Embed Size (px)
Citation preview
at SciVerse ScienceDirect
Energy 53 (2013) 230e236
Contents lists available
Energy
journal homepage: www.elsevier .com/locate/energy
Differential output growth across regions and carbon dioxideemissions: Evidence from U.S. and China
Chunhua Wang*
School of International Trade and Economics, University of International Business and Economics, 10 Huixin Dong Jie, Chaoyang District,Beijing 100029, China
a r t i c l e i n f o
Article history:Received 9 October 2012Received in revised form28 January 2013Accepted 16 February 2013Available online 24 March 2013
Keywords:Carbon dioxide emissionsDifferential growthDecomposition
* Tel.: þ86 10 6449 3581; fax: þ86 10 6449 3042.E-mail addresses: [email protected]
0360-5442/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.energy.2013.02.044
a b s t r a c t
This paper explores the importance of differential output growth across regions within a country inreducing the country’s total carbon dioxide emissions from the combustion of fossil fuels. It proposes aframework that decomposes changes in emissions into sources attributable to 1) national growth rate ofgross domestic product (GDP), 2) differential GDP growth across regions, 3) changes in energy intensity,and 4) changes in CO2 emission coefficient of energy. Data for China (1995e2009) and the United States(1990e2009) are analyzed. Uneven growth across regions reduced carbon dioxide emissions in bothcountries.
� 2013 Elsevier Ltd. All rights reserved.
1. Introduction
Policymakers inmany countries face the challenge of controllinggreenhouse gas emissions while trying to achieve economicgrowth. In the world’s two largest emitters e China and the UnitedStates, climate change and greenhouse gas policies mainly target onthe carbon intensity of economy (i.e., carbon emissions per unit ofoutput). To decrease the carbon intensity, different measures havebeen proposed and implemented to achieve lower levels of energyintensity of economy (energy-GDP ratio) and carbon coefficient ofenergy (carbon-energy ratio).
Many previous studies in the literature focus on the dynamics ofcarbon intensity and per capita carbon dioxide emissions (see, forexample, [1,2]). As for analyzing the factors that drive the changesin total emissions, the Kaya identity [3] offered a simple and usefulframework. A majority of findings suggest that output growth is amain source of emission increments and falling energy intensitycontributes to a decline in emissions [4e11]. The causal relation-ships between economic growth, carbon emission, and fossil fuelsconsumption have also been investigated by many authors [12,13].Studies in these lines of literature rely on sector data for specificcountries and industries [14].
All rights reserved.
The spatial distribution of economic activities is highly unevenwithin a country, which has significant implications for under-standing the trends in carbon dioxide (CO2) emissions if regions inthe country differ in carbon intensity. In fact, variations in bothcarbon intensity and economic growth rates across regions do existand are significant. However, the effects of differential gross do-mestic product growth across regions on the changes in totalemissions between time periods in a country are largely neglectedin the literature. The present paper contributes to understandingthis.
This paper tries to quantify the effects on changes in emissionsof differential growth across regions. To accomplish it, a frameworkis proposed to decompose changes in CO2 emissions into sourcesattributable to 1) national growth rate of GDP, 2) differentialgrowth rates across regions within the country, 3) changes in en-ergy intensity of economy, and 4) changes in CO2 emission coef-ficient of energy. The effects of differential growth may increase ordecrease CO2 emissions. The net effects depend on the distribu-tions of carbon intensity and economic growth rates acrossregions.
This paper focuses on the world’s two largest emitters e Chinaand the United States. We apply our decomposition framework toU.S. data at the state level and Chinese data at the provincial level.Our empirical results highlight the importance of uneven outputgrowth for reducing CO2 emissions. Differential growth acrossprovinces in China saved about 525.46 million tons of CO2 from1996 to 2009. Had all states followed the national pattern of
Table 1Summary statistics.
Variable Year/period Mean Std dev Min Max # Obs.
U.S. statesa
Annual growth rate (%)CO2 emissions 1990e1997 1.677 1.318 �2.072 6.267 51
1997e2009 �0.214 1.034 �2.969 2.293 51Energyconsumption
1990e1997 1.730 1.222 �2.276 4.695 511997e2009 0.203 0.983 �1.361 3.275 51
GRP 1990e1997 3.479 1.767 �0.602 7.833 511997e2009 2.239 0.987 �0.348 4.482 51
Absolute levelCarbon intensity 1990 0.840 0.618 0.065 3.246 51
1997 0.734 0.528 0.065 2.852 512009 0.552 0.368 0.036 1.788 51
Energy intensity 1990 12.458 4.935 2.582 24.803 511997 10.996 4.333 2.740 23.673 512009 8.655 3.364 2.078 17.528 51
C. Wang / Energy 53 (2013) 230e236 231
economic growth, total emissions in the United States during1998e2009 would have been 994.38 million tons more. It has amore important role than changes in carbon coefficient of energywith regards to decreasing emissions and carbon intensity duringsome subperiods in our sample for both countries.
