Contributions to sector-level carbon intensity change：
an integrated decomposition analysis
Qunwei Wanga, Ye Hangb, Bin Sub,, Peng Zhoua
a College
of Economics and Management, Nanjing University of Aeronautics
and Astronautics
bSchool
cEnergy
of Business, Soochow University
Studies Institute, National University of Singapore
Outline
1. Introduction
2. Methodology
3. Case study
4. Conclusion
2
Introduction
 Background

Environmental-economic balance requirement of tackling climate change

Carbon intensity (i.e. carbon emissions per unit of GDP) is an important
metric for measuring energy and environmental performance

Carbon intensity reduction targets

China: 13th Five Year Plan (2016-2020), reduce carbon intensity by
18% compared to 2015 level (State Council, 2016)

India: Intended Nationally Determined Contribution (INDC), reduce
carbon intensity by 33-35% by 2030 from 2005 level (UNFCCC, 2015)
Driving factors？
3
Introduction
 Decomposition method

Index decomposition analysis (IDA)

Structural decomposition analysis (SDA)

Attribution analysis (AA)
(Ang and Zhang, 2000; Xu and Ang, 2013)
(Rose and Casler, 1996; Su and Ang, 2012)
(Choi and Ang, 2012; Su and Ang, 2014)

Quantify the contributions of the individual components to an aggregate indicator change

AA based on LMDI in IDA; AA based on generalized Fisher index in SDA
The impacts of production technology related components cannot be directly
investigated under the IDA and SDA frameworks


Production-theoretical decomposition analysis (PDA)
(Zhou and Ang, 2008)

Production technology related components (technical efficiency & technological change)

Potential effects deflated by technical efficiency
PDA and IDA combined approach
(Kim and Kim, 2012; Lin and Du, 2014)

Production technology related components & Potential effects

4
Avoid misleading results of PDA in quantifying industrial structure and energy mix effects
Introduction
 Summary of the studies combining PDA and IDA methods
Application area
Study
Indicator
Variable
Effect
SDF
Level
Period
Input
Output
act
cef
√
√
1. Kim and Kim (2012)
C
National
1990-2006
E
E
Y, C
2. Lin and Du (2014)
EI
Regional
2005-2010
Y
E, K, L
Y
3. Du and Lin (2015)
E
Regional
2003-2010
Y
E, K, L
Y
√
4. Du et al. (2017)
C
Regional
2006-2012
E
E
Y, C
√
5. Li et al. (2017)
C
Regional
2001-2011
E, Y
E, K, L
Y, C
6. Wang et al. (2017b)
C
2005-2010
E, C
E, L
7. Zhou et al. (2017)
E
1991-2012/
2001-2012
E, Y
This study (PDA-IDA-AA)
CI
2006-2014
E, Y
National/Regi
onal/Sectoral
National/
Sectoral
Sectoral
pcef
emx
pei
str
e-pr
√
√
√
√
√
√
√
√
√
√
√
√
√
√
√
√
√
Y, C
√
√
E, K, L
Y
√
E
Y, C
√
sub
√
√
√
c-pr
√
√
√
y-pr
ros
pros
gap
√
√
√
√
√
√
√
√
√
√
√
√
√
√
√
√
The existing PDA and IDA combined approach cannot further capture the
contribution of each region to the individual driving factor
Note:
1.
2.
3.
4.
The indicators “C”, “EI”, “E” and “CI” stand for the changes in carbon emissions, energy intensity, energy consumption and carbon intensity, respectively.
The abbreviation “SDF” refers to Shephard distance function.
The variables “E”, “K”, “L”, “Y” and “C” refer to energy, capital, labor, desirable output and undesirable output, respectively.
The abbreviations “act”, “cef”, “pcef”, emx”, “pei”, “str”, “e-pr”, “y-pr”, “c-pr”, “sub”, “ros”, “pros” and “gap” refer to the various effects in the
decomposition identity, i.e. activity, carbon emission coefficient, potential carbon emission coefficient, energy mix, potential energy intensity, economy
structure, energy use productivity (i.e. energy use technical efficiency and energy use technological change), desirable output productivity (i.e. desirable
output technical efficiency and desirable output technological change), carbon emission productivity, substitution, regional output structure, potential
regional output structure and output gap effects, respectively.
5
Introduction
 Objective
Proposing an integrated decomposition approach combining PDA, IDA and AA
to explore the contributions to sector-level carbon intensity change
 Contribution

