In the format provided by the authors and unedited.
Suppplementary Material for:
Unequal household carbon footprints in China
Dominik Wiedenhofer
a,
*, Dabo Guan
b,
*, Zhu Liu
c
, Jing Meng
d
, Ning Zhang
e
, Yi-
Ming
We
i
f,
*
a.
Institute of Social
Ecology (SEC), IFF
- Vienna, Alpen
-Adria University Klagenfurt, Austria
b.
School of International Development, University of East Anglia, Norwich, NR4 7JT
, UK
;
c.
Kennedy School of Government, Harvard University, Cambridge, MA 02138, US.
d.
School of Environmental
Sciences, University of East Anglia, Norwich, NR4 7JT
, UK
;
e.
College of Economics, Jinan University, Guangzhou,
510632
, China.
f.
Center for Energy and Environmental Policy Research, Beijing Institute of Technology, Beijing 100081,
China
.
* Corresponding email:
Dominik.wiedenhofer@aau.at
; dabo.guan@uea.ac.uk
; wei@bit.edu.cn
Keywords: climate justice; carbon footprint; inequality; s
ustainable consumption; multi
-regional input
-output
analysis;
Figure
S1: Detailed c
arbon footprints of Chinese household consumption
per capita,
for 13 income
groups in 2012
UNEQUAL HOUSEHOLD CARBON FOOTPRINTS IN CHINA
Unequal household carbon footprints in China
©
2016 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
SUPPLEMENTARY INFORMATION
DOI: 10.1038/NCLIMATE3165
NATURE CLIMATE CHANGE
|
www.nature.com/natureclimatechange
1
Figure
S
2
: Absolute carbon
footprints of Chinese and international household consumption in
2012/2011
Table
S
1
: Consumption
-
Based Carbon
-
Gini coefficients of Chinese household consumption across 13 income groups
Housing
Mobility
Food
Goods
Services
T
o
tal CF
Exp
enditure
National
2012
0.35
0.53
0.28
0.44
0.46
0.39
0.41
2007
0.38
0.56
0.32
0.50
0.51
0.43
0.45
Urban
2012
0.32
0.48
0.20
0.35
0.33
0.33
0.33
2007
0.31
0.49
0.20
0.36
0.34
0.33
0.32
Rural
2012
0.23
0.29
0.20
0.25
0.24
0.23
0.23
2007
0.22
0.31
0.17
0.27
0.27
0.23
0.23
Figure
S
3
: Carbon footprint elasticities for 2007 and 2012.
CF elasticities were calculated using the basic income elasticity approach, where
the relative change of each income groups’
CF/cap from the average CF/cap is divided by the relative change of each income groups’
expenditure/cap from the average exp/cap in 2012
Table
S
2
: Main results for Chinese household carbon footprints i
n 2012 and 2007
.
CF elasticities were calculated using the basic income elasticity approach, where the relative change of each income groups’
CF/cap from
the average CF/cap is divided by the relative change of each income groups’ expenditure/cap from the a
verage exp/cap in 2012.
2012
2007
CF
cap
(tCO2)
Total CF
(Mt CO2)
Expenditure
($2012MER)
CF
elasticity of
expenditure
Population
(Million people)
CF
cap
(tCO2)
Total CF
(Mt CO2)
Expenditure
($2007MER)
CF
elasticity of
expenditure
Population
(Million
people)
China, total
1.72
2,332
1,908
1.00
1,354
1.48
1,954
951
1.00
1,321
Urban, total
2.44
1,738
2,803
0.97
712
2.41
1,429
1,583
0.98
594
Rural, total
0.93
594
916
1.12
642
0.72
525
435
1.01
728
Urban China
Very rich
6.39
455
7,237
0.98
71
6.29
374
4,026
0.98
59
Rich
3.73
266
4,298
0.97
71
3.70
220
2,439
0.97
59
Middle
-
high
2.75
392
3,159
0.97
142
2.71
322
1,814
0.97
119
Middle
2.00
285
2,334
0.95
142
1.99
236
1,330
0.97
119
Lower
-
middle
1.49
212
1,725
0.96
142
1.47
175
978
0.99
119
Poor
1.12
80
1,270
0.98
71
1.09
65
710
1.00
59
Very poor
0.75
27
838
1.00
36
0.71
21
460
1.00
30
Extremely
poor
0.58
21
650
1.00
36
0.57
17
367
1.05
30
Rural China
Highest
1.64
210
1,611
1.13
128
1.27
185
780
1.05
146
middle
-
high
1.07
138
1,054
1.13
128
0.80
117
479
1.08
146
middle
0.79
102
785
1.13
128
0.63
92
378
1.08
146
Poor
0.62
80
625
1.10
128
0.50
73
302
1.07
146
Extremely
poor
0.51
65
506
1.11
128
0.40
58
236
1.09
146
(1)
Me
thod
and Data
I
n this study environmentally extended input
-
output analysis (EE
-
IO)
was used
to estimate
emissions, energy or resource use
linked with the
production and
supply
of
final demand
and
household consumption.