An advantage of the framework is that it allows us to analyzeregional data. It should be noted that the decomposition analysisin this paper is an accounting exercise which generates new re-sults regarding the sources of changes in CO2 emissions. As inprevious studies using sector data, our decomposition results donot provide explanations about fundamental drivers of thechanges.
The organization of the remainder of the paper is as follows. Thenext section describes data and presents some stylized facts. Sec-tion 3 proposes the decomposition framework. Section 4 presentsempirical results. The last section concludes this paper.
Carbon coefficient 1990 0.063 0.023 0.025 0.144 511997 0.063 0.023 0.024 0.149 512009 0.060 0.020 0.017 0.124 51
Chinese provincesb
Annual growth rate (%)CO2 emissions 1995e2004 5.677 3.317 1.389 14.960 29
2005e2009 8.262 3.085 3.527 16.111 30Energyconsumption
1995e2004 6.587 2.844 2.583 13.229 292005e2009 8.100 1.650 4.440 12.247 30
GRP 1995e2004 10.443 1.083 8.558 12.410 292005e2009 13.067 1.632 10.565 18.246 30
Absolute levelCarbon intensity 1995 5.282 2.732 1.399 13.521 29
2005 3.706 2.374 1.577 11.524 302009 3.160 2.127 1.268 10.337 30
Energy intensity 1995 2.710 1.422 0.836 7.752 292005 1.960 1.087 0.941 5.663 302009 1.639 0.912 0.814 4.728 30
Carbon coefficient 1995 1.948 0.264 1.372 2.855 292005 1.857 0.321 1.101 2.511 302009 1.865 0.327 1.315 2.714 30
a For the United States: Carbon intensity is measured in metric tons of carbondioxide per thousand (2000) dollar. Energy intensity in thousand British thermalunits per (2005) U.S. dollar. Carbon coefficient in metric tons of carbon dioxide permillion Btu.
b For China: Carbon intensity is measured in metric tons of carbon dioxide per 104
(1995) RMB yuan. Energy intensity in metric tons of standard coal per 104 (1995)RMB yuan. Carbon coefficient in metric tons of carbon dioxide per metric tons ofstandard coal.
2. Data description and some stylized facts
Data on energy consumption and CO2 emission for U.S. statescome directly from the U.S. Energy Information Administration1
and the Environmental Protection Agency,2 respectively. Realgross domestic product by State in chained (2005) U.S. dollars ispublished by the U.S. Department of Commerce, Bureau of Eco-nomic Analysis, beginning in 1977. Note that gross domesticproduct by State has a structural break in 1997 resulting from theshift from SIC to NAICS in that year. We use the 1997 data from bothsystems.
The unit of observation is the province in China. We constructedprovince-level CO2 emissions estimates based on fossil fuel com-bustion data for the 1995e2009 period. Yearly data on fuel type andconsumption are obtained directly from various editions of ChinaEnergy Statistical Yearbook [15].3 Energy consumption are convertedto CO2 emissions using fuel-specific emissions factors. Similar ap-plications of this approach have been used by many previousstudies [1,16e18].
Table 1 provides summary statistics of the data. We reportaverage annual growth rates of gross regional product (GRP here-after which is obtained from Ref. [19]), regional CO2 emissions andenergy consumption for two subperiods. Absolute values forselected variables are also reported for selected years. It is observedthat growth rates in GRP, energy consumption, and CO2 emissionsvary substantially across regions (states or provinces) in bothcountries. On average, regions in both countries experienced sig-nificant declines in both energy intensity and carbon intensity overthe years. Carbon coefficient did not change much in these timeperiods. Obviously, the pattern of differential growth of GRP acrossregions matters for CO2 emissions because carbon intensity ofregional economies is hugely disparate across regions in bothcountries. If regions in a country with lower levels of carbon in-tensity grow faster than the national level, the country wouldproduce fewer emissions.
To describe the dynamics of carbon intensity distributions, weconduct a simple nonparametric analysis. Let COs;t
2 denote carbonemissions in region s at year t, Ys,t GRP, and Es,t energy consump-tion. Carbon intensity of the economy can be written as theproduct of energy intensity of the economy and carbon coefficientof energy,
1 The website is http://www.eia.gov/state/seds.2 The website is http://www.epa.gov/statelocalclimate/resources/state_
energyco2inv.3 Data for Ningxia in 2000, 2001 and 2002 and for Hainan in 2002 are missing. So
the provinces are excluded from our analyses for these and related years.
