The driving forces of China's energy use from 1992 to 2010:An empirical study of input–output and structuraldecomposition analysis

Shi-Chen XieGraduate School of International Development, Nagoya University, Nagoya 464-8601, Japan

H I G H L I G H T S

� Hybrid energy I–O models on both energy supply and demand side are developed.� Two energy flow charts are provided to describe China's energy system.� Analysis of structural change, and technology change effects on China's energy use.� China's current energy consumption is export and investment driven demand.� Policy implications of the structural decomposition analyses are discussed.

a r t i c l e i n f o

Article history:Received 12 February 2014Received in revised form21 April 2014Accepted 19 May 2014Available online 18 June 2014

Keywords:Structural decomposition analysisEnergy input–output modelEnergy flow chart

a b s t r a c t

The energy consumption in China has accelerated since the early 2000s, and China became the largestenergy consumer in the world by 2010. To examine the driving forces of China's energy use, this paperconducts a structural decomposition analysis based on hybrid input–output tables. In addition, wedescribe the framework of China's energy system by using two energy flow charts. The results show thatChina's current energy use is investment-led demand. Between 1992 and 2007, the three main final-demand categories – gross fixed capital formation, household consumption and exports – contributedapproximately one-third each to the changes of total energy use in China. Between 2007 and 2010,however, three-quarters of energy consumption changes came from investment activity only. Techno-logical improvement saved approximately five percent of the total energy use annually during theperiods of 1992–1997, 1997–2002 and 2007–2010. In the period of 2002–2007, however, its contributiondropped to only three percent p.a. due to the rise of the indirect energy requirement coefficient in theconstruction sector. These results suggest that adjusting the final demand structure and improvingenergy efficiency further will meet China's energy challenges in the future.

& 2014 Elsevier Ltd. All rights reserved.

1. Introduction

An economy requires a different input level of energy resourcesin different development phases. As Bernardini and Galli (1993)have identified, national energy intensity follows a bell shape orinverted U-shape pattern in the long term, which is defined as thetheory of dematerialization. In other words, an economy's totalenergy use typically grows faster than its gross domestic product(GDP) and eventually grows slower than the GDP during thecourse of economic development. Galli (1998) examined theenergy intensity trends for 10 Asian emerging countries during1973–1990 and found that energy intensity decreases beyondsome threshold level of GDP per capita. Medlock and Soligo

(2001) analyzed the effect of sector-specific energy use growthrates on final energy consumption and identified energy intensitypeaks at approximately $2600 (1985 PPP$).

The trends of energy1,2 intensity3 in China follow this invertedU-shape pattern (Fig. 1).4 Between 1949 and 1978, because of thepursuit of the Soviet style of industrialization in China, the share of

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/enpol

Energy Policy

http://dx.doi.org/10.1016/j.enpol.2014.05.0350301-4215/& 2014 Elsevier Ltd. All rights reserved.

E-mail address: shichen_xie@163.com

1 The term “energy” used in this paper refers to commercial energy only. Traditionalfuels, mainly biomass, are excluded from this analysis due to its data availability.

2 1 ton of coal equivalent (tce)¼0.7 ton of oil equivalent (toe).3 The term “energy intensity” is defined as the ratio of energy consumption in

its physical quantity over GDP in constant prices.4 The non-thermal electricity included here is not measured in calorific value

but divided by the thermal-electric conversion efficiency in the same year. If thenon-thermal electricity is not measured in calorific value in this paper, we will notethis explicitly here.

Energy Policy 73 (2014) 401–415

industrial economic output increased from 18 to 44 percent, andthe energy–GDP ratio almost tripled (Rosen and Houser, 2007).After the launch of the Reform and Opening-up policy in 1978, theenergy intensity in China began falling, while Chinese per capitaGDP remained at approximately hundreds of US$. During theperiod between 1978 and 2001, the energy–GDP ratio fell bytwo-thirds (68 percent), decreasing at an annual average rate offive percent (NBS, 2010a, 2011, 2012a). Zhang (2003) noted thatthis achievement is rarely accomplished in countries at this levelof development.

However, this declining energy intensity trend has reversedsince 2002. Over the period of 2002–2010, the energy–GDP ratiodecreased at an annual average rate of only two percent. Given itsunparalleled economic growth at average of 10 percent p.a. overthree decades, China became the largest energy consumer in theworld by 2010 (BP, 2013) even though its energy intensity droppedabout three-quarters during the same period. At the same time,there is also growing concern over the environment on both localand global scales associated with the increasing emissions fromfossil fuel combustion. It is critical to understand the driving forcesof China's energy use for the design of both Chinese and globalenergy and environmental policy.

Total energy use is the product of economic output and energyintensity. The energy intensity is determined by three fundamentalfactors: changes in the structure of final demand, increases in theefficiency of energy use, and the substitution of alternative inputs(Bernardini and Galli, 1993). To gauge the effects of driving forces onenergy uses change, index decomposition analysis (IDA) and structuraldecomposition analysis (SDA) are two widely used techniques byresearchers. The reviews of IDA can be found in Ang (2004) and Angand Zhang (2000), and of SDA can be found in Miller and Blair (2009)and Su and Ang (2012b). Empirical studies conducted by Huang(1993), Liao et al. (2007), Ma and Stern (2008), Sinton and Levine(1994), Zhang (2003), and Zhao et al. (2010) used IDA decomposedChina's industrial energy use change into production effect, structuraleffect, and technical effect.5 These studies found that technical effect isthe dominant contributor for the energy intensity decline during1980s and 1990s, and the excessive expansion of energy-intensivesectors caused China's energy intensity fluctuation after 2001.

While the IDA studies mentioned above only discuss thechange in industrial energy use, the SDA studies developed froman input–output table can describe the economy-wide energy use.Although IDA is applied to track economy-wide energy efficiencytrends (Ang, 2006; Ang et al., 2010), the differences between thescope of these two techniques in energy studies still exist. Forexample, typical SDA studies are able to provide more detailedfactors, such as a Leontief effect (or technical effect) and a final

demand effect by both sector and demand source, and to estimatethe indirect effect, such as the energy embodied in steel rebarwhich is used in construction. More similarities and differencesbetween SDA and IDA can be found in studies by Hoekstra and vanden Bergh (2003) and Su and Ang (2012b).

By using SDA and based on the input–output tables in 1981 and1987, Lin and Polenske (1995) examined primary energy use changesin China from 1981 to 1987. They found that production technologychanges reduced China's energy requirements, while final demandshifts increased energy use, and the driving force of the decline inenergy intensity was energy efficiency improvements. Based on theinput–output tables from 1987 and 1992, Garbaccio et al. (1999)decomposed the reduction in energy use into technical change andstructural change. Their main conclusion is that technical changewithin sectors accounted for most of the fall in the energy–outputratio during the period of 1987–1992, and structural change actuallyincreased the use of energy. Based on China's input–output tables in30 sectors in 1992, 1997, 2002 and 2004, Chai et al. (2009) suggest thatthe fluctuation of energy intensity is mainly due to technologicaladvances and the corresponding change in the industrial structure.

This paper applies SDA to analyze the driving forces of China'senergy use and its change between 1992 and 2010 – the years forwhich we have both detailed energy balance tables and input–output tables. It can be considered as an update of the study by Linand Polenske (1995). This study differs from the previous analysesin three important aspects. Firstly, hybrid I–O tables used in thisstudy are compiled based on the published statistics. Secondly, twoenergy flow charts are provided to describe China's energy system.Thirdly, we provide hybrid energy I–O models on both energysupply and demand side. These models allow us to capture bothdirect and indirect energy demands associated with final demand,and eliminate the energy price variations across all sectors and finaldemand categories. This paper decomposes total energy use and itschanges in China by using the energy demand side model.

The remainder of this paper is organized as follows. Section 2introduces the energy input–output model and the application ofLMDI in this model. The compilation of hybrid input–output tablesis described in Section 3. Section 4 describes China's energy flowcharts and the decomposition of total energy use and its change.Finally, Section 5 discusses conclusions and policy implications.