T
he IO approach conceptually is similar to the consumer lifestyle
approach
1,2
, insofar as all direct and indirect emissions associated with a specific lifestyle or
consumption pattern are quantified
3
. The strength of the IO method lies in the complete and
systematic coverage of
the
entire
upstream supply chain
and especially of all indirect linkages
between industrial sectors.
For details of the method
s
and extensive mathematical treatments
we have to refer elsewhere
4
–
6
and only summarize and foc
u
s on the specifics
of this study
.
Firstly,
data
on
Chinese h
ousehold expenditure from 2007
and 2012
by income groups
for
the
entire
population were used to discern the
consumption
patterns of 5 rural and 8 urban income
groups
.
In contrast to most other s
tudies
which aggregated the IO tables to fit the expenditure
data or time series of IO tables,
we
mapped this expenditure data to the
detailed
national
input
-
ou
t
p
ut table
s
for 2007
and 2012
, in order
to preserve as much resolution as possible,
which has be
en shown to
substantially
improve accuracy of results
7
–
9
.
This data until 2007
has been used in the literature before to estimate the development and trajectories of Chinese
household footprints
10
–
13
, investigations of urban vs. rural household carbon footprints
2
, over
time
14
,
15
or for specific cities
16
. For this study we disaggregated household consumption into
13 different income groups (8 urban and 5 rural), using the
China Urban Life and Price
Yearbook
17
, which reports 8 urban and 5 rural income groups. The Yearbooks list average
incomes and consumption expenditur
e patterns for each group, yielding the sum of total
household final use reported in the Chinese IOT. The data discerns 8 major classes of
expenditure items and 58 sector specific items, which is different than the 135 sectors of the
IOT. We then disaggreg
ated the 58 specific items unto the 135 sectors based on their
according products.
Direct energy use
and emissions
of the 13 income groups
were
estimated via urban and rural
energy prices derived from the Input
-
Output Tables and the separate reporting of
t
otal
urban
and rural
household
energy consumption
in the Chinese energy statistics
.
Because
Chinese industry and consumers
indirectly and directly
demand imports we
furthermore used
a
so
-
called
multi
-
regional input
-
output model
18
–
20
derived from the most
recent GTAP database
21,22
(v8 for 2007
and v9 f
or 201
1
)
. This model
cover
s
the
majority of
the
world economy
including all international supply chain linkages.
With this we go beyond
existing studies on Chinese household footprints, which
rely
on
the so
-
called domestic
technology assumption
to approxim
ate emissions from imports
2,14,15,23
.
The total carbon footprint of Chinese household consumption
q
hh
then consists of
the
domestic
indirect
emissions
q
d
om
,
the international upstream emissions
q
mrio
and the
emissions
from direct
energy use of households
q
direct
.
푞
ℎℎ
=
q
direct
+
푞
푑표푚
+
푞
푚푟푖표
Eq.1.
(2)
Estimating
the
domestic
direct and indirect
carbon footprints
of
Chinese household consumption
퐪
퐝퐢퐫퐞퐜퐭
풂풏풅
풒
풅풐풎
The
latest
input
-
output data
for 2012 and 2007 is available
from the Chinese National Bureau
of Statistics
24
.
The input
-
output tables were used in full detail of 135 sectors, going beyond
existing studies
which use more aggregated IO tables (8
-
40 sectors) due to for example a
focus on time series analysis and the constraint of backwards compatibility of input
-
output
tables
2,14,15,25,26
.
CO2 from fossil fuels energy use data has been compiled
from the latest
revised data
27
, which is up to 10% lower than the official
Chin
ese
Energy Statistical
Yearbook
28
. This data
covers 18 types of fuel, heat and electricity consumption in physical
units, at
a 4
5
sectors disaggregation. We used
a
mapping of
the
emissions dataset to the 135
input
-
output sectoral resolution developed in
previous work
10
.
