C-1: Autocorrelation graphs for Employment_by sector_ continued
-0.50 0.00 0.50 1.00 1.50
P artia l au to corre la tio ns of D .n 11
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
A utoco rrela tio ns of D .n1 1
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50 1.00 1.50
P artia l au to corre la tio ns of D .n 12
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50
A utoco rrela tio ns of D .n1 2
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

164
Table C-1: Autocorrelation graphs for Employment_by sector_ continued
-0.50 0.00 0.50
A utoco rrela tio ns of D .n1 3
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rrela tio ns of D .n1 4
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A utoco rrelatio ns of n 1 5
1 2 3 4 5 6
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rrela tio ns of D .n1 6
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

165
Table C-1: Autocorrelation graphs for Employment_by sector_ continued
-0.50 0.00 0.50
A utoco rrela tio ns of D .n1 7
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-1.00 -0.50 0.00 0.50 1.00
A utoco rrelatio ns of n 1 8
1 2 3 4 5 6
Lag
Bartlett's formula for MA(q) 95% confidence bands

-0.50 0.00 0.50
A utoco rrelatio ns of n 1 9
1 2 3 4 5 6
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
A utoco rrela tio ns of D .n2 0
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands

166
Table C-1: Autocorrelation graphs for Employment_by sector_ continued
-0.50 0.00 0.50
A utoco rrelatio ns of n 2 1
1 2 3 4 5 6
Lag
Bartlett's formula for MA(q) 95% confidence bands

167
Table C-2: Partial Autocorrelation graphs for Employment_by sector
-0.50 0.00 0.50 1.00 1.50
P a rtia l au to corre la tions of D .n 1
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.60 -0.40 -0.20 0.00 0.20 0.40
P artia l a u to co rre la tio ns of n 2
1 2 3 4 5 6
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.50 0.00 0.50
P artia l a u to co rre la tio ns of n 3
1 2 3 4 5 6
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.60 -0.40 -0.20 0.00 0.20 0.40
P artia l a u to co rre la tio ns of n 4
1 2 3 4 5 6
Lag
95% Confidence bands [se = 1/sqrt(n)]

168
Table C-2: Partial Autocorrelation graphs for Employment_by sector_ continued
-0.50 0.00 0.50 1.00
P a rtia l au to corre la tions of D .n 5
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.60 -0.40 -0.20 0.00 0.20 0.40
P artia l a u to co rre la tio ns of n 6
1 2 3 4 5 6
Lag
95% Confidence bands [se = 1/sqrt(n)]

-1.50 -1.00 -0.50 0.00 0.50
P artia l a u to co rre la tio ns of n 7
1 2 3 4 5 6
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50 1.00
P a rtia l au to corre la tions of D .n 8
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

169
Table C-2: Partial Autocorrelation graphs for Employment_by sector_ continued
-0.50 0.00 0.50
A utoco rrelatio ns of D .n8
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
P artia l a u to co rre la tio ns of n 9
1 2 3 4 5 6
Lag
95% Confidence bands [se = 1/sqrt(n)]

-1.00 -0.50 0.00 0.50 1.00
A u toco rrelatio ns of n 9
1 2 3 4 5 6
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50
P artia l au to corre la tio ns of D .n 10
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

170
Table C-2: Partial Autocorrelation graphs for Employment_by sector_ continued
-0.50 0.00 0.50 1.00 1.50
P artia l au to corre la tio ns of D .n 11
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50 1.00 1.50
P artia l au to corre la tio ns of D .n 12
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.50 0.00 0.50 1.00 1.50
P artia l au to corre la tio ns of D .n 13
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50 1.00 1.50
P artia l au to corre la tio ns of D .n 14
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