2. The energy input–output model and decomposition method

2.1. Energy input–output model

The normal input–output (I–O) model analyzes the flows ofproducts between industries in an economy. The I–O model isdeveloped from an inter-industry transaction table, called aninput–output table. China's I–O table shows imports within the

Gre

at L

eap

Forw

ard

Cul

tura

l Rev

olut

ion

Ope

ning

up

and

Ref

orm

Pol

icy

Den

g’s

Sou

ther

n To

ur s

peec

h

Ent

ers

WTO

0

100

200

300

400

500

600

Ene

rgy

inte

nsity

, tce

/mill

ion

RM

B(2

000P

)

1950 1960 1970 1980 1990 2000 2010

Year

1978

1992

2001

0

100

200

300

400

500

600

0 5 10 15 20

GDP per capita, thousand RMB (2000 price)

Fig. 1. Energy intensity trends in China, 1953–2011. Data source: NBS (2010a, 2011, 2012a).

5 The terms “efficiency effect” and “real intensity effect” are also used.

S.-C. Xie / Energy Policy 73 (2014) 401–415402

transaction table, which is assumed as competitive imports(commodities that are also produced domestically). The rows ofsuch tables describe the distribution of an industry's outputthroughout the economy, which can be summarized in matrixnotation as

ZiþFi�m¼ x ð1Þwhere Z is the n�n matrix of intermediate inputs; F is the n� omatrix of the total final demand; m and x are n� 1 (column)vectors of imports and total outputs, respectively; i is the n� 1vector of ones, which is known as a summation vector. The totalfinal demand includes domestic final demand and exports (foreigndemand)

F¼ FdþFEx ð2Þwhere Fd is the n� omatrix of the domestic final demand in whichthe elements of exports are zeros; FEx6 is the n� o matrix ofexports in which the elements of the domestic final demandare zeros.

An economy's total energy use consists of all the energyembodied in domestically produced goods and services, whileenergy embodied in imported goods and services is the energydemand in foreign countries. Thus, to measure the impacts ofdomestic production on total energy use, we need to divide thetransaction flows of a competitive import type I–O table intodomestic products and imported products (Table 1). The rows ofeach part can be expressed as

ZdiþFdi¼ x ð3Þ

ZmiþFmi¼m ð4Þwhere Zd ¼ ðI�MÞZ and Zm ¼M � Z are the n�n matrices ofintermediate inputs of domestic and imported products, respec-tively; Fd ¼ ðI�MÞFdþFEx is the n� o matrix of the total finaldemand for domestic products; Fm ¼M � Fd is the n� o matrix ofthe total final demand for imported products; M¼ m̂ð dZiþFdiÞ�1 isthe n�n diagonal matrix of the import ratio based on theassumption that imported products would not be exporteddirectly and distributed at the same percent of total inputs to allintermediate sectors and domestic final demand categories. Millerand Blair (2009) and Su and Ang (2013) provide more detaileddiscussions on this non-competitive imports' assumption.

Using the definitions of the technical coefficient matrix fordomestic products, Ad ¼ Zdx̂ �1, and for imported products,Am ¼ Zmx̂ �1, Eqs. (3) and (4) can be rewritten as

AdxþFdi¼ x ð5Þ

AmxþFmi¼m ð6ÞThe summation matrix A¼ AdþAm is the direct input coefficientin the import exogenous I–O model. The solution to Eq. (5) is givenby

x¼ ðI�AdÞ�1Fdi¼ LdFdi; ð7Þwhere Ld ¼ ðI�AdÞ�1 is the n�n matrix of the Leontief inverse ortotal input requirements for domestic products. Eqs. (3)–(7) areknown as the non-competitive import type I–O model or importendogenous I–O model.

To extend the general I–O framework to energy studies, mostempirical studies applied to China simply multiply the total output bythe energy coefficient. The discussion of the strengths and weakness ofthis approach can be found in the studies by Dietzenbacher and Stage

(2006) and Miller and Blair (2009). In this study, we use hybrid I–Otables where the monetary energy flows are replaced by physicalenergy flows.

In the hybrid I–O table, the rows of energy-related sectorspresent the distribution of energy products throughout economy,which can be summarized similar to Eqs. (5) and (6) as

ε̂Adxþ ε̂Fdi¼ ε̂x ð8Þ

ε̂Amxþ ε̂Fmi¼ ε̂m ð9Þwhere ε is the n� 1 vector consisting of ones and zeros, elementsof which are ones corresponding to energy sectors and zeros for allthe other sectors. The diagonal matrix, ε̂, is used to select rowsof energy sectors from the hybrid I–O table. For example, ε̂Adx inEq. (8) represents the intermediate inputs of domestic energy.In hybrid I–O Eqs. (8) and (9), the units for each element in thematrices of Fd, Fm, x and m are expressed as ½ tceRMB�, and for Ad andAm are ½ tce=tceRMB=tce

tce=RMBRMB=RMB�. The solution to Eq. (8) is given by

ε̂x¼ ε̂ðI�AdÞ�1Fdi¼ ε̂LdFdi: ð10ÞSimilar to identity (6), this identity explains the total energyproduction via the product of total requirement matrix for domesticenergy, ε̂Ld, and the total final demand for domestic products, Fdi.

The total energy use of an economy can be measured fromeither the energy supply side or the energy demand side, accord-ing to the energy balance table. From the energy supply side, thetotal primary energy supply is the summation of primary energyproduction and other energy supply sources, such as imports, etc.The information of primary energy production, however, is notavailable from the hybrid I–O table directly. The reason is that theprimary supply of electricity (hydro power, nuclear power, andwind power) is counted together with thermal power. The primarysupplies of heat and coke (recovery energy7) have similar pro-blems. Lin and Polenske (1995) solve this problem by introducinga hypothetical sector into the I–O model. In this paper, we bring ann� 1 vector, α, to separate primary electricity from thermalelectricity. The elements in the vector α for secondary energysectors are ratios of primary sources over total energy outputs, andfor all the other sectors are ones.

Therefore, the total primary energy supply (TPES) is given by

eTPES ¼ α̂exþem�eEx�eΔS ð11Þwhere eTPES is the m� 1 vector of TPES; α̂ex , em, eEx and eΔS arethe vectors of the total primary energy production, importedenergy, exported energy and energy stock changes respectively.By substituting Eqs. (9) and (10) into (11) and rearranging them,we have

eTPES ¼ ε̂α̂LdFdiþðε̂AmLdFdiþ ε̂FmiÞþ ^εFu ð12Þwhere ε̂α̂LdFdi is the total primary energy production; ε̂AmLdFdiis the intermediate input of imported energy; ε̂Fmi is the final

Table 1Non-competitive import type input–output table.

Economicactivities

Intermediateuse

Final demand Total output/Import

Intermediate input Zd Fd ¼ ðI�MÞFdþFEx x

Import Zm Fm ¼M � Fd mValue added VTotal input x0

6 In China's I–O table, there is an “error” column in the final demand block. Itwas not shown in the equation because it represents the statistical differencebetween inputs and outputs only. We treat it the same as exports.

7 The recovery energy is excluded from the total primary energy supply in thelatest version of China's energy statistics. We still count it as a primary energysupply in this paper.

S.-C. Xie / Energy Policy 73 (2014) 401–415 403

demand for imported energy; ^εFu is the energy delivered toexports and stock changes; u is the o� 1 vector consisting ofminus ones for exports and stock changes and zeros for all otherfinal demand categories.

From the energy demand side, the total energy use equals thesummation of energy consumed in intermediate sectors and inhouseholds. In the intermediate input matrix, energy inputs totransformation sectors are not completely consumed but areconverted into secondary energy, which appears again in the rowsof secondary energy sectors. We introduce an n� 1 vector, β, toavoid this double accounting, elements of which are the percen-tage of secondary energy outputs over total energy inputs tocorresponding transformation sectors and zeros for all the othersectors. The vector, β, can be defined as the net energy transfor-mation efficiency. It provides a net energy output ratio for theenergy transformation sector, the figures of which are slightlysmaller than normal energy transformation efficiency.