The sector-level carbon intensity
Developing an integrated decomposition
framework combining PDA, IDA and AA

PDA
IDA
Contribution
Creating two new pre-defined factors, i.e.
Driving factor 1
Driving factor 2 ... Driving factor k
AA
the potential regional output structure
effect and the output gap effect

Contribution
Contribution
Contribution
Region 1
Region 1
Region 1
Region 2
Region 2
Region 2
Focusing on the decomposition of carbon
intensity, which is less well documented
compared to other indicators
Region i
Region i
...
the single sectoral level
...
Obtaining the decomposition results at
...

Region i
6
Outline
1. Introduction
2. Methodology
3. Case study
4. Conclusion
7
Methodology
 Environmental production technology (Zhou and Ang, 2008)
Sn 
 E ,Y
n
S 
 E , Y
n
n
n
n

, C n  : E n can produce (Y n , C n )
: z E
I
,C
n
n
i
i
i 1
I
zY
i 1
n
 Yi'n
n
i
 Cin'
i i
I
zC
i 1
i
 Ein'
zi  0, i  1,..., I 
8
Methodology
 Shephard distance function and estimation models

Shephard input distance function
DEn ( E n , Y n , C n )  sup  : ( E n /  , Y n , C n )  S n 

Shephard desirable output distance function

Estimation models: DEA linear programming
DYn ( E n , Y n , C n )  inf  : ( E n , Y n /  , C n )  S n 
1
 DEm ( E n , Y n , C n ) 
 DEm ( E n , Y n , C n )   min 
I
s.t.
z E
i 1
m
i
i
zY
i 1
m
i i
I
(Zhou and Ang, 2008)
zC
i 1
i
 min 
I
 E
n
i'
s.t.
z E
i 1
m
i
i
I
I
Feasible
1
m
i
 Yi'
n
Infeasible
C
n
i'
z i  0, i  1,..., I
m, n  t , t  1
zY
i 1
m
 Yi'n
m
i
 Cin'   i'
i i
I
(Wang et al., 2015)
zC
i 1
i
  Ein'
z i  0,  i'  0, i  1,..., I
m, n  t , t  1 and m  n
9
Methodology
 Decomposition approach — PDA
Cijt
Eijt
Eit Yi t
CI   t  t  t  t
Ei Yi Y
i 1 j 1 Eij
I
J
t
Emission coefficient, energy mix, energy intensity, regional output structure

I
J
CI t  
i 1 j 1
Eijt
Eit

Yi t  DYt ( E t , Y t , C t )  DYt 1 ( E t , Y t , C t ) 
I
Y
i
i 1

Cijt
 t
1/2
t
t
t
t
t 1
t
t
t
Eij
 DE ( E , Y , C )  DE ( E , Y , C ) 
Y
i 1
i
t
 D ( E , Y , C )  D ( E , Y , C ) 
t
Y
t
t
t
t 1
Y
t
t
t
Yt
 D t 1 ( E t , Y t , C t ) 
 D ( E , Y , C )   Et t t t 
 DE ( E , Y , C ) 
t
E
t
t
1/2
 DYt ( E t , Y t , C t )  DYt 1 ( E t , Y t , C t ) 
Eit Yi t
I

t
1/2
t
 Dt ( E t , Y t , C t ) 
1
 t t t t   tY1 t t t 
DY ( E , Y , C )  DY ( E , Y , C ) 
1/2
1/2
Five categories (nine components)
1/2
1. Structure
Energy mix (EMX)
Potential regional output structure (PIS)
2. Intensity
Potential energy intensity (PEI)
Emission coefficient (CEF)
3. Output gap
Output gap (ISG)
4. Energy use productivity
Energy use technical efficiency (EUE)
Energy use technological change (EST)
5. Desirable output productivity
Desirable output technical efficiency (YOE)
Desirable output technological change (YCT)
10
Methodology
 Decomposition approach — IDA
CI t  EMX t  PIS t  PEI t  CEF t  ISG t  EUE t  EST t  YOE t  YCT t
CI t 1  EMX t 1  PIS t 1  PEI t 1  CEF t 1  ISG t 1  EUE t 1  EST t 1  YOE t 1  YCT t 1

