Understanding China’s past and future energy demand: An exergy
efﬁciency and decomposition analysis
Paul E. Brockway
, Julia K. Steinberger, John R. Barrett, Timothy J. Foxon
Sustainability Research Institute, School of Earth and Environment, University of Leeds, LS2 9JT, UK
We complete the ﬁrst time series exergy and useful work study of China (1971–2010).
Novel exergy approach to understand China’s past and future energy consumption.
China’s exergy efﬁciency rose from 5% to 13%, and is now above US (11%).
Decomposition ﬁnds this is due to structural change not technical leapfrogging.
Results suggests current models may underestimate China’s future energy demand.
Received 12 February 2015
Received in revised form 3 May 2015
Accepted 22 May 2015
There are very few useful work and exergy analysis studies for China, and fewer still that consider how
the results inform drivers of past and future energy consumption. This is surprising: China is the world’s
largest energy consumer, whilst exergy analysis provides a robust thermodynamic framework for analys-
ing the technical efﬁciency of energy use. In response, we develop three novel sub-analyses. First we per-
form a long-term whole economy time-series exergy analysis for China (1971–2010). We ﬁnd a 10-fold
growth in China’s useful work since 1971, which is supplied by a 4-fold increase in primary energy cou-
pled to a 2.5-fold gain in aggregate exergy conversion efﬁciency to useful work: from 5% to 12.5%. Second,
using index decomposition we expose the key driver of efﬁciency growth as not ‘technological leapfrog-
ging’ but structural change: i.e. increasing reliance on thermodynamically efﬁcient (but very energy
intensive) heavy industrial activities. Third, we extend our useful work analysis to estimate China’s future
primary energy demand, and ﬁnd values for 2030 that are signiﬁcantly above mainstream projections.
Ó2015 The Authors. Published by Elsevier Ltd. This is an open accessarticle under the CC BY license (http://
As the world’s economic powerhouse and largest energy
consumer , much effort is spent understanding China’s historical
energy consumption (e.g. [2–4]) and future energy demand [5–7].
However these studies typically examine primary or ﬁnal energy
data, rather than useful work values obtained using an exergy anal-
ysis based technique. This is the research gap that this paper seeks
to address. Exergy analysis takes a broader, whole system
approach to energy analysis, giving ‘‘a measure of the thermody-
namic quality of an energy carrier’’ (p. 686, ), thereby enabling
a robust view of useful work consumed in provision of energy ser-
vices. Exergy analysis also has the beneﬁt of taking into account
more aspects of the energy supply chain than traditional energy
analysis, and in a more consistent way. A ﬂow visualisation of pri-
mary exergy to useful work is given in Fig. 1.
A key assumption in this study is that useful work is a better
‘energy parameter’ than primary energy on which to analyse end
energy use and economic activity, since – as Fig. 1 shows – it is
the last thermodynamic place where energy is measured before
it is exchanged for energy services. We are not alone in this view.
Numerous authors (e.g. [8–10]) suggest second law exergy analy-
ses can help understand national-scale energy use. For economic
insights, Percebois  suggested in 1979 that energy intensity
metrics (i.e. energy consumption relative to GDP) were better
undertaken at the energy output stage, since it ‘‘allows us to anal-
yse structural change in energy supply and situates our analysis at
the level of satisﬁed needs’’. Serrenho et al.  recent work on
useful work intensity supports this assertion. Meanwhile, Warr
and Ayres , Santos et al.  and Guevara et al. , all found
empirical evidence suggesting useful work is a better candidate as
0306-2619/Ó2015 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
E-mail address: firstname.lastname@example.org (P.E. Brockway).
Applied Energy 155 (2015) 892–903
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/apenergy
a factor of production (than primary energy) to explain economic
growth. This gets us to the crux of our argument: if it is useful work
and not primary energy that supplies economic needs, then we
should conduct energy and economic analyses at that level.
The few published time-series studies of useful work account-
ing have focussed largely on industrialised countries including
the US, UK and Japan (e.g. [16–18] and later all EU-15 countries
. Somewhat curiously, these country-scale analyses typically
focus on economic implications and linkages, rather than
energy-based conclusions. Brockway et al.  set out to address
this imbalance, by undertaking a 50 year time-series analysis
(1960–2010) of the US and UK. They found the US and UK may
no longer be increasing their aggregate exergy efﬁciency, as
increases in process level efﬁciencies are offset by efﬁciency dilu-
tion taking place , following the case of Japan . In short:
individual technology gains in efﬁciency are being overtaken by
using increasing amounts of less efﬁcient processes, such as
air-conditioning. This raises the question: could the same be hap-
pening in China?
Numerous Extended Exergy Accounting (EEA) studies have been
published on China (e.g. [20–32]). EEA is a biophysical exergy anal-
ysis method, developed largely by Wall and Sciubba in the 1990s
(e.g. [33–35]) to examine the embedded exergy of all natural
resources inputs (e.g. energy, natural materials) and associated
outputs of the economy (e.g. food, materials, wastes). This valuable
technique helps understand societal exergy consumption. It is
complementary to the useful work accounting method applied
here, which is based on an ‘‘energy carriers for energy use’’
approach  introduced at a national-scale by Reistad  in
1975, which examines the exergy destruction of energy conversion
processes from primary exergy to end useful work. The key distinc-
tion is that EEA is akin to a mass-balance analysis (except it studies
exergy content not mass) whereas Reistad’s approach estimates
the thermodynamic work done by the energy system to deliver
energy services. It is the latter approach we require for detailed
energy system analysis – and such national-scale useful work
accounting studies for China are rare (e.g. ), and none to date
examine a long time-series.
To address the lack of exergy-based analyses in China which
examine time-series results through an energy demand lens, we
pose the following research question: What new insights can useful
work analysis provide for historical and future energy demand in
China? In response, we provide three novel, linked analyses. To
start, we undertake the ﬁrst historical exergy efﬁciency and useful
work analysis for China, covering the period 1971–2010. Next, we
adopt an index decomposition analysis to identify the key drivers
of change in China’s useful work. Last, we develop a useful work
based method for projecting China’s primary energy demand to
2030, and also test implications of potential future declines in
the rate of exergy efﬁciency improvement.
The paper proceeds as follows. After the Introduction, Section 2
contains Methods and Data, Results and Discussions are in
Section 3, with Conclusions in Section 4.
2. Methods and data
2.1. Historical useful work analysis (1971–2010)
2.1.1. Method Summary
Reistad  deﬁned exergy as ‘available energy’. As depicted
in Fig. 1, at a country-scale, primary exergy of energy carriers
(e.g. coal, oil, gas, renewables, food and feed) is transformed into
ready to use ‘ﬁnal energy’ (e.g., diesel or electricity), which is
then used to provide ‘useful work’ (i.e. through heat, mechanical
drive, manual labour or electrical devices), to ultimately provide
energy services (e.g. warmth, light, cooling, sustenance).
Carnahan et al.  deﬁned task-level ‘useful work’ ðU
‘‘the minimum exergy input to achieve that task work transfer’’.
For our purposes, task-level means sub-class (j) (e.g. diesel road
transport or low temperature heat) levels nesting within overall
main classes (i) of energy use (i.e. heat, muscle work, transport,
mechanical drive). Task-level exergy efﬁciency ð
the second law thermodynamic efﬁciency of the energy conver-
sion from primary exergy to end useful work, deﬁned by
Carnahan et al.  as:
Primary exergy values at task-level ðE
Þare then multiplied
with their associated task-level exergy efﬁciencies (
) to give an
estimate for task-level useful work ðU
Þ. When summed, we derive
an overall estimate for the total national-scale useful work
Þvia Eq. (2). Finally, national exergy (second law) efﬁ-
) is given by Eq. (3), which – following Carnahan et al.
 – we adopt as a country-scale measure of energy efﬁciency,
and use it as a term throughout this paper for consistency. Eq.
(2) also reveals the obvious (but important, as we see later) obser-
vation that useful work changes are supplied by changes in pri-
mary exergy and/or exergy efﬁciency.
Fig. 1. Conceptual diagram of primary exergy to useful work.
¼The minimum exergy input to achieve that task work transfer
Maximum amount of reversible work done as system reaches equlibrium ð1Þ
P.E. Brockway et al. / Applied Energy 155 (2015) 892–903 893
Our country-scale useful work accounting approach builds on
the methodology developed by numerous authors including
Reistad , Wall , Ayres and Warr , and more recently
Serrenho et al. , who introduced a consistent International
Energy Agency (IEA) based input energy mapping framework.
Brockway et al.  made further advances to electricity applica-
tions and mechanical drive classes, which is also used in this study
for consistency and comparability. We apply these advances to
produce a ﬁrst time-series analysis of China. Fig. 2 gives an over-
view of the basic stages:
2.1.2. Input data
Primary exergy inputs, E
;are ﬁrst derived. IEA energy datasets
1971–2010 (1) for fossil fuel and biomass (combustible renew-
ables) provided much of the base data. IEA primary energy values
are converted to primary exergy inputs using chemical exergy
coefﬁcients . At an aggregate level, total primary exergy is
around 5% higher than the IEA’s Total Primary Energy Supply
(TPES) values. The inputs E
are then mapped to three main classes
(heat, mechanical drive and electricity) and to task-levels where
possible (e.g. Low Temperature Heat (LTH)), following recent
approaches [19,42]. The task-levels are listed in Appendix A.In
some cases, we extend the IEA end energy use breakdown to more
granular levels (e.g. road fuel split between transport modes) by
supplementing Chinese end consumption data in three key areas:
buildings [3,44–48]; transport [49–53]; and industry [54–58].
Next, task-level exergy efﬁciencies ð
Þfor transport, heat, and
electricity are added. Previous US–UK values  are modiﬁed by
Chinese data as follows. For transport, local fuel economy data
was used for road and rail [52,53,59,60]. For calculating Carnot efﬁ-
ciencies (for heat exergy efﬁciencies), for external temperatures we
used 1971–2010 China monthly air temperature data , whilst
indoor temperatures (for LTH efﬁciencies) were weighted for
China’s city/rural split and assume a 20 year lag in comfort levels
versus UK data . LTH ﬁrst law efﬁciencies are based on Warr
et al. , Chen et al.  and Edwards et al. . Steel and ammo-
nia industries are adopted (as with US–UK study) as representative
of High Temperature Heat (HTH) efﬁciencies, by virtue of having
the two highest proportions of Chinese industrial energy use
. Process (GJ/tes) efﬁciency data for steel [54–57,64,65] and
ammonia (taken as 75% of UK values, based on average values from
Phylipsen et al. ) and the IEA  are combined with temper-
ature data to calculate time-series exergy efﬁciencies. For electric-
ity application efﬁciencies, values of 80% of those from the US–UK
analysis were typically used, based on evidence that China’s aver-
age devices were 10–20 years behind US–UK values across indus-
try, commerce and residential sectors [3,67].
Then, we calculated primary exergy and useful work values for
a fourth main class: muscle work. For human labour, estimates fol-
low Brockway et al. approach : using manual labour popula-
tion [69,70], food intake data [71,72], and Smil’s estimated 13%
conversion efﬁciency of food to human useful work . For
draught animals, we assumed 100 million draught animals in
China in 1990 , and a 1% annual decline in numbers from
1971 to 2010, mirroring India . For animal useful work outputs,
we assumed 400 W average power output for a 5 h working day
over 120 working days/year, based on published data [74,76,77].
Estimates of intake feed requirements were based on
Ramaswamy and Krausmann [74,78].
Last, a note on data quality. For input energy data, two system-
atic discrepancies mean our national-level datasets underestimate
actual primary energy use. First, at a national-scale, IEA-based TPES
values are 5% lower than those of Lawrence Berkeley National
Laboratory (LBNL) China Energy Databook . Second, reported
aggregate primary energy consumption in China is 10% higher
from aggregated regional versus national datasets . However,
these differences are expected to be systematic, and thus have lim-
ited overall effect for our trends analysis. For task-level efﬁciencies,
whilst the China data sources are weaker in many instances than
the previous US–UK studies , overall trends and comparison
to US–UK results remain valid.
2.1.3. Useful work accounting outputs
Appendix A shows the task-level outputs of useful work, pri-
mary exergy and exergy efﬁciency. This data serves as task-level
inputs to the Logarithmic Mean Divisia Index (LMDI) decomposi-
tion analysis, or is summed to give useful work or exergy efﬁcien-
cies at main class level (i.e. heat, mechanical drive, electricity and
muscle work) and country-scale.
2.2. LMDI decomposition (1971–2010)
LMDI decomposition is now the mainstream Index
Decomposition Analysis (IDA) technique for analysing drivers of
changes in CO
emissions (e.g. [80,81]) and sectoral energy use
such as manufacturing and transport (e.g. [82,83]). Using the
LMDI approach, we develop a new approach to reveal the relative
contribution of energy and efﬁciency drivers to China’s historical
useful work (U). First, we expand Eq. (2) (U=PE
) to yield Eq.
(4), which is based on task-level useful work (U
) and primary
), enabling the historical results to act as the input data
for the LMDI analysis. Eqs. (5)–(9) give the four drivers of useful
work changes: Input Exergy (D
); Main class structure (D
IEA Primary Energy data
(fossil fuels and biomass)
IEA Primary energy mapping
to task-level end uses
Food and feed data
Exergy chemical equivalent
Primary exergy mapping to
Useful Work, Eij
Local country energy data
(e.g. electricity end use)
Task level exergy
results, U ij
Fig. 2. Useful work analysis ﬂowchart.
894 P.E. Brockway et al. / Applied Energy 155 (2015) 892–903
sub-class (i.e. task) level structural change (D
); and task-level
). This shows how LMDI decomposition can be used
to breakdown the overall exergy efﬁciency changes (from the main
analysis results in Section 3.1) into three parts (D
! ! ð6Þ
! ! ð7Þ
! ! ð8Þ
! ! ð9Þ
E= Primary exergy input to economy
= Main class exergy input
= Task-level exergy input
= Task-level useful work output
= log mean weighting function
X= Exergy input
S= Main class share of exergy input (E
L= Task-level share of exergy input within main class (E
F= Task-level exergy efﬁciency(U
= change in overall exergy input (E)
= change in share of exergy inputs between main classes (E
= change in task-level shares (E
) of exergy inputs within
each main class
2.3. China energy demand scenarios 2010–2030
After conducting the historical and decomposition analyses, we
develop and trial a new useful work-based methodology to esti-
mate primary energy demand to 2030, based on projections of
GDP and extrapolations of task-level exergy efﬁciencies under
illustrative constant and declining exergy efﬁciency growth rate
scenarios. Four steps were required. The ﬁrst estimates China’s
useful work requirement for 2010–2030. To do this, 1971–2010
overall useful work energy intensity (UW/GDP) – calculated from
historical GDP data  – is extrapolated using a best-ﬁtting curve
to 2030. Using World Bank forecasts of GDP for 2011–2030  –
see also the Supplementary Information – Section S1, China’s total
useful work (to deliver that GDP) in 2030 is then estimated.
Second, total projected useful work to 2030 is allocated to
task-levels. To start, useful work proportions from main classes
are estimated based on historic trend comparison in UK, US and
China. China and US allocations are shown in Fig. 3. Then, task level
allocations are derived, also based on comparisons to previous
US-UK values, which place China as 40 years behind US–UK allo-
cations. These results at task-level are given in the Supplementary
Information – Section S2.
Third, task-level exergy efﬁciencies are projected to 2030 under
two illustrative scenarios which have different efﬁciency gains
assumptions. In Scenario 1 (constant efﬁciency gains), China’s
1990–2010 task-level exergy efﬁciency changes are extended to
2010–2030. Typically this places China’s task-level efﬁciencies in
2030 as those of average US-UK values in 2010. In Scenario 2
(declining efﬁciency gains), only half of China’s 1990–2010 efﬁ-
ciency gains are extended to 2010–2030, with two thirds of these
reduced gains assumed to occur in 2010–2020. There is some jus-
tiﬁcation for the declining gains scenario, as Brockway et al. 
found that efﬁciency gains in important task-levels (e.g. residential
electricity and LTH) slowed or reversed in 1990–2010 (versus
1970–1990). Assuming an average 20 year lag for China, this could
mean similar effects exhibited in China by 2030. More detailed efﬁ-
ciency results at task-level are given in the Supplementary
Information – Section S3. Whilst other efﬁciency scenarios are pos-
sible (and indeed probable), our two selected cases are intended to
represent the possible envelope of task-level efﬁciencies for 2010–
2030, and are thus valid to study the effects of declining efﬁciency
Fourth, estimates of total primary energy demand for 2010–
2030 are made at task-level (Eq. (11)) and aggregate level (Eq.
(12)). Sufﬁx 1 and 2 refer to Scenario 1 and 2. Finally, the chemical
exergy conversion ratios  are removed to reveal primary
Fig. 3. China (1971–2030) and US (1860–2010) useful work allocations.
P.E. Brockway et al. / Applied Energy 155 (2015) 892–903 895
energy (i.e. TPES) projections to 2030 under these two scenarios,
with differences suggesting impacts of declining exergy efﬁciency
gains on primary energy demand.
3. Results and discussion
3.1. 1971–2010. useful work accounting results
Table 1 summarises useful work, primary exergy and exergy
efﬁciency results for 1971–2010, with task-level results given in
Appendix A for 1971 and 2010. China’s end useful work has
increased 10-fold since 1971, with electricity applications and
HTH industrial uses growing from 30% to 53% of total useful work.
Conversely, muscle work and low temperature heat have together
declined from 40% of total useful work to 8%.
Aggregate exergy efﬁciency has grown almost linearly from
5.3% to 12.5%. Table 1 (together with Appendix A) suggest a key
factor is the structural shift from (low efﬁciency) muscle work
and low temperature heat (20 °C) to (high efﬁciency) HTH. Fig. 4
illustrates a second reason: the strong growth in mechanical drive
and heat class efﬁciencies – which make up over half of total pri-
mary exergy inputs. The question of whether this linear aggregate
efﬁciency trend can continue is considered via the future scenario
analysis in Section 3.3.
Fig. 4 also compares China’s aggregate efﬁciency growth to the
stable US (10–11%) values from the previous US-UK study .
China’s exergy efﬁciency overtakes the US by around 2004. At ﬁrst,
it is tempting to see China’s overtaking of the US’s aggregate efﬁ-
ciency as ‘technological leapfrogging’ (e.g. ) – i.e. rapidly adopt-
ing high-efﬁciency technologies without having to deal with the
legacy of past low efﬁciency capital stock. In fact this is not the
case, since task-level exergy efﬁciencies are generally lower than
the US (except mechanical drive, which is a small component of
China’s energy use). This result implies structural differences make
a signiﬁcant contribution to China’s increasing efﬁciency: i.e. its
production-focused industrial economy uses more high tempera-
ture heat and industrial processes versus the US’s mature con-
sumer economy. The index decomposition results in Section 3.2
support this view. In turn, this implies as China’s economy also
matures and its structure shifts towards that of the US, that this
may have a diluting effect on future overall exergy efﬁciency, as
seen later in Section 3.3.
Summary of useful work analysis results 1971–2010.
Useful work analysis: output category 1971 1980 1990 2000 2010
Main category end use PJ % of total PJ % of total PJ % of total PJ % of total PJ % of total
Direct heat 1087 71 1848 71 2787 68 3901 59 7602 51
Mechanical Drive 126 8 268 10 477 12 1140 17 2633 18
Electricity end uses 154 10 343 13 684 17 1460 22 4665 31
Muscle work 157 10 151 6 148 4 137 2 127 1
Total 1524 100 2610 100 4096 100 6638 100 15027 100
Direct heat 15370 54 21560 56 28271 55 31504 49 51983 43
Mechanical Drive 1090 4 1971 5 2988 6 5625 9 12720 11
Electricity end uses 1592 6 3519 9 6283 12 13951 22 42507 35
Muscle work 10661 37 11760 30 13489 26 13398 21 13159 11
Total 28713 100 38810 100 51032 100 64478 100 120,369 100
Exergy efﬁciency (useful work/primary exergy)
Main category end use % efﬁciency % efﬁciency % efﬁciency % efﬁciency % efﬁciency
Direct heat 7.1 8.6 9.9 12.4 14.6
Mechanical Drive 11.6 13.6 16.0 20.3 20.7
Electricity end uses 9.7 9.7 10.9 10.5 11.0
Muscle work 1.5 1.3 1.1 1.0 1.0
Total 5.3 6.7 8.0 10.3 12.5
Fig. 4. China’s exergy efﬁciency by end use 1971–2010, compared to US aggregate efﬁciency.
896 P.E. Brockway et al. / Applied Energy 155 (2015) 892–903
Few comparative estimates are available of aggregate Chinese
efﬁciencies. Chen and Chen  calculate a value of 20%, twice that
of our 10% value for China in 2000. The main reasons are due their
exclusion of muscle work, and higher industry efﬁciency (e.g. 78%
for the chemical sector). Nakicenovic  estimated reforming
countries (e.g. China) exergy efﬁciencies in 1990 to be 10%, of a
similar order to our 8% estimate for 1990.
Fig. 5 shows how China’s 10-fold useful work growth was sup-
plied by a 4-fold increase in primary energy coupled to a 2.5-fold
gain in aggregate exergy efﬁciency: from 5% to 12.5%. In other
words, if China’s exergy efﬁciency had stayed at 5%, a 10-fold gain
in primary exergy would have been required to achieve the same
useful work supply level.
Finally, to understand the overall ﬂow of exergy to end useful
work, and the exergy losses that occur during the various conver-
sion processes, useful work-based Sankey diagrams of China are
constructed for 1971 and 2010, as shown in Appendix B. They
show the transformation of China in 40 years from a largely agri-
cultural to industrial economy. By 2010, China is dominated by
energy dense fossil fuel inputs (versus food and feed for muscle
work) and energy intensive end uses, particularly in industry,
which underpins the rise in overall exergy efﬁciency.
3.2. LMDI decomposition results 1971–2010
The multiplicative factors are summarised in Table 2 for the
period 1971–2010, comparing three countries: China, the UK and
US. For China, the largest contribution to useful work growth is pri-
mary exergy, conﬁrming the result of Fig. 5 Importantly, the overall
efﬁciency gain factor (2.5) is now split into three parts. First, the
main class structural change (1.39) tracks the move from less efﬁ-
cient (i.e. muscle work) to more efﬁcient (i.e. heat) main classes.
Second, we ﬁnd sub-class structural change (1.19) is above 1.00,
which means that within each main class there has also been an
efﬁciency ‘concentration’ effect. (In contrast note the efﬁciency
‘dilution’ values of 0.87–0.88 for the US and UK). This is due to
China’s transition from agricultural society to industrial powerhouse,
causing structural shifts within main classes from lower to higher
efﬁciency categories (e.g. LTH to HTH). Third, task-level efﬁciency
gains (1.48) are the largest of the three efﬁciency gain factors.
The value of using the LMDI approach is also highlighted by
Table 2. Firstly, it conﬁrms and quantiﬁes the assertion stated in
Section 3.2: that overall structural change (1.66) is at least as
important to overall efﬁciency gains as task-level efﬁciency gains
(1.48). Secondly, we can directly compare factors to other coun-
tries. In this case, we see that China has not reached the point of
efﬁciency ‘dilution’ that can be seen in the US and UK – where
would be below 1.00 – as found earlier by Williams et al.
 for Japan. China’s improvements to task-level efﬁciencies
(1.48) are similar to US (1.29) and UK (1.58) values, conﬁrming that
instead of technological leapfrogging, it is overall structural change
(1.66 for China versus 0.90 for US and UK) that has been responsi-
ble for China’s rise in overall aggregate efﬁciency to overtake the
3.3. Future exergy efﬁciency: impacts on primary energy projections
3.3.1. Step 1 – Useful work projection to 2030
China’s useful work and primary energy intensities (of eco-
nomic activity) are shown in Fig. 6, based on constant price GDP.
It shows a 66% reduction in useful work intensity from 12.0
(GJ/2005$US) in 1971 to 3.9 (GJ/2005$US) in 2010, compared to
an 86% reduction in primary energy intensity (210.7 to
29.8 GJ/2005$US) – the standard metric for energy intensity (e.g.
) – over the same period. The greater stability of useful work
intensity suggests useful work is more closely linked to GDP than
primary energy – supporting the key assumption noted earlier.
Useful work and primary energy intensities are projected to 2030
using best-ﬁtting trendlines as shown in Fig. 6.
The World Bank’s GDP forecast for China in 2030  is
$13.5 Trillion(US2005), a 3.5-fold increase from the
$3.8 Trillion(US2005) value in 2010. Using the useful work inten-
sity projection of 2.45 (GJ/US$2005) for 2030, this gives a useful
work estimate of 33.1EJ in 2030 (just over double the 15.0EJ con-
sumed in 2010) – to deliver that level of GDP.
3.3.2. Step 2 – Allocation of task-level useful work
Fig. 7 shows the projected annual useful work growth to 2030 is
almost linearly 27–28 Mtoe/year. This is due to two effects can-
celling each other out: a slowdown in GDP growth mirroring useful
work intensity reductions. At a main class level, as China’s econ-
omy matures, a slowdown in heat’s contribution to useful work
is offset by growth in electricity and mechanical drive (mainly
transport) classes. This appears broadly consistent with other eco-
nomic forecasts for China used in energy modelling (e.g. ).
Summary of LMDI decomposition factors 1971–2010 for China, US, UK.
Country U Dex Dstr Ddil Deff
China 9.76 3.96 1.39 1.19 1.48
US 1.53 1.32 1.03 0.88 1.29
UK 1.43 1.01 1.04 0.87 1.58
Overall structural change Task-level
China 9.76 3.96 1.66 1.48
US 1.53 1.32 0.90 1.29
UK 1.43 1.01 0.90 1.58
Fig. 5. China 1971–2010 useful work analysis results vs 1971 datum.
P.E. Brockway et al. / Applied Energy 155 (2015) 892–903 897
3.3.3. Step 3 – Task-level exergy efﬁciencies
Next, task-level exergy efﬁciencies are projected based on the
linear and declining gains scenarios described earlier – see
Supplementary Information. The results at main class level are
shown in Fig. 8. In Scenario 1, stable gains in task-level exergy efﬁ-
ciencies are combined with structural change in China in 2011–
2030 – moving towards a more service sector-based economy,
with associated decreases in higher efﬁciency processes (e.g. high
temperature heat) and increases in low-efﬁciency activities (e.g.
residential and commercial electricity), as shown earlier in Fig. 3.
This results in only a small increase in national aggregate exergy
efﬁciency from 12.5% to 13.0% in 2030. The green wedge in Fig. 8
illustrates the effect of this structural change, compared to a sim-
ple extrapolation of China’s 1990–2010 aggregate efﬁciency, which
would result in aggregate exergy efﬁciency of around 17% in 2030.
In Scenario 2, which includes both structural change and slowing
of task-level efﬁciency gains, aggregate exergy efﬁciency peaks at
12.8% before 2025, then reduces to 12.5% by 2030. Therefore most
of the reduction in overall efﬁciency is due to assumed structural
change than the difference in task-level efﬁciencies under the
For heat and mechanical drive classes, the projected efﬁciency
dilution is so strong (i.e. less industrial usage and more consumer/
commercial use), their efﬁciencies decline by 2030 under both sce-
narios. As electricity provides an increasing share of useful work by
2030, this accelerates the slowdown (scenario 1) and decline (sce-
nario 2) in overall exergy efﬁciency. Mechanical drive efﬁciency
stagnates in this analysis under both scenarios, since it balances
task-level efﬁciencies that were increasing (e.g. static motors and
aviation) and decreasing (e.g. road transport – due to more
Primary Energy Intensity (GJ/US$2005)
Fig. 6. Comparison of China primary energy and useful work intensities.
Fig. 7. China – useful work projection to 2030.
Eﬃciency scenario diﬀerences
Fig. 8. China – exergy efﬁciency scenario results.
898 P.E. Brockway et al. / Applied Energy 155 (2015) 892–903
cars/less motorcycles, and more heavy duty-trucks). However, as
the smallest of the three main classes, this effect has limited
impact on the aggregate exergy efﬁciency.
3.3.4. Step 4 – Primary end demand in 2030
Finally, the useful work-based primary energy estimates are cal-
culated based on the assumed efﬁciency scenarios. The results are
compared in Fig. 9 to ﬁve published reference (i.e. current policies)
scenarios [67,89–92] and a top-down primary energy intensity
(TPES/GDP) based estimate (derived econometrically via the
best-ﬁt TPES/GDP projection shown earlier in Fig. 6). By 2030, our
Scenario 1 (6000 Mtoe/year) requires 900 Mtoe/year more primary
energy than the econometric estimate, whilst Scenario 2 – due to
assumed declining efﬁciency gains – requires an additional
300 Mtoe/year (compared to Scenario 1). The TPES/GDP derived pri-
mary energy estimate (as with the other ﬁve reference projections)
slows over time, following the assumed slow-down in GDP growth.
In contrast, our useful work derived projections show more linear
increases, as with ﬂat overall exergy efﬁciencies (shown earlier in
Fig. 8), the linear projected growth in useful work required (see ear-
lier Fig. 7) is passed on to required primary energy inputs.
Our useful work-based projections are signiﬁcantly higher than
the ﬁve reference cases. The three reference scenarios using a 2010
base year [89–91] produce estimates of 4300–5000 Mtoe/year in
2030, whilst the two scenarios with a 2005 base year [67,92] esti-
mate primary energy consumption as 3200 Mtoe in 2030. A key
aspect therefore appears the choice of base year, with the 2005
base year models missing China’s step up in energy consumption,
and so undercut the projections of later base year models.
Perhaps this illustrates how tricky energy forecasting is, as Smil
 notes: ‘‘long-range energy forecasters have missed every
important shift of the past 2 generations..[and they]..will continue
to be wrong’’.
Nevertheless, the fact remains the traditional energy models
give lower estimates of primary energy than our simple useful
work-based approach – so it’s worth reﬂecting on this. Most
importantly, we base our projections on a different energy inten-
sity metric versus mainstream models – ours is based on useful
work (U/GDP), as this measures the energy level delivered to eco-
nomic activities, rather than on primary energy (TPES/GDP) enter-
ing the economy. Moreover, our TPES/GDP based projection is 20%
below our U/GDP based projections – showing that this distinction
is an important one. The GDP projections that we use are consis-
tent with other models (e.g. ). Our methodology is also
top-down: it starts from an aggregate demand estimation, and
then builds up its constituent elements from task-share trends.
Other energy models tend to be bottom-up, using demand and
technology trends of various sectors. We attach more detailed sce-
nario data in the Supplementary Information.
Whilst we believe the useful work based approach to primary
energy forecasting is justiﬁed by the observed links between
aggregate economic activity and useful work, signiﬁcant caveats
exist around the accuracy of the underlying data to our energy pro-
jection conclusions. For the useful work calculations for 1971–
2010, though the primary exergy data is relatively robust (relying
mainly on IEA energy balance data), the task-level efﬁciencies have
greater uncertainty, being based on often partial data. In turn, pro-
jecting task-level useful work allocations and exergy efﬁciencies to
2030 ampliﬁes any data inaccuracies. However the driving ratio-
nale of the paper was to develop a new technique based on useful
work. The result highlights the possible importance of this method
and thus mandate for further study.
To address the lack of time-series exergy analyses for China
which examine energy demand drivers and implications, we set
the following research question: What new insights can useful work
analysis provide for historical and future energy demand in China?
First, our historical analysis found China’s exergy efﬁciency grew
linearly from 5.3% (1971) to an impressive 12.5% (2010), placing
it between the US (11%) and the UK (15%). In addition, a striking
10-fold rise in China’s useful work occurred from 1971 to 2010,
supplied by a 4-fold increase in primary exergy and a 2.5-fold
increase in exergy efﬁciency. Second, using LMDI decomposition
we found efﬁciency growth was split evenly between task-level
efﬁciency gains and structural change (e.g. moving from muscle
work to mechanical drive). Third, a new useful work-based energy
forecasting technique is developed and trialled, which – based on
two illustrative exergy efﬁciency scenarios – projects China’s
2030 primary energy demand in the range of 6000–6300 Mtoe, sig-
niﬁcantly higher than the 4500–5200 Mtoe estimates from pub-
lished sources using traditional energy models which use the
same 2010 baseline year.
The results allow several key insights. Firstly, if China’s exergy
efﬁciency had stayed at 5%, a 10-fold (rather than 4-fold) gain in pri-
mary exergy would have been required to achieve the same useful
work supply level. Through the mechanism of the macro-economic
rebound effect, however, as Ayres et al.  and Schipper and
Grubb  established, lower efﬁciency gains may in fact translate
to lower economic growth, and hence lower required useful work.
Second, the application of LMDI decomposition to useful work
results provided robust insights: revealing China’s efﬁciency rise
above the US was not due to technological leapfrogging, but greater
use of energy intensive (yet more exergy efﬁcient) industrial pro-
cesses. Third, in common with the US and UK, China may approach
an asymptotic exergy efﬁciency maximum by 2030, as its economy
matures and efﬁciency dilution starts. Such dilution is already
Fig. 9. China – primary energy (TPES) forecasts to 2030.
P.E. Brockway et al. / Applied Energy 155 (2015) 892–903 899
forecast: the modal shift to cars  will reduce mechanical drive
exergy efﬁciency; a rapid increase in residential electricity ;
and a peaking in the share of HTH allied to a shift to greater residen-
tial LTH. Fourth, our extension of useful work based technique pro-
jects higher primary energy demand in China by 2030 versus
traditional bottom-up energy model estimates (i.e. based on pri-
mary or ﬁnal energy). Further studies investigating the possible
reasons (e.g. differences in assumed future energy efﬁciency sav-
ings, structural consumption, energy rebound and efﬁciency dilu-
tion) would therefore be beneﬁcial.
Overall, the useful work method appears a valuable technique to
give new insights into Chinese energy consumption and efﬁciency –
past, present and future. Given the implications to future energy
demand and associated policies, further research is encouraged.
First, work to improve the consistency of the useful work method
would be of beneﬁt – such as the treatment of renewables,
non-energy use, active/passive system efﬁciencies, or extending the
analysis boundary to include energy services, as others suggest
[38,87,96]. Second, contrast the construction of traditional (primary
and ﬁnal energy) versus useful work energy models, to uncover the
reasons for energy projection differences. Third, undertake further
economic analysis to test the key assumption underpinning this
work: that useful work is a more suitable parameter for energy and
economic analysis than primary energy. Lastly, policy implications
could be explored – such as how to meet higher (than expected) pri-
mary energy demand, or how to amend micro-efﬁciency policies to
capture energy savings before rebound occurs.
We gratefully acknowledge the support of Engineering and
Physical Sciences Research Council (EPSRC) and Arup for contribut-
ing to the PhD CASE (Collaborative Award in Science and
Engineering) scholarship of the ﬁrst author. The support of the
Economic and Social Research Council (ESRC) is also gratefully
acknowledged. The work contributes to the programme of the
ESRC Centre for Climate Change Economics and Policy. Also, we
thank João de Santos and Tiago Domingos of Instituto Superior
Técnico, Lisbon, for sharing their unpublished work. We would like
to thank the two anonymous reviewers and the editor of Applied
Energy for their useful suggestions and valuable comments.
Appendix A. Useful work accounting outputs: China – 1971,
(see Table A1)
Useful work accounting outputs: China – 1971, 2010.
Main class, iTask level, j1971 2010
PJ PJ (%) PJ PJ (%)
Heat LTH (Low Temperature Heating 20 °C) 435 10,103 4.3 973 20,281 4.8
MTH1 (Medium Temperature Heating 100 °C) 30 247 12.1 592 4428 13.4
MTH2 (Medium Temperature Heating 200 °C) 301 2657 11.3 1913 10,506 18.2
HTH (High Temperature Heating 600 °C) 283 2362 12.0 3963 16,695 23.7
Sub total 1049 15,370 6.8 7441 51,910 14.3
Mechanical drive – Gas/diesel oil (assume diesel road
20 114 17.3 821 3728 22.0
Mechanical drive – Domestic Aviation fuel, jet fuel 0 0 n/a 158 642 24.6
Mechanical drive – Gasoline fuel (Petrol road vehicles 42 242 17.3 691 3920 17.6
Mechanical drive – Diesel/gas oil fuel (Boat engines) 2 17 13.0 188 915 20.5
Mechanical drive – Industry static motors (diesel
28 118 23.5 522 1934 27.0
Mechanical drive – Gas/diesel fuel (diesel trains) 1 8 13.0 64 310 20.5
Mechanical drive – Gas/diesel fuel (tractors) 26 255 10.2 88 799 11.0
Mechanical Drive – bio-diesel/bio-gasoline (road
0 0 n/a 16 80 n/a
Mechanical drive – bio-diesel/bio-gasoline (road
0 0 n/a 0 0 n/a
Mechanical drive – Gas/diesel oil (assume diesel cars) 0 1 20.4 85 385 22.0
Mechanical drive – Gas ﬁred engines (for pipeline
0 0 n/a 2 7 n/a
Mechanical drive – Coal (steam powered trains) 7 333 2.2 0 0 n/a
Mechanical drive – Coal (steam powered boats) 0 0 n/a 0 0 n/a
Mechanical drive sub-total 126 1090 11.6 2633 12,720 20.7
Electricity Lighting 1 84 0.8 46 2836 1.6
Domestic/commercial – Space heating 0 31 1.3 57 3303 1.7
Domestic – Hot water/cooking 1 21 3.0 47 1272 3.7
Industry – HTH process heating 16 199 7.9 443 4537 9.8
Electrolytic end use – Industry 11 153 7.5 370 3490 10.6
Communications/electric devices 0 0 0.1 1 267 0.3
Refrigeration/air conditioning 4 307 1.4 144 8221 1.7
Domestic – Wet/dry motor driven appliances 0 0 10.0 17 138 12.6
Other mechanical drive motors 123 797 15.4 3537 18,442 19.2
Electricity – Sub-total 156 1592 9.8 4662 42507 11.0
Muscle work Human 26 5432 0.5 38 9626 0.5
Draught animals 131 5229 2.5 88 3533 2.5
Muscle work – Sub-total 157 10,661 1.5 127 13,159 1.0
Total Grand total 1488 28,713 5.2 14,863 120,296 12.4
900 P.E. Brockway et al. / Applied Energy 155 (2015) 892–903
Appendix C. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.apenergy.2015.
Appendix D. Data Statement. Supplementary material
A complete results ﬁle, produced following the methodology
and sources described in this paper, has been deposited at the
University of Leeds Data Repository at http://doi.org/10.5518/7.
 International Energy Agency (IEA). Extended world energy balances, IEA World
energy statistics and balances (database); 2013. <http://www.oecd-ilibrary.
org/energy/ data/ie a-world-energy-statis tics-an d-balan ces/ext ended-world-
 International Energy Agency (IEA). Energy technology perspectives 2012:
Pathways to a clean energy system; 2012 <http://www.iea.org/etp/etp2012/>
 Letschert VE, Mcneil MA, Zhou N. Residential electricity demand in China – can
efﬁciency reverse the growth? Lawrence Berkeley National Laboratory (LBNL);
 Energy Foundation China. Background report: vehicle fuel economy in china;
 International Energy Agency (IEA). Energy technology perspectives 2010:
scenarios & strategies to 2050. IEA, France; 2010.
 International Energy Agency (IEA). World energy model documentation 2013
Version; 2013 <http://www.worldenergyoutlook.org/media/weowebsite/
 Organisation for Economic Co-operation and Development (OECD). Energy and
climate policy: bending the technological trajectory. OECD studies on
environmental innovation. OECD Publishing; 2012.
 Hammond GP. Industrial energy analysis, thermodynamics and sustainability.
Appl Energy 2007 Jul;84:675–700.
 Rosen MA, Dincer I, Kanoglu M. Role of exergy in increasing efﬁciency and
sustainability and reducing environmental impact. Energy Pol
 Groscurth HM, Kummel R, van Gool W. Thermodynamic limits to energy
optomisation. Energy 1989;14(5):241–58.
 Percebois J. Is the concept of energy intensity meaningful? Energy Econ 1979.
 Serrenho AC, Sousa T, Warr B, Ayres RU, Domingos T. Decomposition of useful
work intensity: the EU (European Union) – 15 countries from 1960 to 2009.
 Warr BS, Ayres RU. Evidence of causality between the quantity and quality of
energy consumption and economic growth. Energy Elsevier Ltd
 Santos J, Domingos T, Sousa T, Aubyn MS. Useful work & economic growth:
Portugal. TBC; 1960–2009.
 Guevara Z, Sousa T, Domingos T. An approach for a better evaluation of energy
efﬁciency trends using index decomposition analysis and useful work
accounting. Energy Econ 2014.
 Ayres RU, Warr B. Accounting for growth : the role of physical work. Struct
Chang Econ Dyn 2005;16:181–209.
 Williams E, Warr B, Ayres RU. Efﬁciency dilution: long-term exergy conversion
trends in Japan. Environ Sci Technol 2008;42(13):4964–70.
 Warr B, Ayres R, Eisenmenger N, Krausmann F, Schandl H. Energy use and
economic development: a comparative analysis of useful work supply in
Austria, Japan, the United Kingdom and the US during 100 years of economic
growth. Ecol Econ. Elsevier B.V 2010;69(10):1904–17.
 Brockway PE, Barrett JR, Foxon TJ, Steinberger JK. Divergence of trends in US
and UK aggregate exergy efﬁciencies 1960–2010. Environ Sci Technol
 Chen B, Chen GQ. Resource analysis of the Chinese society 1980–2002 based
on exergy—Part 4: Fishery and rangeland. Energy Pol 2007;35(4):2079–86.
 Chen GQ, Chen B. Extended-exergy analysis of the Chinese society. Energy.
Elsevier Ltd 2009;34(9):1127–44.
 Dai J, Chen B. Extended exergy-based ecological accounting of China during
2000–2007. Procedia Environ Sci 2011;5:87–95.
 Chen B, Chen GQ. Exergy analysis for resource conversion of the Chinese
Society 1993 under the material product system. Energy 2006;31(8–
 Dai J, Chen B, Sciubba E. Extended exergy based ecological accounting for the
transportation sector in China. Renew Sustain Energy Rev. Elsevier
 Dai J, Chen B, Sciubba E. Ecological accounting based on extended exergy: a
sustainability perspective. Env Resour Econ 2014;48:9826–33.
 Chen B, Dai J, Sciubba E. Ecological accounting for China based on extended
exergy. Renew Sustain Energy Rev. Elsevier; 2014.
 Dai J, Fath B, Chen B. Constructing a network of the social-economic
consumption system of China using extended exergy analysis. Renew
Sustain Energy Rev Elsevier 2012;16(7):4796–808.
 Dai J, Fath B, Chen B. Constructing a network of the social-economic
consumption system of China using extended exergy analysis. Renew
Sustain Energy Rev Elsevier 2012;16(7):4796–808.
 Ji X, Chen GQ. Exergy analysis of energy utilization in the transportation sector
in China. Energy Pol 2006;34(14):1709–19.
 Chen B, Chen GQ, Yang ZF. Exergy-based resource accounting for China. Ecol
 Chen B, Chen GQ. Modiﬁed ecological footprint accounting and analysis based
on embodied exergy — a case study of the Chinese society 1981–2001. Ecol
 Chen GQQ, Qi ZHH. Systems account of societal exergy utilization: China 2003.
Ecol Modell 2007 Nov;208(2–4):102–18.
 Wall G. Exergy conversion in the Japanese society. Energy 1990;15(5):435–44.
 Wall G, Sciubba E, Naso V. Exergy use in the Italian society. Energy
 Sciubba E. Beyond thermoeconomics? The concept of extended exergy
accounting and its application to the analysis and design of thermal
systems. Exergy Int J 2001 Jan;1(2):68–84.
 Ertesvag IS. Society exergy analysis : a comparison of different societies.
 Reistad G. Available energy conversion and utilization in the United States.
ASME Trans Ser J Eng Power 1975;97:429–34.
 Ma L, Allwood JM, Cullen JM, Li Z. The use of energy in China : tracing the ﬂow
of energy from primary source to demand drivers. Energy Elsevier Ltd
 Carnahan W, Ford KW, Prosperetti A, et al. Technical aspects of the more
efﬁcient utilization of energy: Chapter 2 – Second law efﬁciency: the role of
the second law of thermodynamics in assessing the efﬁciency of energy use.
Am Inst Phys Conf Ser 1975;25:25–51.
 Wall G. Exergy – A useful concept. PhD Thesis, Chalmers Univ Technol Univ
Göteborg; 1986 <www.exergy.se/ftp/ex77c.pd>.
 Ayres RU, Ayres LW, Warr B. Exergy, power and work in the US economy,
1900–1998. Energy 2003;28(3):219–73.
 Serrenho AGCH. Useful Work as an Energy End-Use Accounting Method:
Historical and Economic Transitions and European patterns. PhD Thesis, Univ
 Szargut J, Morris DR, Steward FR. Exergy analysis of thermal, chemical and
metallurgical processes. New York: Hemisphere; 1988.
 Amecke H, Deason J, Hobbs A, Novikova A, Xiu Y, Shengyuan Z. Buildings
energy efﬁciency in China, Germany, and the United States. San Francisco,
USA: Climate Policy Initiative; 2013.
 Murata A, Kondou Y, Hailin M, Weisheng Z. Electricity demand in the Chinese
urban household-sector. Appl Energy 2008;85:1113–25.
 Lawrence Berkley National Laboratory (LBNL). China Energy Databook v8.0;
 Pachauri S, Jiang L. The household energy transition in India and China. Energy
Pol 2008 Nov;36(11):4022–35.
 Catania P. China’s rural energy system and management. Appl Energy 1999
 Hao H, Wang H, Ouyang M. Fuel consumption and life cycle GHG emissions by
China’s on-road trucks: Future trends through 2050 and evaluation of
mitigation measures. Energy Pol Elsevier 2012;43:244–51.
 Huo H, He K, Wang M, Yao Z. Vehicle technologies, fuel-economy policies, and
fuel-consumption rates of Chinese vehicles. Energy Pol Elsevier 2012;43:
 He K, Huo H, Zhang Q, et al. Oil consumption and CO
emissions in China’s road
transport: current status, future trends, and policy implications. Energy Pol
 Qunren L, Yushi M. China’s transportation and its energy use. Energy Sustain
Dev Int Energy Initiat Inc. 2001;5(4):92–9.
 Wang M, Huo H, Johnson L, He D. Projection of Chinese motor vehicle growth,
oil demand, and CO
emissions through 2050. US: Energy Systems Division,
Argonne National Laboratory; 2006.
 Hasanbeigi A, Price L, Aden N. A comparison of iron and steel production
energy use and energy intensity in China and the U.S.; 2011 <http://china.
 He F, Zhang Q, Lei J, Fu W, Xu X. Energy efﬁciency and productivity change of
China’s iron and steel industry Accounting for undesirable outputs. Energy Pol
 Price L, Sinton J, Worrell E, Phylipsen D, Xiulian H, Ji L. Energy use and carbon
dioxide emissions from steel production in China. Energy 2002;27:429–46.
 Hasanbeigi A, Jiang Z, Price L. Analysis of the past and future trends of energy
use in key medium- and large-sized Chinese steel enterprises, 2000–2030.
Berkeley, USA; 2013 <http://china.lbl.gov/sites/all/ﬁles/steel_decom_analysis.
 Hasanbeigi A, Price L, Fino-chen C, Lu H, Ke J. Retrospective and prospective
decomposition analysis of Chinese manufacturing energy use, 1995-2020;
 Hao H, Wang H, Ouyang M. Fuel conservation and GHG (Greenhouse gas)
emissions mitigation scenarios for China’s passenger vehicle ﬂeet. Energy
Elsevier Ltd 2011;36(11):6520–8.
 Teter J. China transport efﬁciency data. UC Davis data sent by email to author;
902 P.E. Brockway et al. / Applied Energy 155 (2015) 892–903
 National Climatic Data Centre. National climatic data centre – climate data
online: dataset discovery; 2014 <http://www.ncdc.noaa.gov/cdo-web/datasets>.
 Department of Energy & Climate Change (DECC). Energy Consumption in the
UK: 2013; 2013 <http://www.decc.gov.uk/en/content/cms/statistics/
 Edwards RD, Smith KR, Zhang J, Ma Y. Implications of changes in household
stoves and fuel use in China. Energy Pol 2004;32:395–411.
 Ross M, Feng L. The energy efﬁciency of the steel industry of China. Energy
 Phylipsen D, Blok K, Worrell E, De BeerJ. Benchmarking the energy efﬁciency of
Dutch industry : an assessment of the expected effect on energy consumption
emissions. Energy Pol 2002;30:663–79.
 Saygin D, Patel M, Tam C, Gielen D. Chemical and petrochemical sector:
potential of best practice technology and other measures for improving energy
efﬁciency. IEA Inf Pap; 2009 <http://www.iea.org/publications/
 Fridley D, Zheng N, Zhou N, Ke J, Hasanbeigi A, Price L. China energy and
emissions paths to 2030 (2nd Edition). Lawrence Berkeley National Laboratory
 Brockway PE, Barrett JR, Foxon TJ, Steinberger JK. Supporting information –
divergence of trends in US and UK aggregate exergy efﬁciencies 1960–2010.
Environ Sci Technol 2014;48:S1–S39.
 Laux R, Werner B, Knight I, Djerf K. Modernising China’s labour market
statistics. Labour Mark Trends April 2003;2003:203–11.
 Brooks R, Tao R. China’s labor market performance and challenges.
International Monetary Fund (IMF) working paper WP/03/210; 2003.
 Wirsenius S. Human Use of Land and Organic materials: Modeling the
Turnover of Biomass in the Global Food System [Internet]. PhD Thesis,
Chalmers Univ Technol; 2000 <http://publications.lib.chalmers.se/
 Food and Agricultural Organisation of the United Nations (FAOSTAT). Food
supply kcal/day; 2013 <http://faostat3.fao.org/faostat-gateway/go/to/download/
 Smil V. Energy in world history. Boulder (Colorado, USA): Westview Press;
 Ramaswamy NS. Draught animals and welfare. Rev Sci Tech Off Int Epiz
 Down to Earth. 65 million draught animals; 2004 <http://www.downtoearth.
 Wilson RT. The environmental ecology of oxen used for draught power. Agric
Ecosyst Environ 2003;97:21–37.
 O’Neill DH, Kemp DC. A comparison of work outputs of draught oxen. J Agric
Engng Res 1989;43:33–44.
 Krausmann F, Erb K, Gingrich S, Lauk C, Haberl H. Global patterns of
socioeconomic biomass ﬂows in the year 2000: a comprehensive assessment
of supply, consumption and constraints. Ecol Econ 2007;65:471–87.
 Guan D, Liu Z, Geng Y, Lindner S, Hubacek K. The gigatonne gap in China’s
carbon dioxide inventories – supporting information. Nat Clim Chang 2012.
 Wang C, Chen J, Zou J. Decomposition of energy-related CO
emission in China:
1957–2000. Energy 2005;30(1):73–83.
 Xu X, Zhao T, Liu N, Kang J. Changes of energy-related GHG emissions in China:
an empirical analysis from sectoral perspective. Appl Energy Elsevier Ltd
 Liu L-C, Fan Y, Wu G, Wei Y-M. Using LMDI method to analyze the change of
China’s industrial CO
emissions from ﬁnal fuel use: an empirical analysis.
Energy Pol 2007 Nov;35(11):5892–900.
 Zhang M, Li H, Zhou M, Mu H. Decomposition analysis of energy consumption
in Chinese transportation sector. Appl Energy Elsevier Ltd
 World Bank. GDP (constant 2005 US$); 2014 <http://data.worldbank.org/
 World Bank, The Development Research Center of State Council the People’s
Republic of China. China 2030: Building a Modern, Harmonious, and
Creative Society; 2012 <http://www.worldbank.org/content/dam/Worldbank/
 Goldemberg J. Viewpoint leapfrog energy technologies. Energy Pol
 Nakicenovic N, Gilli PV, Kurz R. Regional and global exergy and energy
efﬁciencies. Energy 1996;21(3):223–37.
 Liddle B. Revisiting world energy intensity convergence for regional
differences. Appl Energy Elsevier Ltd 2010;87(10):3218–25.
 International Energy Agency (IEA). World energy outlook 2013. Paris, France;
 US Energy Information Administration (USEIA). World primary energy
consumption. Int Energy Outlook 2014; 2014 [cited 2014 Nov 5] <http://
 British Petroleum (BP) Ltd. World Energy Outlook 2035 – Excel Tables. World
Energy Outlook 2035; 2014 [cited 2014 Nov 5] <http://www.bp.com/en/
 IIASA. MESSAGE LIMITS-base (RCP 8.5) Total Primary Energy Consumption
data. LIMITS Scenar. Database (version 1.0.0); 2014 [cited 2014 Nov 5]
 Smil V. Perils of long-range energy forecasting : reﬂections on looking far
ahead. Technol Forecast Soc Change 2000;65:251–64.
 Ayres RU, Turton H, Casten T. Energy efﬁciency, sustainability and economic
growth. Energy 2007;32:634–48.
 Schipper L, Grubb M. On the rebound? Feedback between energy intensities
and energy uses in IEA countries. Energy Pol 2000;28(6–7):367–88.
 Ayres RU. Eco-thermodynamics: economics and the second law. Ecol Econ
P.E. Brockway et al. / Applied Energy 155 (2015) 892–903 903