Transformation losses were
allocated to the respective energy user, for example losses in electricity generation to the
electricity sector
and therefore indirectly to the respective household
s
consuming
electricity.
The Chinese input
-
output tables contain two idiosyncrasies
: Firstly, the tables
contains a
vector “others”, which is part of
C
hinese total output
x
. On an aggregate level this “others”
item is <0.2% of GDP, but on a sectoral level
it can range from
-
4.5% to +8% of total output.
According to the Chinese National Bureau of Statistics
24
it primarily represents differences in
reported data, particularly due to trade. Therefore,
similarly as Minx et al.
10
, we treat this item
as an error term, representing differences in data sources. We exclude this term from further
analysis and treat total output
x
as the sum of
the matrix of industrial
intermediate use
Z
and
final demand
y
, minus “others”.
This new total output
x
is
then used to arrive at the
national
inter
-
industry direct requirements matrix
A
, which represents the production recipe of the
Chinese economy, where ^ represents diagonalization of a vector.
퐴
=
푍
∗
푥
̂
−
1
Eq.1.
Secondly
Chinese IO tables are compiled based on a non
-
competitive imports assumption
29
,
which means that
imports
m
are reported as an aggregate
column
of
total
imports per sector
i
,
with no additional information available to
distinguish
the proportions for intermediate use
and final demand, nor the import
structure of
the 135 sectors. This is not a new problem in
input
-
output analysis and following
previous work
6,10,30
we calculate so
-
called importshares
s
for each sector
i
, where
ex
are all exports of each sector
i
(
Eq. 2
)
.
푠
푖
=
(
푚
푖
)
/
(
푥
푖
+
푚
푖
−
푒푥
푖
)
Eq.2.
These importshares
퐬
퐢
allow
us to separate
the domestic production technology
A
d
from the
direct requirements matrix
A
.
퐴
=
퐴
푑표푚
+
퐴
푚
Eq.3.
퐴
푑표푚
=
푑푖푎푔
(
푠
푖
)
∗
퐴
Eq.4.
The same procedures apply for the separation of
household
final demand
y
_hh
into a
domestically supplied
y
_hh
d
om
and directly imported fraction
y
_hh
m
:
푦
_
ℎℎ
=
푦
_
ℎℎ
푑표푚
+
푦
_
ℎℎ
푚
Eq.6.
푦
_
ℎℎ
푑표푚
=
푑푖푎푔
(
푠
푖
)
∗
푦
_
ℎℎ
Eq.7.
푦
_
ℎℎ
푚
=
(
1
−
푑푖푎푔
(
푠
푖
)
)
∗
푦
_
ℎℎ
Eq.8.
Now the Leontief inverse
L
can be calculated, which represents the direct and indirect
economic activity required to supply a given quantity of final demand for each sector
i
.
퐿
=
(
퐼
−
퐴
)
−
1
Eq.9.
Similarly
the total domestic production multipliers
L
d
are derived.
퐿
푑표푚
=
(
퐼
−
퐿
푑표푚
)
−
1
Eq.10.
As a next step we
normalize the data on total emissions
T
per sector,
by aggre
gat
ing
the new
total output
x
to the sectoral classification used in
the energy and emissions data and then
derived emissions intensities per sector
F,
where
P
represents the mapping of the emissions
data classification
back
to the 135 input
-
output sectors.
This mapping and procedure has been
developed in a previous study o
f the authors
10
.
This procedure rests on the assumption that
sectors in the 135 classificatio
n which are lumped together in the emissions data classification
can be assigned the same emissions multiplier (CO2 / $).
퐹
=
푇
∗
(
푃
∗
푥
̂
)
−
1
∗
푃
Eq.12.
Following standard procedures we
then
apply the Leontief inverse, yielding the usual input
-
output
identity which allows us to estimate the domestic emissions
q
d
om
, which were required
to satisfy a specific level of final demand for domestic production
y
_hh
d
om
.
푞
푑표푚
=
퐹
∗
(
퐼
−
퐴
푑
)
−
1
∗
푦
_
ℎℎ
푑표푚
Eq.13.
Emissions from direct energy use
q
direct
, including electricity, gasoline, natural gas and coal
have been allocated based on the
rural and urban energy prices in the Chinese input
-
output
table for deliveries to households of the sectors ‘coal mining and processing’, ‘gas production
and supply’,
‘electricity generation’ and ‘petroleum and nuclear fuel refining’ of each income
group.