Single-period
CI t 1
t ,t 1
t ,t 1
t ,t 1
t ,t 1
t ,t 1
t ,t 1
t ,t 1
t ,t 1
Dtot 

D

D

D

D

D

D

D









EMX
PIS
PEI
CEF
ISG
EUE
EST
t
CI
Structure
＞1
＝1
＜1
t ,t 1
DPEI
Intensity
 I J S V  PEI it 1  
 exp   wij ln 
t 

P
EI
i

1
j

1
i



Outout gap
S V
ij
w

t ,t 1
YOE
t ,t 1
 DYCT

Desirable output productivity
L  Cijt C t , Cijt 1 C t 1 
 L  C
I
J
i 1 j 1

Energy use productivity
D
t
ij
C t , Cijt C t 1 
Multi-period
T
CI T
CI t 1 T
0,T
t ,t 1
t ,t 1
t ,t 1
t ,t 1
t ,t 1
t ,t 1
t ,t 1
t ,t 1
t ,t 1
Dtot  0   t    DEMX
 DPIS
 DPEI
 DCEF
 DISG
 DEUE
 DEST
 DYOE
 DYCT

CI
t  0 CI
t 0
0,T
0,T
0,T
0,T
0,T
0,T
0,T
0,T
0,T
=DEMX
 DPIS
 DPEI
 DCEF
 DISG
 DEUE
 DEST
 DYOE
 DYCT
T
0,T
PEI
D
t ,t 1
  DPEI
t 0
11
Methodology
 Attribution analysis method (Choi and Ang, 2012)

Single-period

 PEIit 1 
t ,t 1
DPEI  1   rij 
 1
t
PEI
i 1 j 1
i


I
Multi-period
I
J
0,T
PEI
D
rij 
I
J
w
 L  PEI
i 1 j 1
S V
ij
t 1
i
t
i
t
i
PEI
t ,t 1
, PEI DPEI

T
 1   D r
i 1 j 1 t  0
wijS V PEI it
t ,t 1
L  PEI it 1 , PEI it DPEI

J
0,t t ,t 1
PEI ij
 PEI it 1 
 1

t
 PEIi

wijS V ,t ,t 1 PEI it
t ,t 1
ij
r
t ,t 1
L  PEI it 1 , PEI it DPEI


wijS V ,t ,t 1 PEI i ,t
 L  PEI
i
j
t 1
i
t ,t 1
, PEI it DPEI

12
Outline
1. Introduction
2. Methodology
3. Case study
4. Conclusion
13
Case study
 Data

Industrial sector across 30 provinces in China (2006-2014)

China Energy Statistical Yearbook, 2007-2015

China Statistical Yearbook, 2007-2015

2006 IPCC Guidelines for National Greenhouse Gas Inventories
 Variable

Energy consumption (E)

Industrial output (Y)

Carbon emissions (C)
14
Case study
 Results and discussion — Decomposition results

Cumulative effects on industrial carbon intensity change by five categories, 2006-2014
Dtot = 47.63%
52.18%
47.55%
15
Case study
 Results and discussion — Decomposition results

Changes in industrial carbon intensity and its decomposition, 2006-2014
Aggregate
Structure
Intensity
Period
Output
Energy use
Desirable output
gap
productivity
productivity
Dtot
DEMX
DPIS
DPEI
DCEF
DISG
DEUE
DEST
DYOE
DYCT
2006-2007
0.8869
1.0126
1.0565
0.8987
0.9898
1.0995
1.0173
0.9518
0.9574
0.9145
2007-2008
0.8153
0.9929
1.1189
0.8247
1.0008
1.1295
0.8713
1.1170
0.8546
0.9466
2008-2009
1.0503
0.9938
0.9986
1.0442
0.9838
0.9256
1.0176
1.0101
0.9956
1.0877
2009-2010
0.8228
1.0169
1.0191
0.8417
0.9632
1.2388
1.0753
0.9180
1.0003
0.8006
2010-2011
0.8394
1.0087
1.0275
0.8698
1.0096
1.2223
1.0032
0.9359
1.0291
0.7809
2011-2012
0.9685
1.0059
0.9724
0.9836
0.9845
1.0882
1.0210
0.9701
1.0724
0.8846
2012-2013
1.0318
1.0277
0.9978
1.0270
0.9829
0.9895
0.9795
1.0214
1.2669
0.7949
2013-2014
0.9990
1.0048
0.9657
1.0108
0.9886
1.0102
1.0073
0.9983
1.0514
0.9647
g-mean
0.9223
1.0079
1.0185
0.9339
0.9878
1.0830
0.9975
0.9886
1.0228
0.8916
2006-2014
0.5237
1.0646
1.1582
0.5785
0.9066
1.8931
0.9801
0.9126
1.1978
0.3993
47.63%
42.15%
89.31%
60.07%
16
Case study
 Results and discussion — Decomposition results