COs;t2
Ys;t ¼ Es;t
Ys;t �COs;t
2Es;t
hes;t � cs;t ; (1)
Thus, both changes in energy intensity and carbon coefficientcontribute to the change in carbon intensity of the economy be-tween years s and t,
COs;s2
Ys;s
COs;t2
Ys;t
¼ es;s
es;t� cs;s
cs;t: (2)
The above equation can help us investigate the contributions ofthe two components to the changes in carbon intensity over time.Carbon intensity in year s can be constructed by successivelymultiplying carbon intensity at year t by each of the two compo-nents. To isolate the effect on the distribution dynamics of carbonintensity from the change in energy intensity only, we examinethe counterfactual year s carbon-intensity distribution of thevariable
C. Wang / Energy 53 (2013) 230e236232
CIe ¼ COs;t2s;t � es;s
s;t ;
Y eassuming no changes in carbon coefficient. This counterfactual dis-tribution for China is shown in Panel (A) of Fig. 1 along with theactual distributions in 1995 and 2009.4 Note that if energy intensitydid not change for any region in China, the counterfactual distribu-tion in this panel would be identical to the actual 1995 distribution.With the effect of changes in energy intensity alone, the counter-factual distribution moves very close to the actual 2009 distribution,which indicates that this component primarily accounts for carbonintensity distribution change between 1995 and 2009. The effect ofthe changes in carbon coefficient can be deduced by comparing thecounterfactual distribution in Panel (A) with the 2009 actual distri-bution. The effect of such changes slightly shifts the distributionfurther to the left. However, the shape of the distribution does notchange significantly. Panel (B) considers the effect on distributiondynamics of carbon intensity from changes in carbon coefficientonly. The small shift of the counterfactual distribution from theactual 1995 distribution to the left in Panel (B) confirms that im-provements in carbon coefficient of energy played aminor role in thechanges of carbon intensity of Chinese economy. Analyses on theUnited States data also lead to similar conclusions about the distri-bution dynamics of carbon intensity between the years of 1997 and2009. Panels (C) and (D) illustrate the actual distributions in 1997and 2008 and counterfactual distributions for the United States.
We run a series of simple regressions of output growth ratesbetween 1995 and the later years on the 1995 carbon intensity inChinese provinces. We found that those Chinese provinces withlower level of carbon intensity in 1995 grew faster in later yearswith the exception of 2008 and 2009. Such correlation is not sta-tistically significant. Similar regressions are also run for the U.S.data in 1997 and later years. Negative correlations are also foundand they are statistically significant in most of the years. Generally,the estimated coefficient for the U.S. case is greater (in absolute
COs2
COt2¼
8>><>>:
24Ps
�Ys;ses;tcs;t
�PsðYs;tes;tcs;tÞ
35224PsðYs;ses;scs;sÞ
PsðYs;tes;scs;sÞ
352P
s
�Ys;ses;tcs;s
�PsðYs;tes;tcs;sÞ
Ps
�Ys;ses;scs;t
�PsðYs;tes;scs;tÞ
9>>=>>;
1=6
�
8>><>>:
24Ps
�Ys;tes;scs;t
�PsðYs;tes;tcs;tÞ
35224PsðYs;ses;scs;sÞ
PsðYs;ses;tcs;sÞ
352P
s
�Ys;tes;scs;s
�PsðYs;tes;tcs;sÞ
Ps
�Ys;ses;scs;t
�PsðYs;ses;tcs;tÞ
9>>=>>;
1=6
�
8>><>>:
24Ps
�Ys;tes;tcs;s
�PsðYs;tes;tcs;tÞ
35224PsðYs;ses;scs;sÞ
PsðYs;ses;scs;tÞ
352P
s
�Ys;tes;scs;s
�PsðYs;tes;scs;tÞ
Ps
�Ys;ses;tcs;s
�PsðYs;ses;tcs;tÞ
9>>=>>;
1=6
:
5 Similar decomposition framework has been proposed in the regional literatureto investigate the importance of differential growth. For example, Overman et al.[21] use an additive model to decompose the growth in land occupied by residences
value) than that for China which indicates that regional growth inthe U.S. is more responsive to carbon intensity in base years.