(3)
Estimating international
indirect
carbon
footprints
of Chinese
household consumption
풒
풎풓풊풐
To account for the embodied
CO
2
emissions
in imports
q
m
rio
ether
the
so
-
called
domestic
technology assumption
is used
in existing studies on China
,
which
simplifies by
assuming that
imports
were
produced
similarly as domestic output
or the embodied carbon in
imports are
not estimated
2,14,15,25
.
This is often done due to the very large data requirements of a
more
exact
multi
-
regional approach
, but misses important international inter
-
industry feedbacks and
emission transfers
20,31
.
We use the most recent version of the GTAP database as basis for
such an
MRIO, the
c
onstruction and compilation
of which
has been described
before
19,32
. In total
the GTAP
model
covers
57
sectors for 129 regions
in 2007
and 140 regions in 2011
. For the
international emissions data we
include
CO2 emissions from fossil fuels combustion
, cement
production and gas flaring
by sector
, corrected for the latest revisions of Chinese emissions
27
.
Following standard input
-
output procedures the international production structure and carbon
footprints
q
mrio
can be calculated via the direct and indirect economic activity required to
satisfy
final demand
y
mrio
.
푞
푚푟푖표
=
퐹
푚푟푖표
(
퐼
−
퐴
푚푟푖표
)
−
1
∗
푦
푚푟푖표
Eq.15.
The
international
carbon footprint
of
Chinese households
q
m
rio
consists of
three
components:
Firstly
the
Chinese and
international emissions embodied in imports which are
directly
consumed by households
q
m
rio
,
y
.
Secondly
,
the emissions embodied in imports
which are
used
as intermediate inputs in
to
Chinese domestic production
of goods and services
ultimately
consumed by
Chinese household demand
q
m
rio
,
Z
.
And thirdly, all emissions from China, going
into internationa
l intermediate use, being re
-
imported as Chinese intermediate use and then
ending up in Chinese domestic final demand. This third
, probably very small
component
cannot be fully captured with our combination of the national IOT and the MRIO at the
moment
, w
ithout double
-
counting issues
. Therefore we define the international carbon
footprint of Chinese households as (Eq 16):
푞
푚푟푖표
=
푞
푚푟푖표
,
푦
+
푞
푚푟푖표
,
푍
Eq.16.
Because t
he GTAP
-
MRIO
covers all
origin
s
and destinations
of deliveries,
a
vector o
f total
Chinese house
h
old
final
demand
푦
푚푟푖표
=
퐶
ℎ
푖푛푎
can
be extracted
in
which domestic
ally
supplied
household
final demand is discerned from international
ly supplied Chinese household demand.
Because the more detailed national IOT was used to estimate the domestic portion of the
entire
carbon footprint (see section above), we are only interested in the international upstream
emissions
of domestic production
.
Therefore
a two
-
step procedure had t
o be applied. Firstly,
we
remove Chinese
territorial
emissions from the GTAP emissions inventory
퐹
푚푟푖표
,
퐶
ℎ
푖푛푎
=
0
.
By
setting all international
direct
deliveries to Chinese households to zero
a vector is derived
푦
푀푅퐼푂
=
퐶
ℎ
푖푛푎
,
푑표푚푒푠푡푖푐
which only contains the domestically delivered final consumption of
Chinese households
, but still covering all international supply chains
.
This vector is
then
multiplied by the
adjusted
emissions
-
extended MRIO Leontief inverse
, arriving at
푞
푚푟푖표
,
푍
w
hich covers the international upstream emissions of domestic production for domestic
consumption
.
푞
푚푟푖표
,
푍
=
퐹
푚푟푖표
,
퐶
ℎ
푖푛푎
=
0
(
퐼
−
퐴
푚푟푖표
)
−
1
∗
푦
푀푅퐼푂
=
퐶
ℎ
푖푛푎
,
푑표푚푒푠푡푖푐
Eq. 17.
For the
second
part we define a vector
푦
퐶
ℎ
푖푛푎
,
푖푛푡푒푟푛푎푡푖표푛푎푙
where all domestically supplied
Chinese
monetary
final household demand is set to zero, leaving all international imports in
place.
Because this time we are also interested in the Chinese inter
-
industry deliveries
and
s
ubsequently all embodied emissions
from China
to international
intermediate demand
, we
multiply by
the
full emissions
-
extended MRIO Leontief inverse,
arriving
at
푞
푚푟푖표
,
푦
.
푞
푚푟푖표
,
푦
=
퐹
푚푟푖표
(
퐼
−
퐴
푚푟푖표
)
−
1
∗
푦
푀푅퐼푂
=
퐶
ℎ
푖푛푎
,
푖푛푡푒푟푛푎
푡
푖표푛푎푙
Eq. 18.