Using the definition of net energy transformation efficiencyvector, β, the total energy use can be expressed as

eTEU ¼ EIðI� β̂ÞiþeH ð13Þ

where eTEU and eH are the m� 1 vectors of the total energy useand household energy use respectively; EI ¼ ε̂Ax is the m�nmatrix of intermediate inputs of energy, and EIðI� β̂Þ refers tointermediate energy use that eliminates the double accounting. Bycombining Eqs. (10) and (13), we have

eTEU ¼ ε̂AðI� β̂ÞLdFdiþ ε̂Fv¼ ε̂GFdiþ ε̂Fv ð14Þ

where ε̂AðI� β̂Þ is the m�n matrix of direct energy use coeffi-cients; ε̂G¼ ε̂AðI� β̂ÞLd is the m� n matrix of total energy usecoefficients; ε̂GFdi represents the intermediate energy use; ε̂Fv isthe energy delivered to household; v is the o� 1 vector consistingof ones for households and zeros for all the other final demandcategories.

The Leontief inverse matrix can be expressed as Ld ¼ ðI�AdÞ�1

¼ IþAdþðAdÞ2þðAdÞ3þ⋯ (Miller and Blair, 2009). Thus, theintermediate energy use in Eq. (14) can be expressed as asummation of direct energy use, ε̂AðI� β̂ÞðIÞFd, and indirect energyuse, ε̂AðI� β̂ÞðLd�IÞFd. It is also a summation of the intermediateuse of domestic energy, ε̂AdðI� β̂ÞLdFd, and of imported energy,ε̂AmðI� β̂ÞLdFd. In Eq. (14), the household energy use, ε̂Fv, can alsobe expressed as a sum of domestic energy, ε̂Fdv, and of importedenergy, ε̂Fmv.

The difference between the total primary energy supply andthe total energy use is the “statistical difference” in the energybalance table. Both these concepts describe the total energyconsumption in an economy. Lin and Polenske (1995) decomposedChina's energy use from the energy supply side, so that this paperchooses the energy demand side model given by Eq. (14) for thefollowing analysis. Note that the matrices' dimensions, n, m and o,refer to the number of sectors, energy sectors and final demandcategories, respectively.

2.2. Decomposition method

Final demand can be further decomposed into three compo-nents: (1) the overall level of final demand; (2) the distribution oftotal expenditures across final demand categories, such as house-hold consumption and exports; and (3) the product mix of eachindividual final demand category (Lin and Polenske, 1995; Millerand Blair, 2009). In this paper, we added population as anothercomponent so that the final demand can be expressed as

Fd ¼ Bdd̂dydp and F¼ Bd̂yp ð15Þ

where B refers to the product mix effect; d refers to the distribu-tion effect; y refers to the per capita level effect; p refers to thepopulation effect; and the superscript d denotes the effect asso-ciated with domestic products.

Finally, the total energy use in Eq. (14) can be express as

eTEU ¼ ε̂GBdd̂dydpiþ ε̂Bd̂ypv ð16Þ

Based on this equation, the total energy use change can bedecomposed into the parts associated with changes in technology(G) and related to changes in final demand, which can be furtherbroken down into population change (p), the per capita level effect(y and yd), the distribution effect (d̂ and d̂

d) and the product mix

effect (B and Bd)

ΔeTEU ¼ΔeGþΔeBþΔedþΔeyþΔep ð17ÞThere are many ideal decomposition methods reported in

pioneer studies. According to the guidelines for decompositionmethod selection proposed by Ang (2004) and Su and Ang(2012b), the logarithmic mean Divisia index method (LMDI)8 isadopted to decompose the driving forces of energy use. LMDI wasfirst introduced by Ang and Choi (1997) and first applied to energystudy by Wachsmann et al. (2009). It provides an ideal decom-position and is time-reversal invariant and easy to use. The effectsdescribed in Eq. (17) can be estimated by

Δev ¼ ∑m;n;o

eDvΔln v ð18Þ

where v refers to factors G, B, d, y and p; eDv ¼ LðeTv ; e0vÞ9 is the

logarithmic mean function. The detailed formulations are summar-ized in Appendix A. The zero and negative value problems in LMDIare resolved with the “analytical limits” (AL) strategy recommendedby Ang and Liu (2007a,b) and Wood and Lenzen (2006).

3. The data

The main data requirement for this analysis is hybrid I–Otables, where monetary energy flows in the original I–O tableare replaced by physical energy flows. The hybrid I–O tableintegrates the I–O table with the energy balance table, both ofwhich are available from public statistics.

The original I–O tables used in this study are competitiveimport type tables, which have been compiled by the ChineseNational Bureau of Statistics (NBS) every five years since 1987. Weuse five Chinese I–O tables for the years 1992, 1997, 2002, 2007and 2010 (NBS, 2012d, 2009, 2006, 1999, 1996); the first four ofthem are benchmark I–O tables, and the last one is updated from2007 by NBS. The analysis of energy use changes over timerequires a constant price for these I–O tables. We use the priceindices published by the NBS (2012b, 2012c) and the methodssuggested by Liu and Peng (2010) and Su and Ang (2012b) todeflate the I–O tables to constant 2000 prices. The nationalproduction activities are divided into 28 sectors as Appendix Bshows, six of which are energy-related industries.10 Final

8 We adopt the advice from Su Bin and Zhang You Guo on explaining the LMDIdecomposition method.

9 Lða; bÞ ¼ ða;bÞ=ðln a� ln bÞ for aab; Lða; bÞ ¼ a for a¼ b.10 The I–O tables for 2002, 2007 and 2010 count “extraction of crude petroleum

and natural gas” in one sector because petroleum and natural gas are byproducts.We split inputs (column data) in this sector into two sectors based on theproduction share of crude petroleum and natural gas. The outputs (row data) tothis sector are replaced by physical energy flows available from energy statistics. InChina's I–O table, there are not direct natural gas flows to households but via the“gas supply and production” sector. In a hybrid I–O table, natural gas directly flowsto all other sectors, and “gas supply and production” covers coke gas only. The I–O

S.-C. Xie / Energy Policy 73 (2014) 401–415404

demands, which are sourced from seven categories, and imports11

are shown in Appendix B.Physical energy flows are reported in the energy balance table.

The columns in this table describe the balance in the sources ofsupply for each type of energy and its uses. Similar to the I–O table,five energy balance tables for 1992, 1997, 2002, 2007 and 2010 areused in this paper (NBS, 1998, 2010b, 2011). To combine the energybalance table with the I–O table, we aggregate energy types into sixcategories that correspond to the energy production sectors shownin Appendix B. Making a transposition to this table, the rows of thenew table present physical energy flows from the supply side to thedemand side, which can be presented as an energy flow chart (seeAppendix C). In the hybrid I–O table, intermediate energy use coversenergy flows to six energy sectors12 and 22 non-energy sectors;energy consumption in the final demand sectors13 includes energydeliveries to residences, stock exchanges, exports, statistical differ-ences and imports. The relationship between the hybrid energy I–Otable and the energy balance table is given in Appendix B.

Dietzenbacher and Stage (2006) identified that the choice ofmonetary and energy units would affect the estimation of leveland mix effects when using the hybrid I–O table. In order toovercome this problem, their proposed solution is based on usinga sum of monetary units instead, which requires full informationon the prices paid for the final demand energy. In this paper, weexamine the impacts from the choosing different energy units onthe decomposed effects (Fig. 2). We found that the three effects –

per capita level effect, distribution effect and mix effect – aredetermined by the choosing of energy units. The shares of theseeffects in the change of total energy use are following a sigmoidfunction when energy unit shifting from 1Eþ3 to 1Eþ9 tce andmonetary unit fixes on billion RMB. Most of these curves aresymmetrical about the energy unit of 1Eþ6 tce, except the periodof 1992–1997 which might be caused by the inconsistent of energystatistics before and after 1996. The median values avoid theimpacts from units on the decomposed effects. Thus, we choose“million tce” as the energy unit and “billion RMB” as the monetaryunit in the following analysis.