171
Table C-2: Partial Autocorrelation graphs for Employment_by sector_ continued
-0.50 0.00 0.50
A utoco rrela tio ns of D .n1 4
1 2 3 4 5
Lag
Bartlett's formula for MA(q) 95% confidence bands
-1.00 -0.50 0.00 0.50
P artia l au to corre la tions of n 15
1 2 3 4 5 6
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.50 0.00 0.50
A utoco rre lation s o f n1 5
1 2 3 4 5 6
Lag
Bartlett's formula for MA(q) 95% confidence bands
-0.50 0.00 0.50 1.00
P artia l au to corre la tio ns of D .n 16
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

172
Table C-2: Partial Autocorrelation graphs for Employment_by sector_ continued
-1.00 -0.50 0.00 0.50
P artia l au to corre la tio ns of D .n 17
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.40 -0.20 0.00 0.20 0.40 0.60
P artia l au to corre la tions of n 18
1 2 3 4 5 6
Lag
95% Confidence bands [se = 1/sqrt(n)]

-0.60 -0.40 -0.20 0.00 0.20 0.40
P artia l a u to corre la tion s o f n 19
1 2 3 4 5 6
Lag
95% Confidence bands [se = 1/sqrt(n)]
-0.50 0.00 0.50 1.00
P artia l au to corre la tio ns of D .n 20
1 2 3 4 5
Lag
95% Confidence bands [se = 1/sqrt(n)]

173
Table C-2: Partial Autocorrelation graphs for Employment_by sector_ continued
-1.00 -0.50 0.00 0.50
P artia l au to corre la tions of n 21
1 2 3 4 5 6
Lag
95% Confidence bands [se = 1/sqrt(n)]

174
Table C-3: Time series forecast by sector __Employment
200 300 400 500 600 700
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) N1
0 200 400 600 800
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) N2

0 100 200 300 400 500
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) N3
0 100 200 300
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) N4

175
Table C-3: Time series forecast by sector __Employment_ continued
0 50 100 150
1985 1990 1995 2000 2005 2010
year
0 50 100
1985 1990 1995 2000 2005 2010
year

176
Table C-3: Time series forecast by sector __Employment_ continued
0 50 100 150
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) N9

200 300 400 500
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) N11

177
Table C-3: Time series forecast by sector __Employment_ continued
40 60 80 100 120
1985 1990 1995 2000 2005 2010
year

0 100 200 300
1985 1990 1995 2000 2005 2010
year
200 400 600 800
1985 1990 1995 2000 2005 2010
year

178
Table C-3: Time series forecast by sector __Employment_ continued
200 250 300 350
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) N17
50 100 150 200

0 100 200 300 400
1000 1500 2000 2500 3000
1985 1990 1995 2000 2005 2010
year

179
Table C-3: Time series forecast by sector __Employment_ continued
50 100 150 200 250
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) N21

180
Appendix D: Time series forecast of Purchasing price of fuel
and raw materials

Table D-1: Time series forecast __Purchasing price of fuel and raw materials
0 100 200 300 400
1985 1990 1995 2000 2005 2010
year
y prediction, dyn(2001) P