3. The decomposition framework
The previous section suggests that regions within either countrydiffer highly in GRP growth rates and the levels of energy intensityand carbon coefficient. Changes in gross regional product, energy
4 In the estimation of the densities in the two years and all other densities in thispaper, we use the Gaussian kernel function and the Sheather and Jones [20] selectorto determine the “optimal” bandwidth.
intensity, and carbon coefficient certainly affect total CO2 emissionsof the whole country. To quantify the effects of differential growth,we propose a framework which decomposes changes in CO2emissions between years.5 Note that, for region s at year t,
COs;t2 ¼ Ys;t � es;t � cs;t ; (3)
inwhich carbon dioxide emission, COs;t2 , is written as the product of
GRP, Ys,t, energy intensity of the economy in the region, es,t, andcarbon coefficient of energy in the region, cs,t. A simple aggregationover the regions leads to a representation for total emission of thewhole country at year t:
COt2 ¼
Xs
�Ys;t � es;t � cs;t
�: (4)
Then the changes in CO2 emission in the country between twotime periods t and s can be decomposed into three components:output growth across regions, changes in carbon intensity, andchanges in energy intensity,
COs2
COt2¼
PsðYs;s � es;s � cs;sÞ
PsðYs;t � es;t � cs;tÞ : (5)
To isolate the effects on CO2 emission changes between the twotime periods, we decompose it in six different ways with threefactors in each of them. For example, one can write that6
COs2
COt2¼
Ps
�Ys;ses;tcs;t
�PsðYs;tes;tcs;tÞ �
Ps
�Ys;ses;scs;t
�PsðYs;ses;tcs;tÞ �
PsðYs;ses;scs;sÞ
PsðYs;ses;scs;tÞ :
Each of the 18 factors measures the effect of the change in one ofY, e, and c with two others fixed between the two periods. We takethe geometric mean of the six decompositions and rearrange theminto three groups resulting in the following equation7
This decomposition leads to several Fisher ideal index [22].Furthermore, GRP in region s can be rewritten as
Ys;t ¼ Yt � Ys;t=Yt , which is the product of national output Yt and
in the United States and find that the shift of population across states made sig-nificant contributions to the increase in residential land.
6 The other five decomposition is available from the author upon request.7 SeeSeigel [22] forageneralmathematicaldiscussion.Angetal. [23]givesanexample
of application in energy studies. Recent applications can be found in Refs. [24,25].
Fig. 1. Counterfactual distributions of carbon intensity.
C. Wang / Energy 53 (2013) 230e236 233
the share of this region. Let ys,t h Ys,t/Yt. Then we have this region’sgrowth
Ys;s
Ys;t ¼ Ys
Yt �Ys;s
Ys
Ys;t
Yt
hYs
Yt �ys;s
ys;t; (6)
where the first part denotes national pattern of economic growth.The second part gives the differential growth of this region’s GRP.Combining the above two equations, we find that
COs2
COt2¼ Ys
Yt �
8>><>>:
24Ps
�ys;ses;tcs;t
�Psðys;tes;tcs;tÞ
35224Psðys;ses;scs;sÞ
Psðys;tes;scs;sÞ
352P
s
�ys;ses;tcs;s
�Psðys;tes;tcs;sÞ
Ps
�y
Psðy
�
8>><>>:
24Ps
�ys;tes;scs;t
�Psðys;tes;tcs;tÞ
35224Psðys;ses;scs;sÞ
Psðys;ses;tcs;sÞ
352P
s
�ys;tes;scs;s
�Psðys;tes;tcs;sÞ
Ps
�ys;se
Psðys;se
�
8>><>>:
24Ps
�ys;tes;tcs;s
�Psðys;tes;tcs;tÞ
35224Psðys;ses;scs;sÞ
Psðys;ses;scs;tÞ
352P
s
�ys;tes;scs;s
�Psðys;tes;scs;tÞ
Ps
�ys;se
Psðys;se
The four factors in the above decomposition describe the sour-ces for the change in CO2 emissions in the whole country betweentime periods t and s. The four factors are: 1) national growth rate ofGDP; 2) differential growth rates across regions within the country;3) changes in energy intensity; and 4) changes in CO2 emissioncoefficient of energy. It can be easily verified that the indexes satisfyperfect decomposition of the index of change in carbon emissions.If a factor takes a value less than one, it indicates that it helps lowerthe emissions in time period s. Obviously, the last three compo-nents in the above equation are the contributors to changes incarbon intensity of the economy at the national level.
s;ses;scs;t�
s;tes;scs;tÞ
9>>=>>;
1=6
s;scs;t�
s;tcs;tÞ
9>>=>>;
1=6
s;tcs;s�
s;tcs;tÞ
9>>=>>;
1=6
:
Table 2Carbon emissions dioxide growth and its sources in China: 1995e2009.
