ARTICLE IN PRESS

Energy Policy 33 (2005) 867–871

The LMDI approach to decomposition analysis: a practical guide
B.W. Ang*
Department of Industrial and Systems Engineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260, Singapore

Abstract

In a recent study, Ang (Energy Policy 32 (2004)) compared various index decomposition analysis methods and concluded that the
logarithmic mean Divisia index method is the preferred method. Since the literature on the method tends to be either too technical or
specific for most potential users, this paper provides a practical guide that includes the general formulation process, summary tables
for easy reference and examples.
r 2003 Elsevier Ltd. All rights reserved.

Keywords: Decomposition analysis; Index decomposition analysis; Divisia index; LMDI

1. The LMDI formulation process kth factor on the right-hand side of Eqs. (2) and (3) are
respectively:
Let V be an energy-related aggregate. Assume that !!
X LðV T ; V 0 Þ x T
k;i
there are n factors contributing to changes in V over i i
Dxk ¼ exp ln 0
time and each is associated with a quantifiable variable i
LðV T ; V 0 Þ xk;i
whereby there are n variables, x1 ; x2 ; y; xn : Let sub- X ðV T  V 0 Þ=ðln V T  ln V 0 Þ
i i i i
script i be a sub-category of the aggregate for which ¼ exp T  V 0 Þ=ðln V T  ln V 0 Þ
i
ðV
structural change is to be studied. At the sub-category !!
level the relationship Vi ¼ x1;i x2;i ?xn;i holds. The xTk;i
general index decomposition analysis (IDA) identity is  ln ; ð4Þ
x0k;i
given by
X X !
V¼ Vi ¼ x1;i x2;i ?xn;i : ð1Þ X xTk;i
DVxk ¼ LðViT ; Vi0 Þln
i i x0k;i
0 0 0
P 0
i
!
P from V ¼ i x1;i x2;i yxn;i in
The aggregate changes X ViT  Vi0 xTk;i
period 0 to V T ¼ i xT1;i xT2;i yxTn;i in period T: In ¼ ln 0 ; ð5Þ
multiplicative decomposition, we decompose the ratio: i
ln ViT  ln Vi0 xk;i

Dtot ¼ V T =V 0 ¼ Dx1 Dx2 yDxn : ð2Þ where Lða; bÞ ¼ ða  bÞ=ðln a  ln bÞ as defined in Ang
(2004).2 The general formulae in the formulation process
In additive decomposition we decompose the difference:
are summarized in Table 6 in Appendix A.
DVtot ¼ V T  V 0 ¼ DVx1 þ DVx2 þ ? þ DVxn : ð3Þ
The subscript tot represents the total or overall change
2. Two illustrative cases
and the terms on the right-hand side give the effects
associated with the respective factors in Eq. (1).
Changes in industrial energy consumption may be
In the logarithmic mean Divisia index (LMDI)
studied by quantifying the impacts of changes in three
approach,1 the general formulae for the effect of the
different factors: overall industrial activity (activity
*Tel.: +65-687-422-03; fax: +65-677-714-34. effect), activity mix (structure effect) and sectoral energy
E-mail address: iseangbw@nus.edu.sg (B.W. Ang). intensity (intensity effect). The sub-category of the
1
The LMDI is used here to refer to the logarithmic mean Divisia
2
method I (LMDI I). A related version, the LMDI II, has a weighting For more on Eqs. (4) and (5), see Ang and Liu (2001) and Ang et al.
scheme slightly more complex than LMDI I (Ang et al., 2003). (1998), respectively.

0301-4215/$ - see front matter r 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.enpol.2003.10.010