181
Appendix E: Weight matrix for GV AR model

Table E-1: Weight matrix for GV AR model
Country ARTWORK CHEMICAL CULTURE FERROUS FIBERS FOOD FURNITURE LEATHER MACHINERY MEDICINES
ARTWORK 0.0000 0.0574 0.0631 0.0575 0.0605 0.0601 0.0690 0.0595 0.0568 0.0590
CHEMICAL 0.0122 0.0000 0.0128 0.0117 0.0123 0.0116 0.0140 0.0120 0.0115 0.0120
CULTURE 0.1089 0.1042 0.0000 0.1044 0.1098 0.1071 0.1251 0.1079 0.1031 0.1070
FERROUS 0.0147 0.0140 0.0154 0.0000 0.0148 0.0137 0.0168 0.0145 0.0139 0.0144
FIBERS 0.0655 0.0625 0.0689 0.0627 0.0000 0.0638 0.0750 0.0648 0.0619 0.0643
FOOD 0.0074 0.0071 0.0078 0.0071 0.0074 0.0000 0.0085 0.0073 0.0070 0.0073
FURNITURE 0.1867 0.1780 0.1964 0.1785 0.1876 0.1728 0.0000 0.1844 0.1767 0.1835
LEATHER 0.0486 0.0464 0.0511 0.0465 0.0490 0.0467 0.0557 0.0000 0.0461 0.0478
MACHINERY 0.0032 0.0030 0.0033 0.0031 0.0032 0.0029 0.0037 0.0032 0.0000 0.0031
MEDICINES 0.0422 0.0403 0.0443 0.0404 0.0424 0.0394 0.0484 0.0418 0.0399 0.0000
METAL 0.0270 0.0258 0.0284 0.0258 0.0272 0.0259 0.0309 0.0267 0.0255 0.0265
NONFERROUS 0.0357 0.0341 0.0375 0.0342 0.0359 0.0335 0.0410 0.0354 0.0337 0.0350
NONMETALIC 0.0168 0.0160 0.0177 0.0161 0.0169 0.0162 0.0192 0.0166 0.0159 0.0165
PAPER 0.0456 0.0435 0.0479 0.0436 0.0458 0.0426 0.0522 0.0451 0.0431 0.0448
PETROLEUM 0.0230 0.0219 0.0241 0.0219 0.0230 0.0198 0.0262 0.0227 0.0217 0.0225
PLASTICS 0.0381 0.0364 0.0401 0.0365 0.0384 0.0359 0.0437 0.0377 0.0361 0.0374
PRINTING 0.1026 0.0978 0.1078 0.0981 0.1031 0.0977 0.1173 0.1013 0.0971 0.1008
RUBBER 0.0748 0.0713 0.0786 0.0715 0.0752 0.0741 0.0855 0.0739 0.0708 0.0735
TEXTILE 0.0128 0.0122 0.0134 0.0122 0.0128 0.0123 0.0146 0.0126 0.0121 0.0125
TIMBER 0.1049 0.0997 0.1104 0.1001 0.1051 0.0953 0.1194 0.1034 0.0992 0.1031
WEARING 0.0295 0.0282 0.0310 0.0282 0.0297 0.0286 0.0338 0.0292 0.0279 0.0289

182
Table E-1: Weight matrix for GV AR model_ continued
Country METAL NONFERROUS NONMETALIC PAPER PETROLEUM PLASTICS PRINTING RUBBER TEXTILE TIMBER WEARING
ARTWORK 0.0582 0.0586 0.0579 0.0593 0.0581 0.0588 0.0629 0.0611 0.0575 0.0632 0.0583
CHEMICAL 0.0118 0.0119 0.0117 0.0120 0.0118 0.0119 0.0127 0.0124 0.0116 0.0127 0.0118
CULTURE 0.1055 0.1062 0.1051 0.1076 0.1054 0.1067 0.1141 0.1108 0.1044 0.1148 0.1056
FERROUS 0.0142 0.0143 0.0141 0.0145 0.0142 0.0144 0.0154 0.0149 0.0140 0.0154 0.0143
FIBERS 0.0634 0.0639 0.0629 0.0646 0.0631 0.0641 0.0684 0.0665 0.0626 0.0685 0.0636
FOOD 0.0072 0.0072 0.0071 0.0073 0.0071 0.0072 0.0077 0.0075 0.0071 0.0077 0.0072
FURNITURE 0.1807 0.1824 0.1784 0.1839 0.1798 0.1827 0.1947 0.1893 0.1780 0.1946 0.1812
LEATHER 0.0471 0.0475 0.0466 0.0480 0.0469 0.0476 0.0508 0.0493 0.0464 0.0509 0.0472
MACHINERY 0.0031 0.0031 0.0031 0.0031 0.0031 0.0031 0.0033 0.0032 0.0031 0.0033 0.0031
MEDICINES 0.0409 0.0411 0.0407 0.0416 0.0408 0.0413 0.0442 0.0429 0.0404 0.0444 0.0409
METAL 0.0000 0.0263 0.0259 0.0266 0.0261 0.0264 0.0282 0.0274 0.0258 0.0283 0.0262
NONFERROUS 0.0346 0.0000 0.0344 0.0352 0.0345 0.0349 0.0374 0.0363 0.0342 0.0376 0.0346
NONMETALIC 0.0163 0.0164 0.0000 0.0165 0.0162 0.0165 0.0175 0.0170 0.0160 0.0174 0.0163
PAPER 0.0441 0.0445 0.0437 0.0000 0.0440 0.0446 0.0476 0.0463 0.0436 0.0477 0.0442
PETROLEUM 0.0222 0.0224 0.0220 0.0226 0.0000 0.0224 0.0239 0.0233 0.0219 0.0239 0.0223
PLASTICS 0.0369 0.0372 0.0367 0.0376 0.0368 0.0000 0.0399 0.0388 0.0365 0.0400 0.0370
PRINTING 0.0993 0.1002 0.0981 0.1011 0.0988 0.1004 0.0000 0.1040 0.0978 0.1070 0.0995
RUBBER 0.0723 0.0730 0.0716 0.0737 0.0721 0.0732 0.0780 0.0000 0.0713 0.0781 0.0725
TEXTILE 0.0123 0.0125 0.0122 0.0126 0.0123 0.0125 0.0133 0.0129 0.0000 0.0133 0.0124
TIMBER 0.1014 0.1025 0.0995 0.1031 0.1005 0.1025 0.1090 0.1060 0.0995 0.0000 0.1018
WEARING 0.0286 0.0288 0.0284 0.0291 0.0285 0.0289 0.0309 0.0300 0.0282 0.0310 0.0000