Year pairs Changes inemissions
NationalGDPgrowth
DifferentialGRP growth
Change inenergyintensity
Change incarboncoefficient
1995e1996 1.0337 1.1185 0.9990 0.9117 1.01471995e1997 1.0462 1.2415 0.9968 0.8324 1.01571995e1998 1.0212 1.3615 0.9950 0.7613 0.99021995e1999 1.0081 1.4812 0.9922 0.7120 0.96331995e2000 1.0556 1.6239 0.9893 0.6986 0.94061995e2001 1.1223 1.7800 0.9886 0.6534 0.97611995e2002 1.2144 1.9745 0.9872 0.6694 0.93081995e2003 1.4006 2.2144 0.9867 0.6839 0.93731995e2004 1.6032 2.5161 0.9883 0.6983 0.92321995e2005 1.9058 2.8485 0.9897 0.7048 0.95921995e2006 2.0763 3.2437 0.9900 0.6853 0.94341995e2007 2.3141 3.7175 0.9895 0.6554 0.95991995e2008 2.4481 4.1637 0.9919 0.6194 0.95701995e2009 2.5763 4.6508 0.9915 0.5868 0.9522
1995e1996 1.0337 1.1185 0.9990 0.9117 1.01471996e1997 1.0121 1.1099 0.9979 0.9131 1.00091997e1998 0.9761 1.0967 0.9979 0.9151 0.97471998e1999 0.9871 1.0879 0.9968 0.9343 0.97631999e2000 1.0475 1.0963 0.9977 0.9785 0.97872000e2001 1.0631 1.0961 0.9993 0.9364 1.03652001e2002 1.0831 1.1085 0.9992 1.0270 0.95222002e2003 1.1421 1.1225 0.9994 1.0121 1.00592003e2004 1.1495 1.1362 1.0004 1.0188 0.99252004e2005 1.1887 1.1321 1.0025 1.0055 1.04172005e2006 1.0895 1.1388 1.0002 0.9691 0.98702006e2007 1.1145 1.1461 1.0006 0.9579 1.01452007e2008 1.0579 1.1200 1.0017 0.9440 0.9989
C. Wang / Energy 53 (2013) 230e236234
The method of decomposing a change in CO2 emissions intocontributing factors provides an advantage over previous tech-niques in its more flexible data requirements that permit theanalysis of cross-regional data. Our method extends the one pro-posed in Ang et al. [23] by singling out the effects of differentialgrowth across regions. It should be noted that the decompositionanalysis in this paper is an accounting exercise which generatesnew results regarding the sources; the decomposition resultsdo not provide explanations about fundamental drivers of thechange.
One can strictly follow the idea of the Kaya identity andextending the proposed decomposition framework by addingpopulation change in it. As this paper mainly focuses on the effectsof output growth and differential growth across regions, we ignorethe effects of population growth in the analysis. Output in thecurrent framework can be regarded as an indicator of human ac-tivities. Furthermore, available data on population for regions inChina is of low quality because it is based on the householdregistration system and we do not have information on labormigration for constructing reliable population data for regions. Inaddition, population growth rates in both countries are relativelysmall in the time periods. It is also useful to examine the effects ofthe changes in the energy mix, particularly regarding renewableenergy.8 We believe that the last one of the four factors from thedecomposition e the effect of changes in carbon coefficient e is arough measure of the composition of energy input in the economy.More careful investigation on the effects is definitely an interestingtopic for future research.
2008e2009 1.0524 1.1170 1.0006 0.9466 0.9947
4. Decomposition results and discussions
In this section, we apply the above decomposition framework toChinese and U.S. data to understand the sources of changes in totalcarbon emissions and carbon intensity of the two economies. Wefocus on the effects of differential output growth across regionswithin each economy.
4.1. Evidence from China
Table 2 reports the sources of changes in CO2 emissions betweenyear pairs in China. The upper panel uses 1995 as the base year anddecomposes changes in emissions between 1995 and each of thefollowing years.9 The first row decomposes emission growth be-tween the years of 1995 and 1996 into four sources: nationalpattern of output growth, differential growth across provinces inChina, changes in energy intensity, and changes in carbon coeffi-cient. In 1996, the level of total emissions in China was about 3.37%higher than that of the previous year. Had all provinces followedthe national pattern of economic growth in China during the twoyears, CO2 emission would have been 0.10% (i.e. 2.71 million tons)higher than its actual 1996 level. Differential output growth con-tributes much more to reducing emissions in the following yearsuntil 2003 when this factor lowered emissions by 49.20 milliontons (i.e. 1.35% of the actual 2003 level). We calculate the changes inemissions between 1995 and each year from 1996 to 2009 and findthat the sum of the changes between 1995 and those years is20,304.18 million tons. Decomposition results from our frameworksuggest that differential growth saved about 528.13 million tons of
8 I thank a reviewer for pointing this out.9 Chongqing City which received provincial status in 1997 is treated as a part of
Sichuan province when 1995 is used as the base year for decomposition. In laterparts of the paper, it is treated as a separate region when 1997 and later years areused as base years.
CO2 from 1996 to 2009 which accounts for about 2.59% (¼528.13/20,304.18) of the total changes over the years.
Results shown in the lower panel of Table 2 are obtained byusing the earlier year in each pair of consecutive years as the baseyear for decomposition analysis. A clear pattern of the effects ofdifferential growth on the changes in CO2 emissions is observed. Itdecreases emissions during each pair of consecutive years before2003. Comparing the magnitudes of the four factors reveals thatdifferential growth is a more important factor than changes incarbon coefficient with regards to reducing emissions in the yearsof 1996, 1997, 2001 and 2003.
China has set a goal of 40e45% reduction from 2005 level incarbon intensity of the economy by 2020. This requires annual rate of3.35e3.91% between 2005 and 2020. Table 2 shows that carbon in-tensity has declined faster than that in most of the years in oursample (especially in years before 2003). While decreases in energyintensity are widely regarded as the key to achieving the goal, wefind that differential growth can also play an important role. Expe-rience in China between 1995 and 2003 suggests that differentialgrowth across regions can contribute much to achieving the goal.
4.2. Evidence from U.S.
In the upper panel of Table 3 for the results from the U.S. data,we use 1997 as the base year and decompose changes in CO2
emissions between 1997 and each of the following years. AggregateCO2 emissions in year 1998 were about 0.09% lower than its level in1997. Had all states followed the national pattern of economicgrowth between the years 1997 and 1998, total CO2 emitted in theU.S. would have been 0.11% (i.e. 6.41 million tons of CO2) more thanits actual 1998 level. A decomposition for the changes in emissionsbetween the years of 1997 and 2006 finds that the 2006 emissionswould be 2.13% more than its actual level if GRP in all states grew atthe national level. The decomposition results also show the
Table 3Carbon dioxide emissions growth and its sources in U.S.: 1990e2009.
Year pairs Changes inemissions
NationalGDPgrowth
DifferentialGRP growth
Change inenergyintensity
Change incarboncoefficient
1997e1998 0.9991 1.0434 0.9989 0.9619 0.99671997e1999 1.0092 1.0929 0.9939 0.9385 0.98991997e2000 1.0386 1.1387 0.9865 0.9273 0.99701997e2001 1.0207 1.1533 0.9859 0.8946 1.00351997e2002 1.0286 1.1732 0.9875 0.8919 0.99541997e2003 1.0376 1.1982 0.9871 0.8764 1.00101997e2004 1.0573 1.2394 0.9864 0.8657 0.99901997e2005 1.0611 1.2741 0.9812 0.8489 1.00001997e2006 1.0477 1.3087 0.9792 0.8223 0.99421997e2007 1.0653 1.3339 0.9793 0.8219 0.99221997e2008 1.0318 1.3295 0.9802 0.8075 0.98051997e2009 0.9610 1.2968 0.9864 0.7815 0.9613
1990e1991 0.9917 0.9984 1.0057 0.9942 0.99331991e1992 1.0190 1.0300 1.0035 0.9831 1.00291992e1993 1.0199 1.0200 1.0039 0.9969 0.99911993e1994 1.0144 1.0466 1.0059 0.9678 0.99561994e1995 1.0105 1.0343 1.0015 0.9870 0.98831995e1996 1.0366 1.0437 0.9997 0.9893 1.00421996e1997 1.0164 1.0523 0.9993 0.9568 1.01021997e1998 0.9991 1.0434 0.9989 0.9619 0.99671998e1999 1.0101 1.0475 0.9953 0.9755 0.99331999e2000 1.0291 1.0419 0.9924 0.9881 1.00722000e2001 0.9828 1.0128 0.9993 0.9647 1.00662001e2002 1.0077 1.0173 1.0019 0.9968 0.99192002e2003 1.0087 1.0213 0.9996 0.9826 1.00562003e2004 1.0190 1.0344 0.9993 0.9878 0.99802004e2005 1.0036 1.0280 0.9952 0.9802 1.00092005e2006 0.9873 1.0272 0.9980 0.9688 0.99412006e2007 1.0168 1.0193 1.0001 0.9995 0.99802007e2008 0.9686 0.9967 1.0010 0.9824 0.98822008e2009 0.9313 0.9754 1.0057 0.9686 0.9802
C. Wang / Energy 53 (2013) 230e236 235
contributions of differential growth across states to the declines ofcarbon intensity of the U.S. economy. Nearly 10% of the declinessince 1997 come from it. For example, carbon intensity in 2000 was8.79% lower than its level in 1997. The effect of differential growthlowered it by about 1.35%.