182 1. the activity share and energy intensity of sector i: From Table 2 Eqs. and CO2 emission Results of industry energy-related CO2 emission decomposition for factors (emission-factor effect). DEtot DEact DEstr DEint Changes in CO2 emissions from industry may be 377. (1) is Aggregate data for Canadian industry. str.963 The subscripts act. by dividing the total emissions for fuel consumption in 3.11 2714. The sub-categories of Canada. The other Dtot ¼ C T =C 0 ¼ Dact Dstr Dint Dmix Demf .8 PJ from 1990 to 2000. sectoral Table 4 energy mix (energy-mix effect). ð6Þ i i Q E i 1990 114.4 From Tables 2 and 3. Q ð¼ i Qi Þ is the total industrial activity level.951 0.044 X X Qi Ei Eij Cij X C¼ Cij ¼ Q ¼ QSi Ii Mij Uij .W. industry activity mix (structure effect). sectoral energy intensity. 4 Due to differences in data source. The IDA identity in Table 1 Eq. 1990–2000: multiplicative decomposition Dtot ¼ E T =E 0 ¼ Dact Dstr Dint . Numerical examples electricity generation by the total final electricity consumption in Canada in the respective years.8 1018. int. 1990–2000: multiplicative decomposition the aggregate are industrial sector and fuel type. Eij is energy sources. Ang / Energy Policy 33 (2005) 867–871 aggregate is industrial sector.814 0.5 295.3 442. industry coverage.31 2336.6 544. we have changes in energy consumption and CO2 emissions are shown in the first column of Tables 2–5.493 0. from Table 6 and they are summarized in Table 7 in 1990–2000: additive decomposition (PJ) Appendix A. . activity structure and sectoral energy intensity. sector We collected the 1990 (Year 0) and 2000 (Year T) classification.2% or The subscripts act. energy consumption in P where Ei ¼ j Eij . 1990 and 2000 X X Qi E i X Year C (MTCO2) E (PJ) Q (gross output.3 The observed Uij ð¼ Cij =Eij Þ: From Eqs. it can be seen that Canadian ¼ DCact þ DCstr þ DCint þ DCmix þ DCemf : ð11Þ industrial energy consumption increased by 16. Results of industrial energy consumption decomposition for Canada. ð10Þ columns of the tables give the decomposition results obtained using the decomposition formulae in DCtot ¼ C T  C 0 Appendix A. Table 3 The LMDI formulae can be readily worked out Results of industrial energy consumption decomposition for Canada. tonnes of CO2 (MTCO2).806 0. 377. industrial activity measurement and decomposition energy and CO2 emission data for industry in Canada technique. respectively.2 2000 135. and Si ð¼ Qi =QÞ and Ii ð¼ Ei =Qi Þ are. (2) and (3). The LMDI formulae Nyboer (2002) and Nyboer and Laurin (2002) do not give CO2 emissions arising from electricity consumption. sectoral energy intensity (intensity effect). We estimated the are summarized in Table 8 in Appendix A.162 1.980 1. mix and emf. ð9Þ ij ij Q Qi Ei Eij ij where C is the total CO2 emissions and Cij is the CO2 database includes a total of 23 industrial sectors and 14 emissions arising from fuel j in industrial sector i.7 96.5 whereP E is the total energy consumption in the industry. The aggregate CO2 emissions in million the consumption of fuel j in industrial sector i.1 studied by quantifying the contributions from changes in five different factors: overall industrial activity (activity effect). 2002). the fuel-mix variable is given petajoules (PJ) and gross industrial output in Canadian by Mij ð¼ Eij =Ej Þ and the CO2 emission factor by dollars (C$) are shown in Table 1. The LMDI decomposition denote the effects associated with overall activity. respectively. 1986 C$ billions) E¼ Ei ¼ Q ¼ QSi Ii . str and int denote the effects associated with the overall activity level. respectively.498 0. The different from those in Natural Resources Canada (2002). the additive decomposition results obtained here are from Nyboer (2002) and Nyboer and Laurin (2002). equivalent emission factors for electricity from the 1990 and 2000 Canadian energy balances in International Energy Agency (1993. (2) and (3). sectoral 3 energy mix and emission factors. ARTICLE IN PRESS 868 B. The Dtot Dact Dstr Dint Dmix Demf IDA identity in Eq. (1) may be written as 1. ð7Þ Dtot Dact Dstr Dint T 0 DEtot ¼ E  E ¼ DEtot ¼ DEact þ DEstr þ DEint : ð8Þ 1. activity structure.