4. Results and discussions

4.1. China's energy flow chart

Before explaining the results of the SDA, let us trace the energyflows throughout the Chinese economy via an energy flow chart.This chart is typically used to visualize energy flows betweensectors. It provides a complete picture of an energy system. China'senergy flow charts for 1992 and 2010 are provided in Appendix C.They are developed based on the pattern of the UK energy flowchart to show Chinese energy statistics (DECC, 2013; Xie et al.,2009). The widths of the lines in China's energy flow chart areshown proportionally to the flow quantity and the colors fordifferent energy categories.

The sources of energy supply are displayed on the left-handside of the energy flow chart (see Appendix C). Between 1992 and2010, driven by rapid economic growth, the total Chinese primaryenergy supply increased nearly three-fold (from 1056 million tcein 1992 to 3074 million tce in 2010), at six percent p.a. (Fig. 4).Obviously, coal is the main energy source for China. In 2010, theChinese economy relied on coal for two-thirds (68 percent) of itsTPES, petroleum for one-fifth (19 percent), natural gas for fourpercent and primary electricity for another nine percent. This coal-based energy supply system has stabilized for more than threedecades since the 1970s. During the two decades before the 1970s,the increasing domestic petroleum production reduced China'scoal dependence. However, with limited petroleum reserves,domestic production was unable to support its continually increas-ing demand. China now relies on international markets for morethan half (53 percent in 2010) of its petroleum consumption, whileit remained an oil net exporter between 1970 and 1992 (Fig. 3).Similar to petroleum, with the substitution of coke gas by naturalgas and inadequate domestic production, the gas supply isbecoming more import dependent in recent years.

The right-hand side of the energy flow chart describes theenergy flows to transformation sectors (yellow rectangles) and tonon-energy industries and residences (see Appendix C). Between1992 and 2010, the proportion of the direct use of coal decreasedfrom more than one-half (57 percent in 1992) to one-quarter (28percent in 2010), while the indirect use of coal for electricitygeneration increased from one-third (30 percent) to one-half (51percent). The electricity generated from fossil fuels (mainly coal,97 percent in 2010) accounted for 80 percent of the total electricitysupply, hydropower for 20–15 percent, nuclear, wind and othersfor the remainder. Driven by economic development and electri-fication in both the residential and industrial sectors, the totalpower generation grew at an average rate of 10 percent p.a., from

y

dB

yd

BydB

y

d

B

−4

−2

0

2

4

1E+4 1E+6 1E+8 1E+4 1E+6 1E+8 1E+4 1E+6 1E+8 1E+4 1E+6 1E+8

1992~1997 1997~2002 2002~2007 2007~2010

y d B median value

The

shar

e of

effe

cts

in to

tal c

hang

e

Energy Units, * tce

Fig. 2. The share of effects in total energy use changes and energy units in China,1992–2010.

0%

20%

40%

60%

80%

100%

120%

140%

160%

1970 1990 1970 1990 1970 1990 1970 1990

Coal Petroleum Natural Gas Primary Electricity

Domestic production ratio Energy structure

Fig. 3. Energy supply structure and domestic production ratio by fuel type, 1953–2010. Data source: NBS (2010a, 2012a).

(footnote continued)table for 2010 places coking and refined petroleum products in the same sector, butwe separate it based on the figures in benchmark I–O table of 2007.

11 Because the benchmark I–O table for 1992 reports net export data only, weuse the ratio data from Li and Xue (1998) to split them into exports and imports.Thanks Youguo Zhang for providing these data.

12 Energy inputs to energy sectors are the sum of transformation inputs, energysector own use and distribution losses.

13 Direct energy use by the government is reported in the service sectoraccording to NBS.

S.-C. Xie / Energy Policy 73 (2014) 401–415 405

754 terawatt-hours (TWh) in 1992 to 4207 TWh in 2010. Theinstalled power capacity rose at an average rate of 10 percent p.a., from 167 GW in 1992 to 966 GW in 2010. In the short term,power generation is highly correlated with the business cycle, andit slowed when financial crises occurred in 1997 and 2008 (Fig. 4).

Finally, the secondary energy production from transformationsectors and the direct use of primary energy flow into non-energyindustrial and residential sectors (see Appendix C). Between 1992and 2010, the most significant change in the final energy demandwas the substitution of coal by electricity in all sectors. Theproportion of coal in residence, for example, fell from 83 percentin 1992 to 32 percent in 2010, while that of electricity rose from6 percent to 29 percent. Industrial sectors are always the largestenergy consumers, accounting for over 60 percent of the total finaluse. The energy used for transportation activity grew slightlyfaster than any other sector, but still accounted for only 17 percentof the total final use in 2010. The energy growth for transportationwill be one of the main energy and environmental challenges forChina in the future.

4.2. Decomposition of total energy use

The energy flow chart provides a qualitative view of China'senergy system. The energy I–O model proposed by Eq. (14) allows usto quantitatively analyze China's energy use from the energy-demand side. In the energy I–Omodel, the total energy requirementsare divided into household use and intermediate use. Intermediateenergy use, ε̂GFd, measures the energy (in six energy types)embodied in domestically produced goods and services (from 28sectors) that are consumed by seven final demand categories. It canbe further broken down into direct use and indirect use or intodomestic energy and imported if needed. The results are summarizedin Table 2, in which the total energy use by final demand categories,by energy types and by sectors is displayed. It should be noted thatthe energy figures displayed in Table 2 are not from the energysupply side but from the energy demand side, so that the total coaluse is not equal to its primary supply.

The total energy use in terms of household consumption covershousehold use and energy embodied in the products that consumedby households. It grew from 447 million tce in 1992 to 931 milliontce in 2010, but its contribution to the total energy use fell from 42percent in 1992 to 30 percent in 2010. This decreasing percentage ismainly due to a much faster growth in other final demandcategories. Among the total energy use in household consumption,direct energy use, such as gas or coal for cooking, electricity for airconditioners, and gasoline for automobiles, accounts for only one-quarter, while more than three-quarters are embodied in the goodsand services that are consumed by the household.

The energy embodied in the goods and services due to house-hold consumption doubled between 1992 and 2010, 70 percent ofwhich is attributable to indirect use, such as energy embodied inthe materials employed to produce automobiles. With increasinghousehold income, the composition of Chinese household expen-ditures shifted to energy-intensive products. Private expenditureson durable goods or energy-intensive goods, such as air condi-tioners and automobiles, have pushed household intermediateenergy use growth but in small quantities. The increasing durablegoods held by households caused the rising household use ofsecondary energy, such as electricity and petroleum products,which need more energy to produce them in energy industries.

Increasingly, rural populations have migrated to urban areasdue to urbanization, which is a result of modernization andindustrialization. By the end of 2010, urban dwellers accountedfor 50 percent of the total population for the first time in China'shistory (NBS, 2012a). The gap of energy use between urban andrural households is getting wider. Compared to the relatively flatenergy consumption in rural households, the energy use in urbanhouseholds tripled between 1992 and 2010. The recent CentralEconomic Working Conference underscored China's commitmentto urbanization (Xin Hua News, 2013), which will boost domesticdemand, thus advancing household energy use in the near future.

The energy use embodied in government consumption expen-ditures was nearly kept constant during 1992–2010. Its share inthe total energy use dropped to only four percent in 2010 fromeight percent in 1992. In China's statistics, the direct energy use bythe government is reported in the service sector, such that only theintermediate energy use by the government is shown in Table 2.

Investment activities – gross fixed capital formation andchanges in inventories – are the main sources of China's currentenergy requirements. As Table 2 shows, the energy embodied ingross fixed capital formation accounts for 39 percent of the totalenergy use by 2010. It increased four-fold from 296 million tce in1992 to 1204 million tce in 2010, the highest growth rate in allfinal demand categories. The investment-led energy consumptionboomed after 2002 when China became a member of WTO. Thechanges in inventories were responsible for less than five percentof total energy consumption. The increasing energy embodied ininventory after 2007 would provide evidence of the overexpansionof investment in heavy industries, such as basic metals.

The booms of investment created a surge in industrial activity.As Table 2 shows, the energy use for gross fixed capital formationmainly comes from construction (accounting for two-thirds) andequipment (accounting for one-fourth), two aggregated sectors. Itshould be noted that the energy required by the constructionsector is not only direct energy use for construction activity itselfbut also indirect energy use for the material inputs to construction,such as the energy use in the production process of steel andcement, among others. The share of indirect energy use reached 93percent of the total use due to investment activity (Table 2), thehighest in all final demand categories.

Exports – the demand from foreign countries – are the last finaldemand category that significantly affects the total energy demandin China. Since the 1990s, China has deeply participated in theinternational division to obtain a comparative advantage, such ascheap labor. Its total trade volume has experienced rapid expansion,at an average growth rate of 17 percent p.a. in the last three decades,and accelerating to 22 percent p.a. after 200214 (NBS, 2010a, 2012a).China's trade surplus expanded over the periods of 1994–1998 and2003–2008 as a result of increasing foreign demand. It is interesting

’84

’90

’92

’99

’07

’85

’90

’95

’98

’04

’08

’90

’92

’98

’03

’08

’88

’02

’06

0%

5%

10%

15%

20%

1990 2000 1990 2000 1990 2000 1990 2000

GDP TPES Generation Capacity

Ann

ual g

row

th ra

te

Fig. 4. Annual growth rates of GDP, TPES, power generation and capacity. Datasource: CEP (2012), NBS (2011, 2012a).

14 The growth rate is estimated based on trade in US dollars withoutconsidering the change in the foreign exchange rate.

S.-C. Xie / Energy Policy 73 (2014) 401–415406

Table 2Total energy use by final demand category, energy type and sector in China, 1992–2010.

S.-C. Xie / Energy Policy 73 (2014) 401–415 407

that the trade surplus soared after 2005, when RMB shifted to amanaged floating exchange rate.

With exports expanding, the energy embodied in exportedgoods and services almost quadrupled during the period 1992–2010, slightly slower than the gross fixed capital formation only.During the period between 2002 and 2007, export-related energyuse grew from 356 million tce to 800 million tce, doubling withinfive years. At the beginning, exports expanded mainly in lightindustries, such as textiles and apparel, and later shifting to heavyindustries, such as machinery and equipment. Exports now con-tribute to about one-quarter of the total energy use (24 percent in2010). The “errors” in Table 2 represent the statistical difference inthe I–O table.

As discussed above, China's total energy use is embodied in thegoods and services from the final demand categories of consump-tion, investment and export. From the import aspect, China avoidsdomestic energy use by imported goods and services. Under theassumption that imported goods and services are produced withthe same technology as domestic goods, the energy avoided byimports (EAI) can be estimated by the total energy use coefficientgiven in Eq. (14). This EAI is a biased measure of embodied energyin imports, considering the difference of energy intensity betweendomestic and imported goods and services (Su and Ang, 2013;Weber et al., 2008). According to the calculation, energy avoidedby imports increased from 219 million tce in 1992 to 674 milliontce in 2010 (Table 3). Imported goods and services mainly includeraw materials and high-technology equipment, such as “Chemicalsand chemical products”, “Basic metals”, “Machinery and equip-ment” and “Electric and telecommunication Equipment”. Thenet effects of international trade activity on China's total energyuse have climbed to 10 percent in 2007 from nearly balancedbefore 2002.

4.3. Decomposition of energy use change

The differences of energy use in sub-periods provide theadditive decomposition of total energy use changes by finaldemand category, energy type and sector, the result of which isdisplayed in Table 4. Before 2002, China's total energy useincreased at an almost uniform average annual growth rate15 offour percent and three percent in the periods of 1992–1997 and1997–2002, respectively. After 2002, however, it increased to 12percent p.a. between 2002 and 2007 and slowed down to fivepercent p.a. during 2007–2010. The year 2002 is the break pointfor the trends of energy requirements in China.

Similar to the total energy use indicated in Section 4.2, thechange in energy use in China mainly came from shifts in grossfixed capital formation, exports and household expenditures(mainly in urban areas). Before 2007, these three final demand

categories each contributed to approximately one-third of the totalenergy use changes in China. After 2007, however, gross fixedcapital formation shifts increased the total energy use by 302million tce and accounted for up to 84 percent of the total change,while the energy embodied in exports decreased by 65 million tce.This is a result of the government's four trillion-RMB stimulus thatwas expected to offset the shrinking foreign demand caused by theglobal financial crisis in 2008 (Xin Hua News, 2008).

China's total energy use change is mainly due to indirect energyuse. The “Gross fixed capital formation”, for example, contributedto most of the total energy change, and over 90 percent of it camefrom indirect energy use, such as energy used to produce rebar.Before 2002, most energy increases in “Gross fixed capital forma-tion” came from indirect energy use in construction. Since 2002,indirect energy use in machinery and equipment has becomeanother major factor.

As mentioned above, the total energy use change can bedecomposed into 6� 28� 7 entries based on the energy I–Omodel, each of which is a product of technology coefficients (G)(not included in household direct energy use change), product mixeffects (B), distribution effects (d), final demand per capita leveleffects (y) and population effects (p). By using LMDI and theenergy I–O model allows us to evaluate the impacts of the shift ineach effect on the total energy use change. Table 5 displays theresults of the SDA on China's energy use change by effects, byenergy type and by sector.

The final demand changes16 are the contributors that pushedthe total growth in energy use. Keeping all other factors constant,the final demand changes, such as the level of economic activitiesor shifts toward more energy-intensive demands, increase thetotal energy use by nine percent, nine percent, 15 percent and 10percent p.a. in the periods of 1992–1997, 1997–2002, 2002–2007and 2007–2010, respectively. The impact of the final demandeffect on energy requirement surged from 2002 to 2007, the firstfive years after China entered the WTO.

This rising pressure on energy demand associated with finaldemand shifts, however, was partly offset by changes in productiontechnology, which reduced the total energy requirement per unit ofgoods and services. Holding all other factors unchanged, theimprovement of production technology decreased the energy useat changeless annual rates of five percent, six percent, three percentand five percent in the periods of 1992–1997, 1997–2002, 2002–2007 and 2007–2010, respectively. The driving force of these energysavings was the improvement in the total energy use coefficients,especially for the final demands of energy-intensive goods andservices from sectors such as “Construction”, “Transport equipment”,“Machinery and equipment”, “Electric machinery and equipment”and “Chemicals and chemical products”.

Table 3Energy avoided by imports (EAI) in China, 1992–2010.

15 The average annual growth rate is defined as the geometric mean of percentchange.

16 Here, the sub-level effects of the final demand for domestic products, Fd , andof the final demand F are aggregated together, as Eq. (17) shows.

S.-C. Xie / Energy Policy 73 (2014) 401–415408

Between 2002 and 2007, the technological change in theconstruction sector abnormally increased the total energy use by10 million tce, which caused the technology change effect to slowdown to only three percent p.a. It should be highlighted again thatthe total energy savings are due to technological change not onlyfrom the final demand sector itself but also from those upstreamsectors. Steel rebar, for example, which is produced by the sectorof “Basic metals”, is finally consumed in “construction”, so thatenergy savings in “construction” cover the technological improve-ments in itself and also in “Basic metals”.

The impact of final demand changes on the total energyrequirements can be further decomposed into four different

dimensions: the population effect (p), level per capita effect (y),distribution effect (d) and products mix effect (B). These effects areintercepts with each other and provide unique insights into thechange in energy demand in China. An increasing populationrequires increased energy consumption. Between 1992 and 2010,the change in population caused around one percent of the totalenergy use growth annually, but this figure was in the process ofbeing reduced. The per capita level effect caused almost all energyuse growth that resulted from final demand shifts, which ledChina's total energy use growth at nine percent, nine percent, 15percent and nine percent annually in the periods of 1992–1997,1997–2002, 2002–2007 and 2007–2010, respectively. The

Table 4Changes of energy use by final demand category, energy type and sector in China, 1992–2010.

S.-C. Xie / Energy Policy 73 (2014) 401–415 409

distribution effect has a minor but positive contribution on China'senergy demand, which means the final demand has been shiftingto energy intensive categories such as investment. In the periodsof 1997–2002 and 2007–2010, especially, the distribution effectrose due to government fiscal stimulus that was expected to offsetthe weak export caused by financial crises in 1997 and 2008.The products mix effect has always cut the energy requirementsin China, since the product mix in each of the final demandcategories is shifting to a lower energy intensive combination.

4.4. Limitations of this study

This study is based on the energy I–O model proposed in Eq.(14), which assumes that a sector uses energy inputs in fixedproportions – the energy technical coefficient. As Miller and Blair(2009) noted, a Leontief production function is under the condi-tion of constant returns to scale. This paper ignores the energy usechange caused by shifts in the economies of production scale andthe substitution of energy with other inputs such as labor orcapital. Lin and Polenske (1995) suggest that the SDA shouldcomplement other types of analysis, such as econometric analysis.

This paper also ignores the effects of price change on totalenergy use in China. Since 2000, the international price ofpetroleum has risen from approximately 20 USD/bbl to currentlyapproximately 100 USD/bbl. In the past three years, the interna-tional petroleum price (Brent) fluctuated around 100–120 USD/bbl, even though the commodity index (CCI) dropped by one-quarter. Given the negative real interest rate in most majoreconomies, a higher oil price in the near future can be expected.Policymakers and consumers should pay attention to these risingoil price trends.

SDA requires a large amount of data; thus, the reliability of thestatistics is crucial to the accuracy of the results. NBS began topublish China's energy balance table since 1980. The reliability ofChina's energy statistics (mainly coal) in the late 1990s and early2000s has been questioned by many researchers (Sinton andFridley, 2002; Sinton, 2001). In 1998, China started to shut downsmall coalmines in order to cut the coalmine death toll andnumber of accidents, while many of those small coalminesreopened after this campaign. NBS, however, did not report thispart of coal production in its energy statistics. After China's secondnational economic census, NBS revised the energy statistics for theyears between 1996 and 2007. Thus, there are changes in thedefinitions and coverage of energy data before and after 1996.These inconsistent energy statistics might cause the household useof coal to drop abnormally by 49 million tce from 1992 to 1997(Table 2).

Non-competitive import type I–O tables are required in thisstudy. However, only competitive type I–O tables are availablefrom the NBS. In Section 2, we assume a fixed import ratio foreach sector and non-direct export for imports when makingnon-competitive I–O tables, and a constant price index for bothdomestic and imported products when compiling I–O tables inconstant prices. In addition, exports reported in China's I–O tablecovers not only normal exports but also processing/assemblingfee. In the I–O table of 2007, for example, over 2 percent of totalexports are processing fee. Different with normal exports, theprocessing fee does not cover any value of material. The energyembodied in the processing fee is very small, since it covers onlydirect energy use for processing/assembling. Thus, there is a slightoverestimation in the energy embodied in exports in this analysis.Lau (2010) and Su and Ang (2013) provide detailed discussion on

Table 5Changes in energy use by effects, energy type and sector in China, 1992–2010.

S.-C. Xie / Energy Policy 73 (2014) 401–415410

this topic. Although inaccuracies exist in China's statistics, thesetables remain a starting point for this study.

The hybrid I–O tables used in this analysis are aggregated into 28sectors, six of which are energy related sectors. The sector aggrega-tion level affects the SDA results in some degree (Lin and Polenske,1995; Su and Ang, 2012a; Su et al., 2010; Weber, 2009). In general, afiner sector classification provides more detailed decompositionanalysis on China's energy use changes. However, that detail levelof energy use, I–O and price index data is not available in every yearand not compatible with each other. Here, we adopt a Monte Carlosimulation to examine the relationship between SDA results andsector numbers (Fig. 5). This simulation is similar to the one designedby Weber (2009). The only modification we made is that therandomly chosen sector aggregated with a sector directly before orafter it, but not only directly after it. The effects that display in Fig. 5are standardized values; and the sector number aggregated from 28down to 7 considering only non-energy industries. As the simulationresults shown in Fig. 5, the linear fitted value of technology changeeffects (G) are decreasing with the increasing of sector disaggregationin the first two sub-periods, and keeping almost constant during theperiods of 2002–2007 and 2007–2010. While the distribution effects(d) are becoming more important with the increasing of sectornumbers. And the variances of the SDA results are reduced with theincreasing of disaggregation level, which might determine by thepossibilities of sector combinations. Let us keep these limitations anduncertainties of SDA in mind.

5. Conclusions and policy implications

From the demand side (the rows of the I–O table), increasing finaldemand from private consumption, investment and exports are themain factors that drive economic growth. The demand for energy isderived from the final demand for goods and services. The energyI–Omodel allows us to investigate both the direct and indirect effectsof final demand changes on total energy requirements. The resultsshow that the main energy challenge for China is a structural

adjustment in final demand. The final demand structure indirectlydetermined the industrial structure in China's economy. For example,final demand for investment indirectly pushed heavy industrygrowth, such as steel and cement production.

According to the calculation, gross fixed capital formation,household consumption expenditures and exports are three mainsources of China's total energy requirements. The impacts of thesefactors on total energy use shifted with the development of theeconomy. In the early 1990s, China experienced a high inflationregime (up to 27.7 in September 1994), and the Chinese govern-ment adopted a range of policies to support the economy, such asdevaluing the RMB by one-third in 1994, tax reform in 1994 andstate-owned enterprise (SOE) reform since 1992. In the 1990s,China's energy use was still mainly driven by household consump-tion, which accounted for about two-fifths of the total energy use(Table 2), but its contribution to energy use change decreased from31 percent (1992–1997) to 20 percent (1997–2002) (Table 4). Incontrast, the contribution of exports to the changes in energy useincreased from 30 percent (1992–1997) to 39 percent (1997–2002). In other words, China was shifting toward becoming anexport-oriented economy in the 1990s.

However, trade surplus fell in the late 1990s, early 2000s and2009. To offset the weakened demand abroad and to reduceoverproduction, the Chinese government boosted investmentthrough the construction of public infrastructure, such as ChinaWestern Development in 1999, Revitalize the Old NortheastIndustrial Bases in 2003, the Rise of Central China Plan in 2004,the New Rural Reconstruction Movement in 2005, and theEconomic Stimulus Plan in 2009, among others. In other words,investment has been a major factor driving China's economicgrowth in the past decade. The contribution of gross fixed capitalformation to China's total energy use increased from approxi-mately 30 percent before 2002 to 39 percent in 2010 (Table 2), andits contribution to the change of energy use swelled to three-quarters during 2007–2010 (Table 4). However, this investment-led economic growth is unsustainable because the heavy depen-dence on investment not only created energy and environmental

−5

0

5

−5

0

5

−5

0

5

−5

0

5

0 10 20 30 0 10 20 30 0 10 20 30 0 10 20 30 0 10 20 30

1992−1997, G 1992−1997, B 1992−1997, d 1992−1997, y 1992−1997, p

1997−2002, G 1997−2002, B 1997−2002, d 1997−2002, y 1997−2002, p

2002−2007, G 2002−2007, B 2002−2007, d 2002−2007, y 2002−2007, p

2007−2010, G 2007−2010, B 2007−2010, d 2007−2010, y 2007−2010, p

Effects Linear fitted value

Sta

ndar

dize

d va

lue

of e

ffect

s

Number of sectors

Fig. 5. The standardized value of effects and sector aggregation levels.

S.-C. Xie / Energy Policy 73 (2014) 401–415 411

problems but also increased systematic risk, including surginglocal government debt. According to the Chinese National AuditOffice (NAO, 2013), the local government debt rose to 17.7 trillionRMB at the end of June 2013, up 70 percent from two-and-a halfyears previous. It is urgent for China to switch its economy fromexport- and investment-based growth to domestic consumption-oriented growth.

With exports and investment driving economic growth, theincome of Chinese households has reached the level at whichenergy-intensive goods are affordable. For example, the averagevehicle holdings per urban household reached 0.22 in 2012.According to the results in Section 4, household consumptionhas contributed 30 percent of the total energy use (Table 2) and toone-quarter of the change in energy use (Table 4) in 2010. GivenChina's population, which is greater 1.3 billion people, and accel-erating urbanization, the commercial and transportation sectorsshould expect to surpass the industrial sector as the main energydemand drivers in the future.

From the supply side (the columns of the I–O table), bothincreasing inputs, such as labor, capital, energy and materials, andimproving the total-factor productivity can boost economic growth.The improvement of total-factor productivity allows an economy touse less energy to produce the same amount of outputs, which can bedefined as the total-factor energy efficiency considering energy inputs.The total-factor energy efficiency consists of two components: tech-nical efficiency and allocative efficiency (Farrell, 1957). Technicalefficiency is simply defined as the ratio of energy inputs over totaloutputs, and allocative efficiency refers to how an economy allocatesinputs by proportion and geography.

This study discusses technical efficiency only. In the energy I–Omodel, the technical energy efficiency is given by the direct energy usecoefficients matrix. Elements of that matrix are in the unit of tce/tcefor energy sectors (because their total outputs are energy fuels) and oftce/RMB for non-energy sectors. According to the results (Table 5), theimprovement of technical energy efficiency saved around 5–6 percentof the total energy use annually in the periods of 1992–1997, 1997–2002 and 2007–2010. In the period of 2002–2007, however, techno-logical change only reduced three percent of the total energy usebecause of the increasing indirect energy use coefficient in theconstruction sector. Finally, we have to emphasize that energy issuesare derivatives of economic and societal development. These systemicissues cannot be solved once and for all. Chinese policymakers shouldcontinue to deepen reform to build up a market-based economy andenergy system. The government should play an important role inproviding public goods, such as energy security and the environment.Let markets but not governments organize the use and distribution ofenergy resources.

Acknowledgments

The China Scholarship Council (CSC) supported this study. Theauthor would like to thank Kiyoshi Fujikawa, Bin Su, YouguoZhang, the editor Michael Jefferson and three anonymous referees

for their helpful advices, comments, and suggestions on the earlierversions of this paper. Errors and omissions remain the soleresponsibility of the author.

Appendix A. LMDI decomposition formulations

Based on identity (14), assuming the aggregate (total energyuse in this paper) is a function of n factors, y¼ f ðx1; x2;…;

xnÞ ¼ x1 � x2…xn. The additive decomposition of the aggregate isthrough the total differential (total changes from all sources)

dy¼ ∂y∂x1

dx1þ∂y∂x2

dx2þ⋯þ ∂y∂xn

dxn ¼ ∑n

i ¼ 1

∂y∂xi

dxi:

Suppose the kth factor changes from xk;0 to xk;T and keeping all elsefactors constant, its impact on the aggregate can be expressed as

Δyxk ¼Z xk;T

xk;0

∂y∂xk

dxk ¼Z xk;T

xk;0

f ð…; xk;…Þxk

dxk ¼Z xk;T

xk;0f ð…; xk;…Þd ln xk:

Using the weight function – the logarithmic mean divisia formula-tion, yD ¼Δy=ΔlnðyÞ, the integral value is estimated by

Δyxk ¼ f ð…; xk;…ÞDΔln xk:

Applying LMDI to identity (18), the technological change effectis given by

ΔeG ¼ ∑m;n;o

ε̂GBdd̂dydp

D

Δln G;

the products mix effect is given by

ΔeB ¼ ∑m;n;o

ε̂GBdd̂dydp

D

Δln Bdþ ∑m;n;o

ε̂Bd̂ypvDΔln B;

the distribution effect is given by

Δed ¼ ∑m;n;o

ε̂GBdd̂dydp

D

Δln ddþ ∑m;n;o

ε̂Bd̂ypvDΔln d;

the level per capita effect is given by

Δey ¼ ∑m;n;o

ε̂GBdd̂dydp

D

Δln ydþ ∑m;n;o

ε̂Bd̂ypvDΔln y;

and the population effect is given by

Δep ¼ ∑m;n;o

ε̂GBdd̂dydp

D

Δln pþ ∑m;n;o

ε̂Bd̂ypvDΔln p:

Appendix B. Sector classification in the hybrid I–O table

Sector classification and final demand categories are given inTables B1 and B2.

S.-C. Xie / Energy Policy 73 (2014) 401–415412

Table B2Final demand categories in the hybrid input–output table.

ID Final demand categories Energy deliveries to final demand categories

F1 Rural household consumption expenditure Rural residential consumptionF2 Urban household consumption expenditure Urban residential consumptionF3 Government consumption expenditureF4 Gross fixed capital formationF5 Changes in inventories Stock changesF6 Exports ExportsF7 Errors Statistical differenceF8 Imports Imports

Table B1Sector classification in the hybrid input–output table.

ID28 ID09 Sector classification ID28 ID09 Sector classification

S01 S01 Mining and washing of coal S15 S03 Chemicals and chemical productsS02 S01 Extraction of crude petroleum S16 S04 Non-metallic mineral productsS03 S01 Extraction of natural gas S17 S05 Basic metalsS04 S01 Refined petroleum products and nuclear fuels S18 S05 Metal productsS05 S01 Coking, production and distribution of gas S19 S06 Machinery and equipmentS06 S01 Electricity, steam production and supply S20 S06 Transport equipmentS07 S02 Agriculture, forestry, and fishing S21 S06 Electric machinery and equipmentS08 S05 Mining of metal ores S22 S06 Electric and telecommunication equipmentS09 S04 Other mining and quarrying S23 S06 Measuring instrument and machinery for cultural activity & office

workS10 S02 Food and tobacco products S24 S06 Artwork, other manufacture, waste, materials recoveryS11 S02 Textiles S25 S06 Water production and distributionS12 S02 Wearing apparel, leather, and related products S26 S07 ConstructionS13 S02 Sawmilling and furniture S27 S08 Traffic, transport, storage and postS14 S02 Paper and paper products, printing and reproduction of recorded

mediaS28 S09 Commerce and service

Fig. C1. China's energy flow chart 1992. Data source: NBS (1998).

S.-C. Xie / Energy Policy 73 (2014) 401–415 413

Appendix C. China's energy flow chart

China's energy flow charts for 1992 and 2010 are shown in Figs.C1 and C2.

References

Ang, B.W., 2004. Decomposition analysis for policymaking in energy: Which is thepreferred method?. Energy Policy 32 (June (9)), 1131–1139.

Ang, B.W., 2006. Monitoring changes in economy-wide energy efficiency: from energy–GDP ratio to composite efficiency index. Energy Policy 34 (March (5)), 574–582.

Ang, B.W., Choi, K., 1997. Decomposition of aggregate energy and gas emissionintensities for industry: a refined Divisia index method. Energy J., 59–73.

Ang, B.W., Liu, N., 2007a. Handling zero values in the logarithmic mean Divisiaindex decomposition approach. Energy Policy 35 (January (1)), 238–246.