Data on actual emissions indicate that total changes between1997 and later years are 2017.85 million tons. Further calcula-tion from our decomposition analysis finds that differentialgrowth across states helps lower total emissions in the UnitedStates by 994.38 million tons. The above numbers suggest that,compared to the 1996 national level of emissions, total changesduring 1997e2009 would be 49.28% (¼994.38/2017.85) more ifGRP in all states grew at the national levels. In China, thecontribution from differential growth from 1995 to 2009 is onlyabout 2.59%.
The lower panel in Table 3 shows decomposition results for theU.S. between every pair of consecutive years from 1990 to 2009. Itsuggests that the effects of differential growth have a moreimportant role than changes in carbon coefficient of energywith regards to decreasing emissions and carbon intensity during1996e2001 with the exception of 1999. The effects increasedemissions in the period of 1990e1995, and then decreased emis-sions for the next 12 years with the exception of 2001e2002. Thepattern reversed again during 2006e2007 when differentialgrowth across states increased CO2 emissions.
Aggregate emission in 1991 was about 0.83% lower than its levelin 1990. Had all states grown at the same rate as the national GDPbetween the years 1990 and 1991, CO2 emission would have been0.57% (i.e. 28.56 million tons) less than its actual 1991 level. Itsuggests that the reduced amount of emission between 1990 and1991 would have been 67.69% more should all states followed thenational pattern of economic growth.
Since 1996, states with higher level of carbon intensity of theeconomy have grown at lower rates. The decomposition results inTable 3 suggest that, had all states followed the national pattern ofeconomic growth between the years 1995 and 1996, emissionwould have been 0.03% higher than its actual 1996 level.
We repeat the exercise using a different series of CO2 data onemissions, which are compiled by Blasing et al. [18] who provideestimates of annual CO2 emitted in each state for each year from1960 through 2001. Qualitatively, the same conclusion is reached. Itshows that the effects of differential growth have a more importantrole than carbon coefficient of energy with regards to decreasingemissions during 1983e1988. The effects decreased emissionsduring the period of 1977e1989 with the exception of 1980e1981,and then increased emissions for the next six years. This trend wasagain reversed in 1996 when differential growth helped loweremissions.
5. Conclusions and discussions
To understand the effects on a country’s total carbon dioxideemissions of differential output growth across regions within thecountry, this paper decomposes changes in CO2 emissions intosources attributable to changes in 1) national growth rate of GDP, 2)differential growth rates across regions, 3) changes in energy in-tensity, and 4) changes in CO2 emission coefficient of energy. Thedecomposition framework is applied to China and the U.S. data. Wealso perform a nonparametric statistical analysis to understand theroles played by changes in energy intensity and carbon coefficientin the distribution dynamics of the Carbon-GDP ratio.
Consistent with previous studies, our decomposition results forboth countries find that output growth is a major source for in-creases in carbon dioxide emissions while decline in energy in-tensity is the main contributor to the reduction of emissions.
Several findings about the role of differential growth of regions’economic output are noteworthy. First, the effects of differentialgrowth can increase or decrease carbon dioxide emissions. The neteffects depend on the distributions of carbon intensity and eco-nomic growth rates across regions. Uneven growth across regions isfound to contribute to reducing carbon dioxide emissions in ourdata. Second, the effects of differential growth are not minimal. Ithas a more important role than changes in carbon coefficient ofenergy with regards to decreasing emissions during some sub-periods in our sample for both countries. Third, differential growthhas a relatively more important role in the U.S. than that in China.
The findings of this study point to policy implications. Deci-sionmakers considering climate change policies will benefit frominformation about spatial distributions of carbon emissions and thesources of the changes in emissions. It is important to note that, toachieve a specific amount of reduction in carbon emissions at thenational level, the distribution of such reduction across regions andthe effects of uneven output growth are critical. The national policyis optimal only if it fully considers regional variations in economicgrowth rate, energy intensity and carbon coefficient.
Acknowledgements
This research is funded by EEPSEA/IDRC (Grant No.106612-012).
References
[1] Aldy JE. Divergence in state-level per capita carbon dioxide emissions. LandEconomics 2007;83:353e69.