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The impact of trade liberlaization on firm performance in developing countries -- new evidence from Pakistani manufacturing sector

PDF

Essays on family planning policies

PDF

Essays on policies to create jobs, improve health, and reduce corruption

PDF

Two essays on financial crisis and banking

PDF

Essays on congestion, agglomeration, and urban spatial structure

PDF

A spatially explicit approach to measuring carbon dynamics for reducing emissions from deforestation and forest degradation: a case study of Chinese forests

PDF

Essays on effects of reputation and judicial transparency

PDF

Essays on the econometrics of program evaluation

PDF

An empirical characterisation of the Bombay Stock Exchange

PDF

Essays on education programs in Costa Rica

PDF

The decisions of migration and remittances in rural China

PDF

Two essays in econometrics: large N T properties of IV, GMM, MLE and least square model selection/averaging

PDF

Essays in political economy and mechanism design

PDF

The economic and political impacts of U.S. federal carbon emissions trading policy across households, sectors and states

PDF

Individual vs. group behavior: an empirical assessment of time preferences using experimental data

PDF

Three essays on supply chain networks and R&D investments

PDF

An empirical analysis of the quality of primary education across countries and over time

PDF

Network-based simulation of air pollution emissions associated with truck operations

PDF

Panel data forecasting and application to epidemic disease

PDF

Essays on health and aging with focus on the spillover of human capital

Asset Metadata

Creator Wei, Wei (author)

Core Title The impact of agglomeration policy on CO₂ emissions: an empirical study using China’s manufacturing data

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Publication Date 07/22/2015

Defense Date 05/28/2015

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag agglomeration,CO₂ emission,Forecasting,Manufacturing industry,OAI-PMH Harvest,program evaluation

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Hsiao, Cheng (committee chair), Dekle, Robert (committee member), Nugent, Jeffrey B. (committee member), Schiler, Marc (committee member)

Creator Email wei1@usc.edu,weiw.04@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-603078

Unique identifier UC11299537

Identifier etd-WeiWei-3674.pdf (filename),usctheses-c3-603078 (legacy record id)

Legacy Identifier etd-WeiWei-3674.pdf

Dmrecord 603078

Document Type Dissertation

Format application/pdf (imt)

Rights Wei, Wei

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA