Energy Policy 58 (2013) 312–318

Contents lists available at SciVerse ScienceDirect

Energy Policy

0301-42http://d

n CorrYuan NO

E-m

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

Carbon dioxide emission drivers for a typical metropolis usinginput–output structural decomposition analysis

Yafei Wang a, Hongyan Zhao b,c, Liying Li d, Zhu Liu b,c, Sai Liang d,e,f,n

a Institute of National Accounts, Beijing Normal University, Beijing 100875, Chinab Institute of Applied Ecology, Chinese Academy of Sciences, Shenyang 110016, Chinac University of Chinese Academy of Sciences, Beijing 100049, Chinad School of Environment, Tsinghua University, Beijing 100084, Chinae State Key Joint Laboratory of Environment Simulation and Pollution Control, Tsinghua University, Beijing 100084, Chinaf School of Natural Resources and Environment, University of Michigan, Ann Arbor, Michigan 48109-1041, United States

H I G H L I G H T S

� Changes in production structure and population are drivers of CO2 increment.

� Changes in CO2 intensity and per capita GDP are forces to offset CO2 increment.� Final demand structure change has limited effect on Beijing's CO2 emission change.� Beijing's key final demand categories and economic sectors are identified.� Policy implications of Beijing's results are analyzed.

a r t i c l e i n f o

Article history:Received 12 October 2012Accepted 13 March 2013Available online 6 April 2013

Keywords:CO2 emissionInput–output analysisStructural decomposition analysis

15/$ - see front matter & 2013 Elsevier Ltd. Ax.doi.org/10.1016/j.enpol.2013.03.022

esponding author at: Tsinghua University, Sch. 1, Beijing 100084, China. Tel.: þ86 10 6279ail address: liangsai09@gmail.com (S. Liang).

a b s t r a c t

As the capital of China, Beijing is regarded as a major metropolis in the world. Study of the variation intemporal CO2 emissions generated by the driving forces in Beijing can provide guidance for policydecisions on CO2 emissions mitigation in global metropolises. Based on input–output structuraldecomposition analysis (IO-SDA), we analysed the driving forces for the increment in CO2 emissions inBeijing from both production and final demand perspectives during 1997–2010. According to our results,the CO2 emission growth in Beijing is driven mainly by production structure change and populationgrowth, partly offset by CO2 emission intensity reduction as well as the decline in per capita final demandvolume during the study period. Final demand structure change has a limited effect on the change in theCO2 emissions in Beijing. From the final demand perspective, urban trades, urban residential consump-tion, government consumption and fixed capital formation are mainly responsible for the boomingemissions. This study showed how the “top-down” IO-SDA methodology was implemented on a cityscale. Policy implications from this study would be helpful for addressing CO2 emissions mitigation inglobal capital cities and metropolises.

& 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Cities are centres of anthropogenic activities, making them majorcontributors to global CO2 emissions (Satterthwaite, 2008). Citiesthus play a crucial role in human actions for curbing global climatechange (Hillman and Ramaswami, 2010; Kennedy et al., 2010; Koehn,2008). The impact of cities on energy use and associated emissions ismore severe in developing countries that are in a state of rapidurbanisation and industrialisation. China, the world largest

ll rights reserved.

ool of Environment, Tsinghua4144; fax: þ86 10 62796956.

developing country, has already become the top primary energyconsumer as well as the top CO2 emitter in the world (Boden et al.,2012; Gregg et al., 2008). About half of the Chinese population nowlives in cities, especially in metropolises such as Beijing and Shanghai(NBSC, 2011). Energy usage in Chinese cities consumes the majorshare (84%) of the national total (Dhakal, 2009). The expectedincrease in urban households and the improvement of quality of lifeimply an even higher emission scenario over the next decades.Understanding the emission status of Chinese cities is thereforecritical for global mitigation.

Based on current knowledge of emission status, uncovering thedriving forces of the increment in emissions is the cornerstone forpresenting a comprehensive analysis and projecting future emission

Y. Wang et al. / Energy Policy 58 (2013) 312–318 313

scenarios (Guan et al., 2008; Liang et al., in press; Wang and Liang, inpress). The relative contributions of socio-economic factors, such aspopulation, economic growth, production, and final demand struc-tures, and CO2 intensity to the change in the CO2 emissions in Chinahave been examined (Guan et al., 2008, 2009; Liang and Zhang,2011d; Minx et al., 2011; Peters et al., 2007; Xu et al., 2011; Zhang,2009). Index decomposing analysis (IDA) and structure decomposinganalysis (SDA) models are two major methods for presenting theanalysis of the driving forces. The IDA model inherited from the IPATframework uses the index number concept for decomposition analy-sis. The IDA model requires less data and thus has an advantage inspatial and time series analysis. Several types of IDA models areavailable, including the most famous Kaya identity as well as the mostgenerally used Laspeyres index decomposition analysis and Divisiaindex decomposition analysis (Ang, 2005; Liu et al., 2012a). Manyresearchers have used IDAmethods to reveal the driving forces for thevariation in the national and regional CO2 emissions in China,including several case studies conducted in cities (Chong et al.,2012; Liu et al., 2012c). However, sectors are highly aggregated whenusing IDA models, and this aggregation limits policy decisions to thesectoral or product scale. Final demands (such as residential con-sumption, capital investment, and export) that drive productionactivities are often neglected when using IDA models due to thelimits of the model framework and data sources. These technologicalgaps can be filled by using the SDA method. The SDA model canpresent decompositions on the sectoral scale and has advantages foranalysing the driving forces for emissions from high-resolutionsectors. Input–output structure decomposing analysis (IO-SDA) usingeconomic input–output tables can shape socio-economic drivers fromboth production and final demand perspectives. The IO-SDA modelhas been applied successfully on the national scale and has resulted inconcrete achievements and policy implications. However, the IO-SDAmodel is rarely applied on the city scale because applying the IO-SDAmodel requires time series input–output tables and sectoral energyuse data that are usually challenging on a city scale.

We present a case study applying the IO-SDA model on the cityscale. Beijing, the capital of China and one of the world's largestmetropolises, is chosen as the case due to its unique features andavailability of data. Beijing has experienced intensive industrialisa-tion and urbanisation in past decades. Along with rapid urbanisa-tion, Beijing is also optimising its economic structure. Servicesectors have been highly promoted in Beijing, and the proportionsof agriculture, industry and construction have gradually decreased.These changes indicate the shift in Beijing's economic structurefrom industry-oriented to service-oriented. Promoted by thegovernment's interest in constructing a modern global metropolis,Beijing is being transformed by the development of intensiveservice and commercial industries. Accompanying this trend, theCO2 emissions trajectory in Beijing has recently decreased slightly.This feature has distinguished Beijing from other Chinese cities,most of which are rapidly increasing their CO2 emissions. Analys-ing the driving forces of the CO2 emission trends in Beijing coulddirect policy decisions on global industrial restructuring andgreenhouse gas (GHG) mitigation, especially for metropolises.

This study contributes mainly to the analysis of the drivingforces of the change in CO2 emissions in Beijing during the timeperiod 1997–2010 using the IO-SDA model. We covered high-resolution industrial sectors and updated time series in this study.The relative contributions of socio-economic factors were investi-gated from both production and final demand perspectives.

2. Methodology and data

Structural decomposition analysis based on the environmentalinput–output (IO-SDA) model is used widely to analyse the

relative contributions of influencing factors to changes in eco-nomic and environmental flows (Dietzenbacher and Los, 1998;Guan et al., 2008, 2009; Haan, 2001; Hoekstra and Bergh, 2002;Liang et al., in press; Liang and Zhang, 2011d; Minx et al., 2011;Peters et al., 2007; Rørmose and Olsen, 2005; Weinzettel andKovanda, 2011; Wood, 2009; Xu et al., 2011; Zhang, 2009). The useof environmental input–output tables within the IO-SDA modelcan describe driving forces such as material intensity, productionstructure, final demand structure, and economic growth on thesectoral scale (Miller and Blair, 2009; Xu et al., 2008). The selectionof driving forces, however, is flexible from various perspectives.We disaggregated driving forces from both production and finaldemand perspectives in this study. From the production perspec-tive, five socio-economic factors were considered, namely, CO2

intensity, production structure, final demand structure, per capitafinal demand volume (final demand volume equals gross domesticproduct (GDP)), and population. From the final demand perspec-tive, nine driving forces are considered: rural residential consump-tion, urban residential consumption, government consumption,fixed capital formation, stock increments, intranational exports,international exports, intranational imports, and internationalimports.

For the decomposing process, there is a technical problem withthe IO-SDA model: the non-uniqueness of decomposing results(Guan et al., 2008; Liang et al., in press; Liang and Zhang, 2011d;Peters et al., 2007). If the number of decomposed factors is n, thenumber of possible decomposition forms is n! (Dietzenbacher andLos, 1998). We calculated the average of all possible first-orderdecomposition forms in this study to address this problem(Dietzenbacher and Los, 1998; Haan, 2001; Hoekstra and Bergh,2002; Liang et al., in press; Rørmose and Olsen, 2005). Details ofthe IO-SDA model can be found in Electronic Annex 1 in the onlineversion of this article.

Cities have close interactions with the surrounding regions.Most of the materials consumed by cities are produced in thesurrounding regions. The analysis of the drivers of the change inCO2 emissions embodied in urban imports is important. We alsoused the IO-SDA model to analyse the relative contributions offour socio-economic factors (including CO2 intensity, productionstructure, import structure, and import volume) to the change inCO2 emissions embodied in imports to Beijing.

To run the IO-SDA model for Beijing, three categories of dataare required: monetary input–output tables (MIOTs), population,and energy consumption. The MIOTs for Beijing in 1997, 2000,2002, 2005, 2007, and 2010 were obtained from the website of theMunicipal Bureau of Statistics of Beijing (http://www.bjstats.gov.cn/2007trcc/xzfw/lssj). To make the classification of economicsectors in the MIOTs for Beijing consistent with its energystatistics, these MIOTs were all changed into the 28-sector format.These MIOTs were all converted into 2000 constant prices by priceindexes. Producer price indexes for detailed economic sectors inBeijing are unavailable. The indexes for Beijing's GDP were there-fore used (MBSB, 1998–2012). The population of Beijing in 1997,2000, 2002, 2005, 2007, and 2010 was obtained from BeijingStatistical Yearbooks (MBSB, 1998–2012).

The volume of energy sources consumed by energy conversion(primary energy used for power generation), agriculture, industry,construction, services, and residents in Beijing is obtained directlyfrom Beijing's Energy Balance Table in China Energy StatisticalYearbooks (NBSC, 1998–2011). Energy consumption for detailedindustrial sectors of Beijing was obtained from Beijing StatisticalYearbooks (MBSB, 1998–2012). We re-allocated energy sourcesused for generating electric and heat power into the sectorProduction and supply of electric and heat power, based on datafrom Beijing's Energy Balance Table (NBSC, 1998–2011). Electricand heat power consumed by each sector was then removed to

-50

0

50

100

150

200

llion

ton

nes

of C

O2

Y. Wang et al. / Energy Policy 58 (2013) 312–318314

avoid double-counting. Finally, we calculated CO2 emissions cover-ing fossil fuel combustion and cement production in Beijing.Specific CO2 emission factors (including carbon content, netcalorific value, and carbon oxidation factors) for China were usedinstead of the default values of the Intergovernmental Panel onClimate Change, as coal usage in China is usually inefficient (Guanet al., 2012). The specific CO2 emission factors for China are shownin the study by Guan et al. (Guan et al., 2012; Liu et al., 2012a).

-200

-150

-100

Rur

al r

esid

ents

Urb

an r

esid

ents

Gov

ernm

ent

Fixe

d ca

pita

l

Stoc

k in

crem

ents

Intr

anat

iona

l exp

orts

Inte

rnat

iona

l exp

orts

Intr

anat

iona

l im

port

s

Inte

rnat

iona

l im

port

s

mi

Fig. 3. Contribution of different final demand categories to Beijing’s CO2 emissionchange during 1997–2010.

3. Results and discussion

3.1. Analysis from the production perspective

Based on the CO2 emission inventory for Beijing during1997–2010, we shaped Beijing's emission trajectory (Fig. 1). Fig. 1shows that CO2 emissions in Beijing presented an unstable trend.The CO2 emissions in Beijing decreased sharply during 2000–2002but rebounded during 2002–2007. Then there was a slight declineduring 2007–2010. The fluctuation of the CO2 emission trajectoryin Beijing emphasises the importance of analysing the hiddendrivers. We will analyse the contribution of the driving forces forthe change in the CO2 emissions in Beijing during 1997–2010below (Figs. 2 and 3).

From the production perspective, production structure changeis the biggest driver for the CO2 emission increment in Beijingduring 1997–2010 (Fig. 2). Production structure change in thisperiod comprises two aspects: (1) Proportions of heavy manufac-turing sectors in the industrial sectors of Beijing increased rapidlyfrom 1997 to 2010 (MBSB, 1998–2012). Those heavy manufactur-ing sectors are usually energy-intensive and CO2-emission-inten-sive. (2) Many small and medium scale enterprises emerged inBeijing in the last decade. These enterprises are usually inefficient

58.0

79.3

69.8

82.6 84.5

78.3

40

50

60

70

80

90

1997 2000 2002 2005 2007 2010

mill

ion

tonn

es o

f C

O2

Fig. 1. CO2 emission from Beijing’s production activities during 1997–2010.

-100

-50

0

50

100

150

1997

-200

0

2000

-200

2

2002

-200

5

2005

-200

7

2007

-201

0

1997

-201

0

mill

ion

tonn

es o

f CO

2

Changes in per capita final demand volumeChanges in final demand structureChanges in production structureChanges in intensityChanges in population

Fig. 2. Contribution of 5 factors to Beijing’s CO2 emission change during1997–2010.

in energy usage. Subsequently, the production structure changeduring 1997–2010 contributed 309.4% (62.6 million tonnes) of theincrement in CO2 emissions in Beijing. This increment implies ahigher proportion of CO2 intensive sectors in Beijing's economy.

Population growth is the second largest driver of Beijing's CO2

emission increment during 1997–2010 (Fig. 2). Beijing's perma-nent population (19.6 million) and permanent floating population(7.0 million) in 2010 increased by 58.2% and 356.1%, respectively,over the 1997 level (MBSB, 1998–2012). Rapid population growthcontributed to 158.9% (32.2 million tonnes) of the CO2 emissionincrement in this period.

Final demand structure change had limited effect on thechange in CO2 emissions in Beijing, contributing only 36.4% (7.4million tonnes) of CO2 increment during 1997–2010 (Fig. 2).

Changes in CO2 emission intensity and per capita final demandvolume are two factors offsetting the increment in CO2 in Beijingfrom 1997 to 2010 (Fig. 2). Beijing has taken efficient measures forenergy conservation and reduction of air pollutants in the 10th(2001–2005) and 11th (2006–2010) Five-year Social and EconomicDevelopment Plans due to mandatory requirements imposed bythe Central Government. These measures resulted in the decline inCO2 emissions, which has significantly offset the total CO2 emis-sions boom during the study period. Beijing also took majormeasures to ensure energy conservation and air quality controlduring the 2008 Olympic Games, such as banning approximatelyhalf of the private vehicles and temporarily stopping productionactivities of high-pollution enterprises. These measures also con-tributed to the reduction of domestic CO2 emissions intensity inBeijing. The decline in CO2 intensity offset Beijing’s CO2 emissionsby 41.2 million tonnes (−203.8%). Per capita final demand volume(also named per capita GDP) is another force offsetting total CO2

emissions. According to data from Beijing’s MIOTs in constantprices, Beijing’s per capita GDP in 2010 decreased by 6.5% over the1997 level. This decline is caused mainly by an increase in thenumber of Beijing’s permanent floating population who areusually of lower income. Finally, the decline in Beijing’s per capitaGDP resulted in a reduction in CO2 emissions by 40.7 milliontonnes (−201.0%).

3.2. Analysis from the final demand perspective

3.2.1. Underlying drivers from final demand perspectiveAnalysing the CO2 emission inventory from the final demand

perspective could uncover underlying drivers throughout wholesupply chains (Liang et al., 2012b, 2013a, in press, 2013c; Liu et al.,

Table 1Allocation of CO2 emission change during 1997–2010 by different final demand categories and economic sectors.

Items Ruralresident(%)

Urbanresident(%)

Government(%)

Fixedcapitalformation(%)

Stockincrements(%)

Intranationalexports (%)

Internationalexports (%)

Intranationalimports (%)

Internationalimports (%)

Agriculture 7.58 16.85 3.29 −0.25 −39.06 −0.08 1.36 1.75 2.26Mining and washing of coal −10.13 −0.04 0.00 0.00 5.14 0.43 1.31 0.59 0.00Extraction of petroleum and naturalgas

−3.20 0.15 0.00 0.00 −24.38 12.02 0.43 1.08 20.69

Mining and processing of metal ores 0.00 0.00 0.00 0.00 83.62 18.02 0.92 0.42 26.23Mining and processing of nonmetalores

4.68 −0.11 0.00 −0.08 −29.98 −0.01 0.53 5.58 −0.04

Processing and manufacture of foods 43.46 2.96 0.00 0.00 −37.42 2.43 0.20 2.50 0.99Manufacture of textile −0.36 3.89 0.00 0.00 −2.29 0.24 −3.25 0.74 −0.37Manufacture of clothes, leather andrelated products

16.30 1.52 0.00 0.00 −2.56 0.19 1.15 1.27 0.08

Manufacture of wood products andfurniture

12.51 −0.67 0.00 −1.17 −11.01 0.12 0.02 0.28 −0.01

Papermaking, printing and articlesmanufacture

1.76 −1.15 0.00 0.00 18.78 0.20 −0.22 1.69 0.06

Petroleum processing and coking −5.77 7.20 0.00 0.00 32.83 2.68 9.45 5.88 1.85Chemistry 20.27 3.03 0.00 0.00 10.50 5.23 1.66 7.29 1.98Manufacture of non-metallic mineralproducts

34.28 −1.94 0.00 −0.54 21.46 2.04 0.10 5.21 −0.27

Smelting and pressing of metals 13.12 −1.40 0.00 −0.05 10.33 5.97 12.19 27.16 2.75Manufacture of metal products −0.37 4.22 0.00 −0.60 73.79 −0.02 −0.64 3.25 −0.33Manufacture of general and specialpurpose machinery

9.79 0.22 0.00 −0.10 −10.57 2.10 −0.51 0.36 −0.23

Manufacture of transport equipment 9.98 8.89 0.00 −12.64 21.03 7.82 3.72 1.89 4.39Manufacture of electrical machineryand equipment

6.34 −5.64 0.00 −7.76 −2.29 1.68 0.64 2.59 0.68

Manufacture of communication andelectronic equipment

5.28 0.90 0.00 −6.73 −37.14 3.76 9.55 8.89 0.11

Manufacture of measuringinstruments and machinery forcultural activity and office work

0.94 0.65 0.00 −0.85 −6.77 0.29 −0.17 −0.34 0.39

Manufacture of artwork and othermanufacturing

6.89 3.51 0.00 0.00 39.90 2.37 0.78 2.75 0.22

Production and supply of electric andheat power

−3.24 −38.18 0.00 0.00 0.00 −0.17 0.00 −1.10 0.00

Production and supply of gas 1.62 1.26 0.00 0.00 0.00 −0.01 0.00 0.40 0.00Production and supply of water −1.03 1.40 0.00 0.00 0.00 0.00 0.00 0.00 0.00Construction 0.00 9.05 0.00 70.55 0.00 8.45 16.96 11.23 3.69Transport, storage and post −4.19 7.54 0.00 0.23 −17.23 0.91 15.85 1.48 22.86Wholesale, retail trade and hotel,restaurants

13.29 9.91 0.00 1.22 3.21 5.22 1.65 3.72 1.63

Other services −79.82 65.99 96.71 58.79 0.12 18.15 26.31 3.42 10.40Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Y. Wang et al. / Energy Policy 58 (2013) 312–318 315

2012b). We thus allocated CO2 emissions by different final demandcategories (Fig. 3) and economic sectors (Table 1).

Urban trades are mainly responsible for the CO2 emissionchanges in Beijing during 1997–2010 (Fig. 3). Changes in netintranational exports caused an increase of 17.0 million tonnes inembodied CO2 emissions in Beijing, accounting for 84.0% of thetotal CO2 emission change. Changes in net international importsavoided 52.9 million tonnes of CO2 emissions, assuming that urbanimports to Beijing are reproduced with the application of Beijingtechnology. Intranational export changes caused a 162.9 milliontonne increase in CO2 emissions (Fig. 3), due mainly to increasingintranational exports of petroleum, natural gas, metal ores, trans-port equipment, construction activities, and other services(Table 1). Similarly, increasing international exports of petroleumproducts, coke, metals, electronics and communication equipment,construction activities, and services are mainly responsible for theCO2 emission increment in Beijing caused by international exportchanges (42.9 million tonnes) (Table 1). Increasing intranationalimports of chemicals, metals, electronics and communicationequipment, and construction activities are mainly responsible forthe CO2 emissions in Beijing that were avoided by intranationalimport changes (145.9 million tonnes) (Table 1). Increasing

international imports of petroleum, natural gas, metal ores, andservices are the main drivers for the avoidance of CO2 emissions inBeijing due to international import changes (95.9 million tonnes)(Table 1).

Urban residential consumption, government consumption andfixed capital formation are three main drivers of the CO2 emissionincrement in Beijing (Fig. 3). The increase in urban residentialconsumption caused an increase of 14.9 million tonnes in CO2

emissions in Beijing, due mainly to increasing demands of agri-cultural products and other services (Table 1). The increment inCO2 emissions in Beijing caused by growth in government con-sumption (17.7 million tonnes) resulted mainly from increasingdemands of agricultural products and other services (Table 1).Increasing fixed capital formation from construction activities andservices is mainly responsible for the increment in CO2 emissionscaused by fixed capital formation (25.8 million tonnes) (Table 1).

3.2.2. Contribution of drivers to the change in CO2 emissionsembodied in urban imports

Fig. 3 shows that urban trade is the biggest contributor to thechange in CO2 emissions in Beijing. This change indicates that

Y. Wang et al. / Energy Policy 58 (2013) 312–318316

Beijing’s import is the underlying driver for CO2 emissions forregions providing goods for Beijing. Optimising Beijing’s importstructure and promoting Beijing’s green consumption to reduce itsimport volume can contribute effectively to global CO2 mitigation.Urban imports to Beijing are internal to the intermediate deliverymatrix of MIOTs for Beijing. The intermediate delivery matrix forBeijing’s domestic production is merged with the matrix forBeijing’s urban imports. Urban imports and urban exports areexpressed by two columns in the final demands of the MIOTs forBeijing. The IO-SDA of Beijing’s CO2 emissions is thus based on theassumption that imports to Beijing are produced using Beijing’sdomestic technologies.

We used the IO-SDA model to investigate the drivers for CO2

emission change caused by urban imports (Fig. 4). Beijing has beenin a rapid urbanisation stage over the period 1997–2010. Increas-ing urban population and improved quality of life produced largerdemands for resources and products, further resulting in rapidgrowth of urban import volume. Subsequently, growth of urbanimport volume is the biggest driver of the increment in Beijing’sCO2 emissions embodied in urban imports. The volume growth ofBeijing’s intranational and international imports increased embo-died CO2 emissions by 115.6 million tonnes (79.3%) and 72.3million tonnes (75.4%), respectively.

Production structure change is the second driver for theincrement in CO2 emissions in Beijing embodied in urban imports.Increasing proportions of heavy manufacturing sectors and manynewly emerged inefficient small scale enterprises in Chinese sub-regions produced production structure changes. Production struc-ture changes subsequently increased Beijing’s CO2 emissionsembodied in intranational and international imports by 84.0million tonnes (57.6%) and 31.8 million tonnes (33.2%), respec-tively, during 1997–2010.

-150

-100

-50

0

50

100

150

200

250

mill

ion

tonn

es o

f CO

2

Changes in intensityChanges in production structureChanges in import structureImport volume

-40

-20

0

20

40

60

80

100

120

1997-2000 2000-2002 2002-2005 2005-2007 2007-2010 1997-2010

1997-2000 2000-2002 2002-2005 2005-2007 2007-2010 1997-2010

mill

ion

tonn

es o

f CO

2

Changes in intensityChanges in production structureChanges in import structureImport volume

Fig. 4. Contribution of 4 factors to the change in CO2 emission caused by Beijing’surban imports during 1997–2010 (a) International imports and (b) Internationalimports.

CO2 intensity change is the main driver to offset the growth ofCO2 emissions embodied in urban imports. China establishedcompulsory targets for energy conservation and air pollutantreduction in the 10th and 11th Five-year Plans. China alsoestablished a CO2 mitigation target of reducing CO2 emissionintensity per unit GDP by 40–45% in 2020 above the 2005 level.China’s sub-regions (such as provinces and cities) have takenvarious measures to achieve these targets, such as promotingcleaner production, adjusting economic structure, and encoura-ging renewable energy usage. Some cities even use power ration-ing measures to limit energy usage. These measures havesignificantly reduced the CO2 emission intensity of goods importedto Beijing. CO2 intensity change during 1997–2010 reduced CO2

emissions in Beijing embodied in intranational and internationalimports by 50.5 million tonnes (−34.6%) and 7.7 million tonnes(−8.0%), respectively.

Urban import structure change has a limited effect on thechange in CO2 emissions in Beijing embodied in intranational andinternational imports, causing a decline of embodied CO2 emis-sions by only 3.3 million tonnes (−2.2%) and 0.6 million tonnes(−0.6%), respectively.

3.3. Policy implications

In summary, as a famous global metropolis, Beijing illustrateshow driving forces compete with one another and then result in afluctuation in the CO2 emission trajectory. The features of the CO2

emission trend in Beijing are summarised below.First, carbonised production structure. According to Fig. 2,

production structure change contributed significantly to the incre-ment in CO2 emissions, especially in the period of 2002–2005 and2007–2010.

Second, dramatic change of per capita final demand volume.The effect of per capita final demand volume change on CO2

mitigation was positive during 1997–2002, negative during 2002–2005, and then positive again during 2005–2010 (Fig. 2). Therecent decline of per capita final demand volume is actually causedby rapid population growth.

Third, effective CO2 emission intensity reduction. CO2 emissionintensity reduction, achieved mainly by technology improvement,has considerably offset the total CO2 emission increment. The CO2

emission intensity increase is the largest contributor to theincrement in CO2 emissions in Beijing during 1997–2000 (Fig. 2).The total CO2 emissions in Beijing also increased dramatically from58.0 million tonnes of CO2 in 1997 to 79.3 million tonnes of CO2 in2000 (Fig. 1). These results indicate that CO2 intensity reductionseems to be the inevitable factor for curbing total increment in CO2

emissions.Finally, importing energy-intensive products rather than pro-

ducing them domestically can also significantly decrease domesticCO2 emissions due to the transition of Beijing’s economic structureto a service orientation, further resulting in rapid development ofdomestic services and the increase in imports. However, thistransition will push CO2 emissions into regions that provide goodsfor Beijing. Reducing the product demand in Beijing to decrease itsimports is preferred to achieve global CO2 mitigation.

Several policy implications follow.

3.3.1. Policy implications for reducing Beijing’s domestic CO2

emissionsBeijing is featured as a centre for China’s modern service

industries. More and more people tend to work and live in Beijingdue to its better social resources. Beijing’s permanent floatingpopulation will most likely continue to grow in the near future.The declining trend of Beijing’s per capita GDP, however, will

Y. Wang et al. / Energy Policy 58 (2013) 312–318 317

gradually be weakened, as the Chinese government is payingspecial attention to people’s livelihoods. In general, Beijing’s GDPwill continue to grow, indicating that the effects of permanentpopulation growth on the increment in CO2 emissions will exceedthe effects of per capita GDP decline on mitigation of CO2

emissions. Mitigation of CO2 emissions in Beijing in the futurewould rely mainly on CO2 emission intensity reduction andstructural optimisation of production and final demand activities.

In addition to conventional measures proposed by nationalgovernments (such as sectoral cleaner production and renewableenergy utilisation), two points should be specifically considered toreduce Beijing’s CO2 emission intensity effectively: energy con-servation for transportation and buildings and carbon capture andstorage. Energy consumption in Beijing is dominated by serviceactivities that are mainly transportation and buildings (MBSB,1998–2012). This situation differs from the national level and thatin many other Chinese cities. Beijing’s manufacturing enterprisesare gradually being moved away, which gradually reduces CO2

emissions from industry. Energy-efficient buildings and publictransportation should be especially encouraged in Beijing toreduce Beijing’s domestic CO2 emissions effectively. In additionto source control of CO2 emissions, end-of-pipe treatment shouldalso be promoted. Beijing should also develop technologies forcarbon capture and storage to reduce CO2 emissions released tothe atmosphere. Technological innovation is challenging. However,Beijing has abundant research resources such as China’s mostfamous universities and institutes. As one of China’s most devel-oped areas (NBSC, 2011), Beijing also has the economic capabilityto support technology innovation for carbon capture and storage.

According to the results shown in Section 3.1, Beijing’s produc-tion structure change is influenced mainly by the change in theproportion of heavy manufacturing sectors and the appearance ofmany small and medium scale enterprises utilising inefficienttechnologies. Beijing is gradually moving its heavy manufacturingenterprises away. To improve production structure further, Beijingshould take action to limit the number of small scale enterpriseswith inefficient technologies. Beijing’s production structurechange during 1997–2010 shows the largest negative contributionto its mitigation of CO2 emissions. Improving production structurecan therefore produce a large potential for mitigation of CO2

emissions in Beijing.According to the results shown in Section 3.1, the change in the

final demand structure in Beijing during 1997–2010 shows anegative contribution to mitigation of CO2 emissions in Beijing.Beijing should therefore also optimise its final demand structure.Green consumption, such as controlling luxury consumption,should be encouraged, resulting in the use of fewer energy-intensive products and using mass transit for transportation.

3.3.2. Policy implications for reducing CO2 emissions embodied inimports

Beijing is a world-famous metropolis. Goods consumed byBeijing’s production and consumption activities depend mainlyon urban imports, indicating that Beijing is the underlying driverfor CO2 emissions in trading partners. Mitigating the CO2 emis-sions embodied in urban imports for Beijing relies not only on CO2

emission intensity reduction and production structure optimisa-tion of trading partners but also on import structure optimisationand import volume reduction for Beijing.

China’s sub-regions will take various measures to reduce theirCO2 emission intensity and improve their production structure dueto mandatory requirements imposed by China’s central govern-ment. From the supply chain perspective, reducing the demand ofthe downstream economy can effectively mitigate CO2 emissionsthroughout whole supply chains. Thus, optimising Beijing’s import

structure and reducing Beijing’s import volume are more signifi-cant for reducing national and global CO2 emissions. Beijing shouldtake measures to encourage the import of goods that are lesscarbon-intensive. Moreover, Beijing should promote sectoral clea-ner production and a green lifestyle to reduce its import volume.

In the near future, China will continue to improve life qualityfor its citizens. Rapid economic growth and urbanisation will stillcontinue for some time. Beijing’s economic structure is approach-ing relative maturity. The success of Beijing’s CO2 emissionmitigation actions will provide a potential pathway for low-carbon development for global metropolises.

3.3.3. Policy implications for administrative measuresBeijing apparently reduced its CO2 emissions during 2007–

2010, unique among Chinese cities. In addition to technologyimprovement and industrial restructuring, Beijing’s administrativemeasures actually play an inevitable role in this dramatic change.To promote the “Green Olympic Games” in Beijing in 2008 (Wuand Zhang, 2008), the Beijing government mandated movingmany energy-intensive manufacturing enterprises into otherregions. The Beijing government also closed enormous inefficientpower plants and manufacturing enterprises. These administrativemeasures had strong effects on both the decline in domesticenergy intensity and the increment in imports. Compared withmandated government-led measures, market mechanisms such asdomestic cap-and-trade systems and carbon taxation should beencouraged to promote further decreases in CO2 emissions.