58 6. which can be seen from the formulae in emissions was much lower because of structural change Appendix A.i Multiplicative decomposition Additive decomposition Change scheme Dtot ¼ V T =V 0 ¼ Dx1 Dx2 ?Dxn DVtot ¼ V T  V 0 ¼ DVx1 þ DVx2 þ ? þ DVxn !! ! P ðViT  Vi0 Þ=ðln ViT  ln Vi0 Þ xTk. Changes in energy mix growth observed was due to structural change in arose from a shift towards cleaner fuels in final energy production and a reduction in sectoral energy intensity. CO2 emissions increased by 18.48 5. actual growth in formulation.80 49.31 Table 6 LMDI formulae for the general case with n factors P P IDA identity V ¼ i Vi ¼ i x1.i LMDI formulae Dxk ¼ exp i ln 0 DVxk ¼ i ln 0 ðV T  V 0 Þ=ðln V T  ln V 0 Þ xk. Some LMDI application issues decomposition results show that the activity effect led to an increase almost two and a half times that margin but.80 MTCO2 from 1990 to 2000.i ln ViT  ln Vi0 xk. changes in energy Commercially available spreadsheet software packages .i :::xn.30 2. use while changes in emission factors arose from an Reduction in sectoral energy is often taken as a measure increase in the share of fossil fuels in electricity of improvement in energy efficiency. The LMDI 4. in energy efficiency. and the much lower led to an increase in emissions.i ¼ 0. From Tables 4 and 5. to a lesser extent. reduction in derived once the IDA identity has been specified. 1990–2000: additive decomposition (MTCO2) DCtot DCact DCstr DCint DCmix DCemf 20. the relative contributions of the structural change in reducing energy consumption.i Note: (a) Where xk. between 1010 and 1020. (b) It can be shown that lnðDtot Þ ¼ lnðDx1 Þ þ lnðDx2 Þ þ ? þ lnðDxn Þ: (c) The following relationship holds: DVtot =ln Dtot ¼ DVx1 =ln Dx1 ¼ DVx2 =ln Dx2 ¼ ? ¼ DVxn =ln Dxn : Table 7 LMDI formulae for decomposing changes in industrial energy consumption IDA identity P P Qi Ei P E¼ i Ei ¼ i Q ¼ i QSi Ii Q Qi Multiplicative decomposition Additive decomposition Change scheme Dtot ¼ E T =E 0 ¼ Dact Dstr Dint DEtot ¼ E T  E 0 ¼ DEact þ DEstr þ DEint   T   T P ðEiT  Ei0 Þ=ðln EiT  ln Ei0 Þ Q P EiT  Ei0 Q LMDI formulae Dact ¼ exp i T 0 T 0 ln DEact ¼ i T 0 ln ðE  E Þ=ðln E  ln E Þ Q0 ln Ei  ln Ei Q0  T 0 T 0  T  T 0  T P ðEi  Ei Þ=ðln Ei  ln Ei Þ S P Ei  Ei S Dstr ¼ exp i ln i0 DEstr ¼ i ln i0 ðE T  E 0 Þ=ðln E T  ln E 0 Þ S ln EiT  ln Ei0 S   Ti   Ti  P ðEiT  Ei0 Þ=ðln EiT  ln Ei0 Þ Ii P EiT  Ei0 Ii Dint ¼ exp i ln 0 DEint ¼ i ln 0 ðE T  E 0 Þ=ðln E T  ln E 0 Þ Ii ln EiT  ln Ei0 Ii results show that the activity effect led to an increase mix led to a reduction but changes in emission factors almost three times that margin. sectoral energy intensity. five factors show the importance of the impact of arising from the shift in the composition of industry industry structure change in reducing the growth of CO2 output towards less energy-intensive sectors. The LMDI formulae can be readily in production and.i P ViT  Vi0 xTk. Overall. The impact of generation. An attractive feature of LMDI is the ease of like the case of energy consumption. In addition.2% or 20.W. was emissions despite a substantial increase in the overall estimated to be nearly six times that of improvement industrial output.g. e. Ang / Energy Policy 33 (2005) 867–871 869 Table 5 Results of industry energy-related CO2 emission decomposition for Canada.84 25. replace all the zeros in the data set by a small positive constant. ARTICLE IN PRESS B.i x2.