Ang, B.W., Liu, N., 2007b. Negative-value problems of the logarithmic mean Divisiaindex decomposition approach. Energy Policy 35 (January (1)), 739–742.

Ang, B.W., Mu, A.R., Zhou, P., 2010. Accounting frameworks for tracking energyefficiency trends. Energy Econ. 32 (September (5)), 1209–1219.

Ang, B.W., Zhang, F.Q., 2000. A survey of index decomposition analysis in energyand environmental studies. Energy 25 (12), 1149–1176.

Bernardini, O., Galli, R., 1993. Dematerialization: long-term trends in the intensityof use of materials and energy. Futures 25 (4), 431–448.

BP, 2013. BP Statistical Review of World Energy 2013. Bp.CEP, 2012. China Electric Power Yearbook. China Electric Power Press, Beijing.Chai, J., Guo, J.-E., Wang, S.-Y., Lai, K.K., 2009. Why does energy intensity fluctuate in

China? Energy Policy 37 (November (12)), 5717–5731.DECC, 2013. UK Energy Flow Chart. URL ⟨https://www.gov.uk/government/collec

tions/energy-flow-charts⟩, July.Dietzenbacher, E., Stage, J., 2006. Mixing oil and water? Using hybrid input–output

tables in a structural decomposition analysis. Econ. Syst. Res. 18 (1), 85–95.Farrell, M., 1957. The measurement of productive efficiency. J. R. Stat. Soc. Ser A

(Gen) 120 (3), 253–290.

Galli, R., 1998. The relationship between energy intensity and income levels:forecasting long term energy demand in Asian emerging countries. Energy J.19 (4), 85–106.

Garbaccio, R., Ho, M.S., Jorgenson, D.W., 1999. Why has the energy–output ratiofallen in China? Energy J. 20, 63–92.

Hoekstra, R., van den Bergh, J., 2003. Comparing structural decomposition analysisand index. Energy Econ. 25 (1), 39–64.

Huang, J., 1993. Industry energy use and structural change: a case study of ThePeople's Republic of China. Energy Econ. 15 (2), 131–136.

Lau, L.J., 2010. Input–occupancy–output models of the non-competitive type andtheir application—an examination of the China-US trade surplus. Soc. Sci. China31 (1), 35–54.

Li, Q., Xue, T.D., 1998. ZHONG GUO JING JI FA ZHAN BU MEN FEN XI—JIAN XIN BIANKE BI JIA TOU RU CHAN CHU XU LIE BIAO (in Chinese). China Statistics Press,Beijing.

Liao, H., Fan, Y., Wei, Y.M., 2007. What induced China's energy intensity to fluctuate:1997–2006?. Energy Policy 35 (September (9)), 4640–4649.

Lin, X., Polenske, K.R., 1995. Input–output anatomy of China's energy use changes inthe 1980s. Econ. Syst. Res. 7 (January (1)), 67–84.

Liu, Q.Y., Peng, Z.L., 2010. ZHONG GUO 1992–2005 NIAN KE BI JIA TOU RU CHANCHU XU LIE BIAO JI FEN XI (in Chinese). China Statistics Press, Beijing.

Ma, C., Stern, D.I., 2008. China's changing energy intensity trend: a decompositionanalysis. Energy Econ. 30 (May (3)), 1037–1053.

Medlock III, K., Soligo, R., 2001. Economic development and end-use energydemand. Energy J. 22 (2), 77–106.

Miller, R.E., Blair, P.D., 2009. Input–Output analysis: foundations and extensions.Cambridge University Press, Cambridge, United Kingdom.

NAO, 2013. The Audit of National Government Debt in China. Technical Report.National Audit Office, December.

NBS, 1996. Input–Output Table of China 1992. China Statistics Press, Beijing.NBS, 1998. China Energy Statistical Yearbook 1991–1996. China Statistics Press,

Beijing.NBS, 1999. Input–Output Table of China 1997. China Statistics Press, Beijing.NBS, 2006. Input–Output Table of China 2002. China Statistics Press, Beijing.NBS, 2009. Input–Output Table of China 2007. China Statistics Press, Beijing.

Fig. C2. China's energy flow chart 2010. Data source: NBS (2011).

S.-C. Xie / Energy Policy 73 (2014) 401–415414

NBS, 2010a. China Compendium of Statistics 1949–2008. China Statistics Press,Beijing.

NBS, 2010b. China Energy Statistical Yearbook 2009. China Statistics Press, Beijing.NBS, 2011. China Energy Statistical Yearbook 2011. China Statistics Press, Beijing.NBS, 2012a. China Statistical Yearbook 2012. China Statistics Press, Beijing.NBS, 2012b. China Urban Life and Price Yearbook 2012. China Statistics Press,

Beijing.NBS, 2012c. China Yearbook of Agricultural Price Survey 2012. China Statistics Press,

Beijing.NBS, 2012d. Input–Output Table of China 2010. China Statistics Press, Beijing.Rosen D H, Houser T., 2007. China energy: a guide for the perplexed. Center for

Strategic and International Studies and Peterson Institute for InternationalEconomics, Washington, DC, United States.

Sinton, J.E., 2001. Accuracy and reliability of China's energy statistics. China Econ.Rev. 12 (4), 373–383.

Sinton, J.E., Fridley, D., 2002. A guide to China's energy statistics. J. Energy Lit. 8,22–35.

Sinton, J.E., Levine, M., 1994. Changing energy intensity in Chinese industry: therelatively importance of structural shift and intensity change. Energy Policy 22(3), 239–255.

Su, B., Ang, B.W., 2012a. Structural decomposition analysis applied to energy andemission: aggregation issues. Econ. Syst. Res. 24 (September (3)), 299–317.

Su, B., Ang, B.W., 2012b. Structural decomposition analysis applied to energy andemissions: some methodological developments. Energy Econ. 34 (January (1)),177–188.

Su, B., Ang, B.W., 2013. Input–Output analysis of CO2 emissions embodied in trade:competitive versus non-competitive imports. Energy Policy 56, 83–87.

Su, B., Huang, H.C., Ang, B.W., Zhou, P., 2010. Input–Output analysis of CO2

emissions embodied in trade: the effects of sector aggregation. Energy Econ.32 (January (1)), 166–175.

Wachsmann, U., Wood, R., Lenzen, M., Schaeffer, R., 2009. Structural decompositionof energy use in Brazil from 1970 to 1996. Appl. Energy 86 (April (4)), 578–587.

Weber, C.L., 2009. Measuring structural change and energy use: decomposition ofthe US economy from 1997 to 2002. Energy Policy 37 (April (4)), 1561–1570.

Weber, C.L., Peters, G.P., Guan, D., Hubacek, K., 2008. The contribution of Chineseexports to climate change. Energy Policy 36 (September (9)), 3572–3577.

Wood, R., Lenzen, M., 2006. Zero-value problems of the logarithmic mean divisiaindex decomposition method. Energy Policy 34 (August (12)), 1326–1331.

Xie, S., Chen, C., Li, L., Huang, C., Cheng, Z., Lu, J., 2009. 2006 China energy flowchart. China Energy 31 (3), 21–23.

Xin Hua News, 2008. China's 4 Trillion Yuan Stimulus to Boost Economy, DomesticDemand. URL ⟨http://news.xinhuanet.com/english/2008-11/09/content_10331324.htm⟩, November.

Xin Hua News, 2013. China Pledges Steady, Human-Centered Urbanization. URL⟨http://news.xinhuanet.com/english/china/2013-12/14/c_132968302.htm⟩,December.

Zhang, Z., 2003. Why did the energy intensity fall in China's industrial sector in the1990s? The relative importance of structural change and intensity change.Energy Econ. 25 (6), 625–638.

Zhao, X., Ma, C., Hong, D., 2010. Why did China's energy intensity increase during1998–2006. Decomposition and policy analysis. Energy Policy 38 (January (3)),1379–1388.

S.-C. Xie / Energy Policy 73 (2014) 401–415 415