[2] Lee C-C, Chang C-P. New evidence on the convergence of per capita carbondioxide emissions from panel seemingly unrelated regressions augmentedDickeyeFuller tests. Energy 2008;33:1468e75.
C. Wang / Energy 53 (2013) 230e236236
[3] Kaya Y, editor. Environment, energy, and economy: strategies for sustain-ability. New York: United Nations University Press; 1997.
[4] Ang BW, Pandiyan G. Decomposition of energy-induced CO2 emissions inmanufacturing. Energy Economics 1997;19:363e74.
[5] Lee K, Oh W. Analysis of CO2 emissions in APEC countries: a time-series and across-sectional decomposition using the log mean division method. EnergyPolicy 2006;34:2779e87.
[6] Liu LC, Wang JN, Wu G, Wei YM. China’s regional carbon emissions change over1997e2007. International Journal of Energy and Environment 2010;1:161e76.
[7] Wang C, Chen J, Zou J. Decomposition of energy-related CO2 emission inChina: 1957e2000. Energy 2005;30:73e83.
[8] Wu L, Kaneko S, Matsuoka S. Driving forces behind the stagnancy of China’senergy-related CO2 emissions from 1996 to 1999: the relative importance ofstructural change, intensity change and scale change. Energy Policy 2005;33:319e35.
[9] Zhang Y, Zhang J, Yang Z, Li S. Regional differences in the factors that influenceChina’s energy-related carbon emissions, and potential mitigation strategies.Energy Policy 2011;39:7712e8.
[10] Adom PK, Bekoel W, Amuakwa-Mensah F, Mensah JT, Botchway E. Carbondioxide emissions, economic growth, industrial structure, and technical effi-ciency: empirical evidence from Ghana, Senegal, and Morocco on the causaldynamics. Energy 2012;47:314e25.
[11] Andreonia V, Galmarinic S. Decoupling economic growth from carbon dioxideemissions: a decomposition analysis of Italian energy consumption. Energy2012;44:682e91.
[12] Lotfalipour MR, Falahi MA, Ashena M. Economic growth, CO2 emissions, andfossil fuels consumption in Iran. Energy 2010;35:5115e20.
[13] Pao H-T, Yu H-C, Yang Y-H. Modeling the CO2 emissions, energy use, andeconomic growth in Russia. Energy 2011;36:5094e100.
[14] Ang BW, Zhang FQ. A survey of index decomposition analysis in energy andenvironmental studies. Energy 2000;25:1149e76.
[15] NBS (Department of Industry and Transport Statistics, National Bureauof Statistics), NDRC (Energy Bureau, National Development and ReformCommission). China energy statistical yearbook. Beijing: China StatisticalPress; 2000e2010.
[16] Lutter R. Developing countries’ green house emissions: uncertainty andimplications for participation in the Kyoto protocol. Energy Journal 2000;21:93e120.
[17] Marland G, Boden TA, Andres RJ. Global, regional and national CO2 emissions. Intrends: a compendium of data on global change. Oak Ridge: Oak Ridge NationalLaboratory. Available online at: http://cdiac.esd.ornl.gov/trends/emis/em_cont.htmlhttp://cdiac.esd.ornl.gov/trends/emis/em_cont.html; 2005.
[18] Blasing T, Broniak C, Marland G. Estimates of annual fossil-fuel CO2 emittedfor each state in the U.S.A. and the district of Columbia for each year from1960 through 2000. Oak Ridge, TN, U.S.A.: Trends: A Compendium of Data onGlobal Change Carbon Dioxide, Information Analysis Center, Oak Ridge Na-tional Laboratory, U.S. Department of Energy; 2004.
[19] National Bureau of Statistics. China statistical yearbook. Beijing: China Sta-tistical Press; 1997e2010.
[20] Sheather SJ, Jones MC. A reliable data based bandwidth selection method forkernel density estimation. Journal of Royal Statistical Society, Series B1991;53:683e90.
[21] Overman HG, Puga D, Turner MT. Decomposing the growth in residential landin the United states. Regional Science and Urban Economics 2008;38:487e97.
[22] Seigel IH. The `ideal’ index-number formula. Journal of the American Statis-tical Association 1945;40:520e3.
[23] Ang BW, Liu FL, Chung H. A generalized Fisher index approach to energyproductivity decomposition analysis. Energy Economics 2004;26:757e63.
[24] Wang C. Decomposing energy productivity change: a distance functionapproach. Energy 2007;32:1326e33.
[25] Wang C. Sources of energy productivity growth and its distribution dynamicsin China. Resource and Energy Economics 2011;33:279e92.