i. A more likely situation is the occurrence of zero values.W. etc. 2 shows a bar chart for the additive case 1. To overcome this Dmix = 0. Presentation of multiplication decomposition results in converge as the small positive constant approaches zero Table 4. respectively. Fig.980 Dint = 0. The results Fig. This is a limitation of LMDI but in IDA negative values seldom occur.814 Demf =1. -30 countries into regions. e. The LMDI method has several practical advantages from the application viewpoint.0 using the numerical results presented in Tables 4 and 5. 2001).i ¼ 0: In the analysis in Section 3. First. (4) and (5). 1. the LMDI formulae contain logarithmic terms and the variables cannot have 0. e. . 1998). 2. Second. LMDI gives 60 20.182 = Dact Dstr Dint Dmix Demf and the computation could proceed as usual. xk.044 From Eqs.80 49. Estimates of an effect at the sub-group level can be aggregated to give the corresponding effect at the -20 group level. 0. Dtot = 1. i. i. 1 shows a radar chart for the multiplicative case while Fig.0 negative values.e. between 1010 and 1020.58 -6.g.5 tion. the results do not contain 50 an unexplained residual term. Fig. which simplifies the result interpretation. Ang / Energy Policy 33 (2005) 867–871 Table 8 LMDI formulae for decomposing changes in energy-related CO2 emissions from industry IDA identity P P Qi Ei Eij Cij P C¼ ij Cij ¼ ij Q ¼ ij QSi Ii Mij Uij Q Qi Ei Eij Multiplicative decomposition Additive decomposition Change scheme Dtot ¼ C T =C 0 ¼ Dact Dstr Dint Dmix Demf DCtot ¼ C T  C 0 ¼ DCact þ DCstr þ DCint þ DCmix þ DCemf  T !  T P ðCijT  Cij0 Þ=ðln CijT  ln Cij0 Þ Q P CijT  Cij0Q LMDI formulae Dact ¼ exp ij ln DCact ¼ ij ln ðC T  C 0 Þ=ðln C T  ln C 0 Þ Q0 ln CijT  ln Cij0 Q0 T 0 T 0  T ! T 0  T P ðCij  Cij Þ=ðln Cij  ln Cij Þ S P Cij  Cij S Dstr ¼ exp ij ln i0 DCstr ¼ ij ln i0 ðC T  C 0 Þ=ðln C T  ln C 0 Þ Si ln CijT  ln Cij0 Si  T !  T P ðCijT  Cij0 Þ=ðln CijT  ln Cij0 Þ I P CijT  Cij0 Ii Dint ¼ exp ij ln 0 DCint ¼ ij ln 0 ðC T  C 0 Þ=ðln C T  ln C 0 Þ I ln CijT  ln Cij0 Ii !! ! T T 0 P ðCij  Cij Þ=ðln Cij  ln Cij Þ 0 MijT P T Cij  Cij 0 MijT Dmix ¼ exp ij ln DCmix ¼ ij ln ðC T  C 0 Þ=ðln C T  ln C 0 Þ Mij0 ln CijT  ln Cij0 Mij0 !! ! T T 0 P ðCij  Cij Þ=ðln Cij  ln Cij Þ 0 UijT P T Cij  Cij 0 UijT Demf ¼ exp ij ln DCemf ¼ ij ln ðC T  C 0 Þ=ðln C T  ln C 0 Þ Uij0 ln CijT  ln Cij0 Uij0 can be adopted to meet the computational needs and Dact = 1. Presentation of additive decomposition results in Table 5. which makes separate decomposition using = + Cstr + Cint + Cmix + the multiplicative and additive schemes unnecessary.84 -25.e. (Ang et al.g. the results given by the multi.493 present the results in a graphical form. a property useful in multi-level aggregation studies.e. ARTICLE IN PRESS 870 B. DVtot =ln Dtot ¼ DVxk =ln Dxk 10 for all k. 40 plicative LMDI possess the following additive property: 30 lnðDtot Þ ¼ lnðDx1 Þ þ lnðDx2 Þ þ ? þ lnðDxn Þ: Third. LMDI is consistent in aggregation (Ang and Ctot Cact Cemf -10 Liu.48 5. there MTCO 2 exists a simple relationship between multiplicative and 20 additive decomposition.951 problem.31 perfect decomposition. 1. this occurs for sectoral energy mix and CO2 emission factors.. all the zeros in the data set may be replaced by a small positive constant.30 -2. 0 Finally. As an illustra. grouping industry activities into sub-groups.5 Dstr = 0.

Conclusion References As a follow-up to the study by Ang (2004). Energy 26. A new energy decomposition method: approach. Paris.. Canada. 2003... Paris. Decomposition analysis for policymaking in energy: gives a practical guide to the LMDI decomposition which is the preferred method? Energy Policy 32 (in press). Liu. Analysis Centre.L..W. Energy Policy 31. We summarise the general and 537–548. . B. 489–495. Liu. International Energy Agency.P. adopting the approach. Chew. J..W. reported in Section 3. 2002. Canadian Industrial Energy End- use Database and Analysis Centre. and Ang. F.. 2002. J. A.. issues are also dealt with. 2004. B. Canadian Industrial Energy End-use Database and illustrative cases in Section 2 are given in Tables 6–8. B. K. 1998. Simon Fraser University. Ottawa. 2001. Energy efficiency trends in Canada Chuen-Wei Chew in carrying out the calculations 1990–2000. Natural Resources Canada. Energy 23. Canada. Office of Energy Efficiency. Energy Balances of OECD The author wishes to acknowledge the assistance of Countries 1999–2000.Q.W. Canada. F. Simon Fraser University. Development of energy intensity indicators for Canadian industry 1990–2000.W. this paper Ang. OECD. It will be useful to practitioners interested in perfect in decomposition and consistent in aggregation. International Energy Agency. OECD. Nyboer. Ang. Perfect decomposition techniques in energy and environmental analysis. Energy Balances of OECD Acknowledgements Countries 1990–1991.. Laurin. Appendix A Nyboer. Ang / Energy Policy 33 (2005) 867–871 871 5. B... specific LMDI formulae for ease of reference. 2002. F. Choi. E.. Zhang. 1993.. ARTICLE IN PRESS B.W. 2002. 1561–1566. Some application energy and environmental indicators through decomposition. Factorizing changes in present two examples using real data.H.L. Development of greenhouse gas intensity indicators for Canadian industry supplementary 1990–2000: CO2 The LMDI formulae for the general case and the two emissions. Ang.