Cumulative changes in industrial carbon intensity and its decomposition, 2006-2014
DPEI
DYCT
17
Case study
 Results and discussion — Attribution results

Multi-period attribution results of DYCT and DPEI, 2006-2014 (base: 2006, %)
Province
DYCT
DPEI
Province
DYCT
DPEI
Beijing
-0.50
-0.43
Hubei
-2.59
-2.61
Tianjin
-1.08
-0.27
Hunan
-1.93
-2.97
Hebei
-5.72
-1.49
Guangdong
-2.81
-0.76
Shanxi
-2.67
-1.37
Guangxi
-1.41
-1.57
Inner Mongolia
-3.17
-2.83
Hainan
-0.19
-0.00
Liaoning
-3.69
-3.71
Chongqing
-1.28
-0.96
Jilin
-1.47
-1.94
Sichuan
-2.74
-2.32
Heilongjiang
-1.15
-0.76
Guizhou
-1.08
-1.82
Shanghai
-1.17
0.09
Yunnan
-1.31
-1.04
Jiangsu
-4.47
-1.54
Shaanxi
-1.41
-0.79
Zhejiang
-1.93
-1.51
Gansu
-0.78
-0.35
Anhui
-1.72
-1.75
Qinghai
0.64
-0.30
Fujian
-1.47
-0.97
Ningxia
-1.66
-0.55
Jiangxi
-1.13
-0.71
Xinjiang
-1.36
0.18
Shandong
-5.39
-3.18
Total change
-60.07
-42.15
Henan
-3.41
-3.90
18
Case study
 Results and discussion — Attribution results