This term now contains the emissions embodied in those imports which are directly consumed
by Chinese households and which accrued during international and Chinese inter
-
industry
intermediate deliveries.
Adding both
term
s to
푞
푚푟
푖표
(Eq. 16) now covers the total international carbon footprint of all
Chinese households, which we still need to allocate to the respective income groups. For this
purpose we utilise the information on the amounts of imported final demand
푦
_
ℎℎ
푚
derived
from the national Chinese IOT in Eq. 8
–
10. This vector
푦
_
ℎℎ
푚
can now be calculated for all
13 income groups
n
(Eq. 10) and after conversion from the original 135 sectors to the 57
GTAP
-
MRIO sectors
, using
two
concordance
s
P_iot_gtap
(see next s
ection),
we arrive at a
matrix
푦
푚푟푖표
,
푚
,
푛
of imported
m
final demand
y
by income group
n
in the 57 MRIO
classification.
Then the
shares
S
of each income group
n
in total imports
m
for each of the 57
sectors are calculated (Eq.19) and by multiplying e
ach income group’s share with
푞
푚푟푖표
, a
complete allocation of international carbon footprints to the respective income group
is
achieved.
푆
푛
,
푚
,
푚푟푖표
=
푦
푚푟푖표
,
푚
,
푛
∑
푦
푚푟푖표
,
푚
,
푛
푛
∗
P
_
iot
_
gtap
Eq.19.
(4)
Translating between the n
ational Chinese
IOT
and the global MRIO:
concordances,
classifications and sectoral splits
To
translate information from the
Chinese IOT
with 135 sectors to
the
GTAP
-
MRIO
with 57
sectors for 129 and 140 countries,
it was necessary to
reconstruct two conco
rdances,
follow
ing
the procedures
by
Prof. Liu Yu
22
, who contributes
the C
hinese
IOTs
to
the GTAP
project
(Table 1
-
3).
T
h
e first
concordance translates the 135 Chinese sectors into
45 sectors
22
(Table 1). The second
c
ontains the mapping of these 45 sectors to the 57 GTAP sectors
22
(Table 2). Several of the GTAP sectors are more disaggregated the
n the original Chinese IOT
data
, such as crops (1 vs 8 groups), livestock (1 vs 4 groups) and meat products (1 vs 2).
The
GTAP team used s
plitting procedures
ba
sed on detailed trade data,
for which we refer to the
documentation of the GTAP database
21
. In order to retrieve these splits ex
-
post, shares in
the
total output of these sectors
in the GTAP
model
were used.
The concordances are available
from the
first
author upon request.
(5)
Limitations of Method and Data
The input
-
output method in general is based on several important assumptions: Firstly, the
constancy of prices in the sectoral and demand
relationships is a generally applied and
accepted part of the methodology
33
.
Secondly, input
-
output analysis makes a so
-
called
proportionality a
ssumption, asserting that changes in final demand are met by proportional
adjustments of total industrial output, based on empirical sectoral interdependencies estimated
from the input
-
output data. Thirdly, when extending the IO model environmentally, for
example with
CO
2
emissions per sector and final demand, both assumptions also apply. This
means that the amount of
CO
2
is treated as being strictly proportional to the (sectoral or total)
outputs and price structures along the production recipes and theref
ore any changes or
fractions of final demand translate into proportional shares of total output and
CO
2
emissions,
based on above mentioned empirical sectoral interlinkages and feedbacks. More detailed
discussions of the limitations of the input
-
ou
tput met
hodology can be found here
6
.
Linking Chinese input
-
output tables with a multi
-
regional input
-
output model also introduces
several issues. Greatest
care has been applied in reproducing the transformation
procedures
and concordances lai
d out by
Prof. Liu Yu
22
, who originally provided the C
hinese IO tables
for the GTAP
database. But on the treatment of the “other” column, which is a statistical
remainder of conflicting data in the
Chinese
input
-
output table
s
, the authors
of this study
diverge and follow
Minx et al.
10
, who treat
this column as uncertainty
and error column
,
rather than integrating it into the overall Chinese input
-
output structure
22
. Furthermore the
GTAP team used detailed trade price indices and trade data to link up the individual country
tables and establish sectoral interlinkages, while for this study only sh
ares of total outputs
and
importshares
6,10
were used to estimate domestic and imported
demand
. For further systematic