4. Conclusion

Using the IO-SDA model, we analysed drivers for the change inCO2 emissions in Beijing during 1997–2010 from both productionand final demand perspectives. Production structure change andpopulation growth were two main drivers for an increment in CO2

emissions during 1997–2010, while CO2 emission intensity reduc-tion and per capita final demand volume decline were two mainforces to offset this increment. Final demand structure change,however, had a small effect on the change in CO2 emissions inBeijing. From the final demand perspective, urban trades, urbanresidential consumption, government consumption, and fixedcapital formation are the main factors responsible for the changein CO2 emissions in Beijing. Mitigation of domestic CO2 emissionsin Beijing will rely on CO2 emission intensity reduction andstructural optimisation of production and final demand activities.Optimising Beijing’s import structure and reducing Beijing’simport volume are more significant for reduction of national andglobal CO2 emissions.

In future work, the drivers of CO2 emissions in more represen-tative cities should be investigated using the IO-SDA model toprovide more guidance for policy decisions for low-carbon cityconstruction. However, developing this guidance requires theconstruction of urban MIOTs because many cities do not compileMIOTs. Liang and Zhang have proposed methods to constructphysical input–output tables (PIOTs) in Chinese cities (Liang et al.,2010; Liang and Zhang, 2011a) and industrial clusters (Liang et al.,2011). These investigators also applied those methods in a typicalChinese city for urban energy system management (Liang et al.,2010; Liang and Zhang, 2011c), urban solid waste recycling (Liangand Zhang, 2012), and urban air pollutant mitigation (Liang et al.,2012b). The PIOT is the basis for the MIOT, as they interact witheach other through sectoral prices (Liang and Zhang, in press;Weisz and Duchin, 2006). The PIOT and MIOT can also becombined to construct hybrid input–output tables (HIOTs) (Liangand Zhang, in press), which can be used for life cycle assessment(Hawkins et al., 2007; Liang et al., 2012a, 2013b, 2012c) and the

Y. Wang et al. / Energy Policy 58 (2013) 312–318318

investigation of policy interactions (Liang and Zhang, 2011b). Infuture work, methods for the construction of urban PIOTs shouldbe improved to reduce uncertainty.

Acknowledgment

This article is financially supported by the Major Program ofNational Philosophy and Social Science Foundation of China(11&ZD045), the National Philosophy and Social Science Founda-tion of China (12BTJ013) and Beijing Natural Science Foundation(8112026).

Appendix A. Supporting information

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.enpol.2013.03.022.

References

Ang, B.W., 2005. The LMDI approach to decomposition analysis: a practical guide.Energy Policy 33, 867–871.

Boden, T.A., Marland, G., Andres, R.J.Global, 2012. Global, regional, and nationalfossil-fuel CO2 emissions. Carbon Dioxide Information Analysis Center.

Chong, W.H.B., Guan, D., Guthrie, P., 2012. Comparative analysis of carbonizationdrivers in China’s Megacities. Journal of Industrial Ecology 16, 564–575.

Dhakal, S., 2009. Urban energy use and carbon emissions from cities in China andpolicy implications. Energy Policy 37, 4208–4219.

Dietzenbacher, E., Los, B., 1998. Structural decomposition techniques: sense andsensitivity. Economic Systems Research 10, 307–323.

Gregg, J.S., Andres, R.J., Marland, G., 2008. China: emissions pattern of the worldleader in CO2 emissions from fossil fuel consumption and cement production.Geophysical Research Letters 35, L08806.

Guan, D., Hubacek, K., Weber, C.L., Peters, G.P., Reiner, D.M., 2008. The drivers ofChinese CO2 emissions from 1980 to 2030. Global Environmental Change-Human and Policy Dimensions 18, 626–634.

Guan, D., Liu, Z., Geng, Y., Lindner, S., Hubacek, K., 2012. The gigatonne gap inChina’s carbon dioxide inventories. Nature Climate Change 2, 672–675.

Guan, D., Peters, G.P., Weber, C.L., Hubacek, K., 2009. Journey to world top emitter:an analysis of the driving forces of China’s recent CO2 emissions surge.Geophysical Research Letters 36, L04709.

Haan, M.D., 2001. A structural decomposition analysis of pollution in the Nether-lands. Economic Systems Research 13, 181–196.

Hawkins, T., Hendrickson, C., Higgins, C., Matthews, H.S., Suh, S., 2007. A mixed-unitinput–output model for environmental life-cycle assessment and material flowanalysis. Environmental Science & Technology 41, 1024–1031.

Hillman, T., Ramaswami, A., 2010. Greenhouse gas emission footprints and energyuse benchmarks for eight U.S. cities. Environmental Science & Technology 44,1902–1910.

Hoekstra, R., Bergh, J.C.J.M.v.d., 2002. Structural decomposition analysis of physicalflows in the economy. Environmental and Resource Economics 23, 357–378.

Kennedy, C., Steinberger, J., Gasson, B., Hansen, Y., Hillman, T., Havranek, M., Pataki,D., Phdungsilp, A., Ramaswami, A., Mendez, G.V., 2010. Methodology forinventorying greenhouse gas emissions from global cities. Energy Policy 38,4828–4837.

Koehn, P.H., 2008. Underneath Kyoto: emerging subnational government initiativesand incipient issue-bundling opportunities in China and the United States.Global Environmental Politics 8, 53–77.

Liang, S., Liu, Z., Xu, M., Zhang, T., 2013a. Waste oil derived biofuels in China bringbrightness for global GHG mitigation. Bioresource Technology 131, 139–145.

Liang, S., Shi, L., Zhang, T., 2011. Achieving dewaterization in industrial parks A casestudy of the Yixing economic development zone. Journal of Industrial Ecology15, 597–613.

Liang, S., Wang, C., Zhang, T., 2010. An improved input–output model for energyanalysis: a case study of Suzhou. Ecological Economics 69, 1805–1813.

Liang, S., Xu, M., Liu, Z., Suh, S., Zhang, T. Socioeconomic drivers of mercuryemissions in China from 1992 to 2007. Environmental Science & Technologyhttp://dx.doi.org/10.1021/es303728d, in press.

Liang, S., Xu, M., Zhang, T., 2012b. Unintended consequences of bioethanol feed-stock choice in China. Bioresource Technology 125, 312–317.

Liang, S., Xu, M., Zhang, T., 2013c. Life cycle assessment of biodiesel production inChina. Bioresource Technology 129, 72–77.

Liang, S., Zhang, T., 2011a. Data acquisition for applying physical input–outputtables in Chinese cities. Journal of Industrial Ecology 15, 825–835.

Liang, S., Zhang, T., 2011b. Interactions of energy technology development and newenergy exploitation with water technology development in China. Energy 36,6960–6966.

Liang, S., Zhang, T., 2011c. Managing urban energy system: a case of Suzhou inChina. Energy Policy 39, 2910–2918.

Liang, S., Zhang, T., 2011d. What is driving CO2 emissions in a typical manufacturingcenter of South China? The case of Jiangsu Province. Energy Policy 39,7078–7083.

Liang, S., Zhang, T., 2012. Comparing urban solid waste recycling from the view-point of urban metabolism based on physical input–output model: a case ofSuzhou in China. Waste Management 32, 220–225.

Liang, S., Zhang, T. Investigating reasons for differences in the results of environ-mental, physical and hybrid input–output models. Journal of Industrial Ecologyhttp://dx.doi.org/10.1111/jiec.12010, in press.

Liang, S., Zhang, T., Jia, X., 2013c. Clustering economic sectors in China on a life cyclebasis to achieve environmental sustainability. Frontiers of EnvironmentalScience & Engineering 7, 97–108.

Liang, S., Zhang, T., Wang, Y., Jia, X., 2012b. Sustainable urban materials manage-ment for air pollutants mitigation based on urban physical input–output model.Energy 42, 387–392.

Liang, S., Zhang, T., Xu, Y., 2012c. Comparisons of four categories of waste recyclingin China’s paper industry based on physical input–output life-cycle assessmentmodel. Waste Management 32, 603–612.

Liu, Z., Geng, Y., Lindner, S., Guan, D., 2012a. Uncovering China’s greenhouse gasemission from regional and sectoral perspectives. Energy 45, 1059–1068.

Liu, Z., Geng, Y., Lindner, S., Zhao, H., Fujita, T., Guan, D., 2012b. Embodied energyuse in China’s industrial sectors. Energy Policy 49, 751–758.

Liu, Z., Liang, S., Geng, Y., Xue, B., Xi, F., Pan, Y., Zhang, T., Fujita, T., 2012c. Features,trajectories and driving forces for energy-related GHG emissions from Chinesemega cites: the case of Beijing, Tianjin, Shanghai and Chongqing. Energy 37,245–254.

MBSB (Ed.), 1998–2012. Beijing Statistical Yearbooks. China Statistics Press, Beijing,in Chinese.

Miller, R.E., Blair, P.D., 2009. Input–Output Analysis: Foundations and Extensions,second ed. Cambridge University Press, Cambridge, UK.

Minx, J.C., Baiocchi, G., Peters, G.P., Weber, C.L., Guan, D., Hubacek, K., 2011. A“Carbonizing Dragon”: China’s fast growing CO2 emissions revisited. Environ-mental Science & Technology 45, 9144–9153.

NBSC (Ed.), 1998-2011. China Energy Statistical Yearbooks. China Statistics Press,Beijing.

NBSC (Ed.), 2011. China Statistical Yearbook 2011. China Statistics Press, Beijing.Peters, G.P., Weber, C.L., Guan, D., Hubacek, K., 2007. China’s growing CO2

emissions: a race between increasing consumption and efficiency gains.Environmental Science & Technology 41, 5939–5944.

Rørmose, P., Olsen, T., 2005. Structural decomposition analysis of air emissions inDenmark 1980–2002. In: 15th International Conference on Input–OutputTechniques. Beijing, China.

Satterthwaite, D., 2008. Cities’ contribution to global warming: notes on theallocation of greenhouse gas emissions. Environment and Urbanization 20,539–549.

Wang, Y., Liang, S. Carbon dioxide mitigation target of China in 2020 and keyeconomic sectors. Energy Policy, http://dx.doi.org/10.1016/j.enpol.2013.02.038,in press.

Weinzettel, J., Kovanda, J., 2011. Structural decomposition analysis of raw materialconsumption. Journal of Industrial Ecology 15, 893–907.

Weisz, H., Duchin, F., 2006. Physical and monetary input–output analysis: whatmakes the difference? Ecological Economics 57, 534–541.

Wood, R., 2009. Structural decomposition analysis of Australia’s greenhouse gasemissions. Energy Policy 37, 4943–4948.

Wu, J., Zhang, Y., 2008. Olympic games promote the reduction in emissions ofgreenhouse gases in Beijing. Energy Policy 36, 3422–3426.

Xu, M., Li, R., Crittenden, J.C., Chen, Y., 2011. CO2 emissions embodied in China’sexports from 2002 to 2008: a structural decomposition analysis. Energy Policy39, 7381–7388.

Xu, M., Zhang, T., Allenby, B., 2008. How Much Will China Weigh? Perspectivesfrom consumption structure and technology development. EnvironmentalScience & Technology 42, 4022–4028.

Zhang, Y., 2009. Structural decomposition analysis of sources of decarbonizingeconomic development in China; 1992–2006. Ecological Economics 68,2399–2405.