Single-period attribution results of DYCT, 2006-2014 (base: previous year, %)
Province
Beijing
Tianjin
Hebei
Shanxi
Inner Mongolia
Liaoning
Jilin
Heilongjiang
Shanghai
Jiangsu
Zhejiang
Anhui
Fujian
Jiangxi
Shandong
Henan
Hubei
Hunan
Guangdong
Guangxi
Hainan
Chongqing
Sichuan
Guizhou
Yunnan
Shaanxi
Gansu
Qinghai
Ningxia
Xinjiang
Total change
2007
-0.09
-0.10
-0.53
-0.31
-0.45
-0.53
-0.21
-0.24
-0.18
-0.74
-0.61
-0.20
-0.23
-0.14
-0.79
-0.55
-0.28
-0.26
-0.66
-0.23
-0.01
-0.10
-0.43
-0.23
-0.25
-0.12
-0.11
0.40
-0.21
-0.12
-8.55
2008
-0.09
-0.12
-0.65
-0.37
1.63
-0.54
-0.24
-0.20
-0.18
-0.68
1.31
-0.24
-0.25
-0.18
-0.83
-0.62
-0.34
-0.33
-0.79
-0.22
-0.02
-0.14
-0.41
-0.21
-0.26
-0.16
-0.13
-0.30
0.38
-0.15
-5.34
2009
0.08
0.14
0.69
0.38
0.45
0.58
0.21
0.20
0.19
0.70
0.18
0.25
0.27
0.20
0.88
0.60
0.36
0.37
0.91
0.21
0.02
0.19
0.35
0.20
0.27
0.17
0.14
0.20
-0.81
0.17
8.77
2010
-0.18
-0.33
-1.69
-0.86
-1.29
-1.31
-0.41
-0.33
-0.35
-1.31
-1.72
-0.53
-0.45
-0.34
-1.62
-1.05
-0.73
-0.64
-1.01
-0.41
-0.05
-0.36
-0.75
-0.36
-0.39
-0.40
-0.24
0.38
-0.78
-0.42
-19.94
2011
-0.16
-0.42
-1.84
-0.88
-1.08
-1.18
-0.60
-0.39
-0.41
-1.76
-0.82
-0.67
-0.59
-0.43
-2.10
-1.32
-1.14
-0.79
-1.06
-0.51
-0.06
-0.49
-1.11
-0.33
-0.51
-0.50
-0.28
0.31
-0.36
-0.46
-21.91
2012
-0.07
-0.21
-1.29
-0.63
-1.37
-0.76
-0.19
-0.13
-0.24
-0.56
-0.19
-0.30
-0.19
-0.21
-0.75
-0.41
-0.44
-0.27
-0.16
-0.17
-0.05
-0.32
-0.37
-0.11
-0.12
-0.35
-0.09
-1.08
-0.18
-0.31
-11.54
2013
-0.09
-0.33
-1.57
-0.81
-3.29
-0.94
-0.46
-0.37
-0.32
-1.41
-0.66
-0.57
-0.46
-0.38
-1.75
-0.99
-0.85
-0.57
-0.83
-0.50
-0.06
-0.55
-0.54
-0.32
-0.44
-0.41
-0.33
-0.16
0.02
-0.54
-20.51
2014
-0.04
-0.16
-1.52
-0.20
-0.11
-0.25
-0.06
-0.07
-0.10
-0.26
0.02
-0.11
-0.07
-0.11
-0.35
-0.15
-0.13
-0.10
0.16
-0.10
-0.05
-0.06
-0.46
-0.01
0.00
-0.28
-0.04
1.16
0.03
-0.10
-3.53
Mean
-0.08
-0.19
-1.05
-0.46
-0.69
-0.62
-0.25
-0.19
-0.20
-0.75
-0.31
-0.30
-0.25
-0.20
-0.91
-0.56
-0.44
-0.32
-0.43
-0.24
-0.04
-0.23
-0.47
-0.17
-0.21
-0.26
-0.13
0.11
-0.24
-0.24
-10.32
19
Case study
 Results and discussion — Attribution results

Single-period attribution results of DPEI, 2006-2014 (base: previous year, %)
Province
Beijing
Tianjin
Hebei
Shanxi
Inner Mongolia
Liaoning
Jilin
Heilongjiang
Shanghai
Jiangsu
Zhejiang
Anhui
Fujian
Jiangxi
Shandong
Henan
Hubei
Hunan
Guangdong
Guangxi
Hainan
Chongqing
Sichuan
Guizhou
Yunnan
Shaanxi
Gansu
Qinghai
Ningxia
Xinjiang
Total change
2007
-0.12
-0.07
-0.42
-0.71
-0.79
-0.89
-0.46
-0.22
0.00
-0.42
-0.36
-0.42
-0.19
-0.02
-0.77
-0.75
-0.32
-0.52
-0.20
-0.39
-0.01
-0.21
-0.52
-0.34
-0.27
-0.11
-0.17
-0.14
-0.24
-0.09
-10.13
2008
-0.22
-0.14
-2.05
-1.32
-0.88
-1.77
-0.48
-0.49
0.01
-0.58
-0.57
-0.60
-0.16
-0.33
-1.41
-1.54
-0.78
-0.51
-0.28
-0.49
-0.02
0.14
-0.64
-0.88
-0.32
-0.25
-0.16
-0.16
-0.26
-0.41
-17.53
2009
-0.02
0.20
1.25
0.49
-0.32
0.80
-0.22
0.59
0.03
0.03
0.20
-0.03
-0.03
0.11
0.30
0.53
-0.40
-0.18
0.06
0.14
0.04
-0.47
-0.21
0.27
0.40
-0.02
0.08
0.04
0.10
0.64
4.42
2010
-0.04
-0.17
-0.82
-0.98
-1.05
-1.36
-0.46
-0.63
-0.01
-0.66
-0.55
-0.71
-0.41
-0.83
-0.80
-1.02
-0.23
-1.09
-0.25
-0.59
-0.03
-0.43
-0.54
-0.22
-0.66
-0.24
-0.09
-0.14
-0.20
-0.62
-15.83
2011
-0.05
-0.09
-1.28
-0.67
-0.21
-1.67
-0.21
-0.40
-0.10
-1.09
-0.50
-0.48
-0.24
-0.12
-0.89
-1.30
-0.50
-0.54
0.08
-0.46
0.01
-0.18
-0.84
-0.23
0.01
-0.25
-0.38
-0.14
-0.10
-0.22
-13.02
2012
-0.01
-0.10
-0.09
0.22
-0.53
-0.35
-0.03
0.09
-0.02
-0.11
0.10
-0.05
-0.04
-0.01
-0.14
-0.66
-0.32
-0.46
-0.13
-0.03
-0.02
1.86
-1.22
-0.18
-0.13
-0.10
0.24
0.10
-0.06
0.54
-1.64
2013
-0.03
0.08
1.97
1.62
0.07
1.04
-0.51
0.07
0.32
1.03
-0.08
0.19
-0.27
0.45
-0.33
-0.03
-0.95
-0.63
-0.18
-0.04
0.05
-2.17
1.31
-0.63
-0.51
-0.04
0.13
0.11
0.08
0.59
2.70
2014
-0.01
-0.05
0.47
0.58
0.34
-0.11
-0.15
0.20
-0.07
0.16
-0.09
0.03
0.05
0.00
0.15
-0.22
-0.15
-0.11
-0.06
-0.09
-0.01
0.13
-0.24
-0.22
0.14
-0.04
0.02
0.05
0.10
0.26
1.08
Mean
-0.06
-0.04
-0.12
-0.10
-0.42
-0.54
-0.32
-0.10
0.02
-0.21
-0.23
-0.26
-0.16
-0.09
-0.48
-0.62
-0.46
-0.50
-0.12
-0.24
0.00
-0.17
-0.36
-0.30
-0.17
-0.13
-0.04
-0.03
-0.07
0.09
-6.24
20
Case study
 Results and discussion — Attribution results

Percentage share of each province in the multi-period attribution results of DYCT
and DPEI, 2006-2014 (%)
high
high
low
high
21
Case study
 Results and discussion — Attribution results

Classification of provinces based on percentage share of each province in the
multi-period attribution results of DYCT and DPEI, 2006-2014

Type Ⅰ (8 provinces)

Type Ⅱ (2 provinces)

Type Ⅲ (14 provinces)

Type Ⅳ (6 provinces)
22
Outline
1. Introduction
2. Methodology
3. Case study
4. Conclusion
23
Conclusion
 Contribution

The integrated decomposition approach extends the existing PDA and IDA
combined approach by quantifying the contribution of each region
attributes to the individual driving factor using the AA method. This helps
set carbon intensity reduction policies targeting a specific factor at the
regional level.

The decomposition model adds two new effects, i.e. the potential regional
output structure effect and the output gap effect. This may provide insights
to different aspects of carbon intensity reduction policymaking.

This study focuses on decomposing sector-level carbon intensity change,
which is relatively less well documented and understood compared to
other indicators/levels in the existing PDA and IDA combined studies.
24
Conclusion
 Case study

Of the five decomposition categories, desirable output productivity and
intensity are the leaders in promoting the reduction of industrial carbon
intensity.

Of the nine decomposition components, desirable output technological
change and potential energy intensity factors play dominant roles in
decreasing the industrial carbon intensity.

The main contributors to the negative value of the desirable output
technological change effect are Hebei, Shandong, Jiangsu, Liaoning and
Henan. The top five provinces contributing to the negative value of the
potential energy intensity effect are Henan, Liaoning, Shandong, Henan
and Inner Mongolia.

Provinces were divided into four types based on the multi-period
attribution results; different industrial carbon intensity reduction policies
should be implemented in different types of provinces.
25
Conclusion- Extension
 Application

Sector-level

Carbon intensity
Economy-wide
Other absolute and intensity indicators, e.g. energy,
water, air emissions and waste
 Methodology

IDA framework

Temporal PDA-IDA/SDA-AA
SDA framework
Spatial (Ang et al., 2015; Su and Ang, 2016 )
Spatial-Temporal (Ang et al., 2016)
26
Thank you!
Qunwei Wang ([email protected])
Bin Su ([email protected] ; [email protected])
27