Specification and aggregation errors in environmentally extended international input-output models*

Transcription

1 19 th Intenational Input-Output Confeence, June 2011, Alexandia, Viginia, USA Specification and aggegation eos in envionmentally extended intenational input-output models* Maaike C. Bouwmeeste a# & Jan Oostehaven a a Univesity of Goningen, Faculty of Economics and Business Postbus 800, 9700 AV Goningen, The Nethelands Abstact An inceasing awaeness of embodied emissions and embodied esouces in taded poducts has esulted in attempts to tack emissions and natual esouce use along the intenational supply chain. An intenational input-output (IIO) model is identified to be the appopiate methodological famewok to undetake this type of envionmental accounting, as diect and indiect, domestic and intenational envionmental impacts can be analyzed within one famewok. In this pape, specification and aggegation eos in envionmental accounting will be studied by means of the EXIOPOL database. Regading the specification eos, we focus on the deviations in envionmental accounting that esult fom (1) assuming that domestic envionmental coefficients can be used to calculate the emissions o esouces embodied in intenational tade, and (2) using a singlecounty famewok vesus using a inte-county famewok to calculate the envionmental impacts. Regading aggegation eos, the EXIOPOL poject has put a majo effot in detailing sectos that ae impotant fom an envionmental point of view. This detail allows us to investigate the eos made when envionmental analysis is done fo moe aggegate industies. Besides, we will compae the IIO model with individual counties, with a model whee counties ae aggegated into one egion, such as the EU27. Moeove, the detail of the database allows us to investigate whethe deviations ae lage fo specific counties o sectos that ae analyzed. We will test the eos made by focusing on CO 2 emissions and wate use. Investigating two quite diffeent envionmental extensions allows fo a genealization of ou findings. Categoy: Wold input-output modeling and databases # Coesponding autho: Tel Fax m.c.bouwmeeste@ug.nl * This wok is pat of the EXIOPOL poject (A New Envionmental Accounting Famewok Using Extenality Data and Input-Output Tools fo Policy Analysis, which is an integated poject funded by the Euopean Union, unde Famewok Pogamme 6, Pioity 6.3 Global Change and Ecosystems, Gant ageement no

2 1 Intoduction Tends of globalization, the ceation of global value chains, stonge intenational inteindusty linkages, and the ise in outsoucing ae all dives in the seach fo intenational compaative data and the constuction of compehensive databases to study these developments. Ove the past decades, concen ove the envionment has steadily isen, and policies ae designed and implemented at all levels of govenance. Some envionmental issues, like global waming though CO 2 emissions, should be addessed at a global level. Othe issues esult in moe localized poblems, like wate shotages, but thei cause can also be global by means intenational tade in intemediate and final poducts. The discussion with egads to the estimation of especially CO 2 emissions has evolved lagely aound measuing the emissions fom a consume pespective as opposed to a poduce pespective. In ode to compute the emissions fom the consume pespective, the emissions embodied in impots ae included, wheeas the emission embodied in expots ae excluded fom the estimates. (G. P. Petes, 2008; G. Petes & Hetwich, 2008; Seano & Dietzenbache, 2010) Including the emissions embodied in impots equies infomation about tading patnes, the poduction technologies of the tading patnes, and the emissions poduced in these counties. Only an envionmentally extended intenational input-output model povides fo this extensive infomation equiement. When using a single-county model fo the calculation of emissions embodied in consumption, the assumption is made that both the envionmental coefficients and the poduction stuctue of the county of the county unde focus moe o less esemble the poduction technologies of all tading patnes. An oveview of models used fo the estimations fom a consumption pespective is given by Wiedmann (2009). The envionmentally extended multi-egional input-output databases compiled in the past decade pomise an impovement ove peviously used methods to estimate envionmental indicatos like pollution in tems of geenhouse gas emissions and use of natual esouces. At the same time, the eos made by appoximating lacking infomation can be studied using the newe geneation of envionmentally extended input-output models. An ealy attempt to do so is Lenzen, Pade, & Munksgaad (2004) in which the authos studied the intenational feedback effects, and hence the eos made when using a single-egion input-output model, in a model including 4 counties and a est of wold egion. They also study the effect of secto aggegation on the estimations. Thei findings suggest that it is impotant to explicitly include tading patne s technologies in tems of poduction stuctue and emission coefficients. Also lage eos ae found when aggegating the sectos to ten sectos. 2

3 In Su, Huang, Ang, & Zhou (2010) it is suggested that a secto detail of aound 40 sectos is sufficient to captue the majoity of CO 2 emissions embodied in poduction. These authos, howeve, use a single county model and only investigate the emissions embodied in expots of China and Singapoe. Combining two databases often equies aggegation o disaggegation of one of the two datasets in ode to match the classifications. Lenzen (2011) addesses the question whethe envionmental data should be aggegated o input-output data should be disaggegated, given that the fist is in a moe detailed classification. Aggegating the envionmental data implies a loss of detail which is undesiable, but disaggegating the input-output data can often only be done with patial data which inceases the uncetainty about the validity of the final dataset. Using Monte-Calo simulations it is shown that disaggegating the input-output data should be pefeed ove aggegating the envionmental data in detemining the input-output multiplies. Andew, Petes, & Lennox (2009) look at the eos in CO 2 estimation intoduced by appoximations of the full model. They find that, in case the numbe of egions in the model is small, a unidiectional tade model can give a easonable appoximation to the full model. Especially including the tade patne that is esponsible fo the lagest shae in emissions embodied in the impots of a county can significantly impove the estimates. They also show that a wold aveage input-output table can offe a suitable substitute fo an aggegate est of the wold table. The authos denounce using a single-county model, although they indicate it is still bette than ignoing impots altogethe. Spatial aggegation is also subject of study by Su & Ang (2010) in which the authos subdivide the data of China into eight egions. They find that the values of CO 2 emissions embodied in expots of China educe as the numbe of egions inceases. In this pape, the specification and aggegation eos made in the calculation of envionmental indicatos when using appoximations in the modelling ae investigated using the EXIOPOL database (Tukke et al., 2009). 2 Methodology 2.1 Data and notation The EXIOPOL database consists of 43 counties; the counties that ae pat of the EU27 and 16 othe lage counties see Appendix 1, fo a list of individual county names. All 43 counties ae linked to each othe though intenational tade flows (Bouwmeeste & Oostehaven, 2008). The level of secto detail in the EXIOPOL database is 129 industies see Appendix 2. Compaed to the 59 by 59 detail level of the ESA-95, the poducts and sectos impotant fom an envionmental point of view have been disaggegated. These 3

4 poducts and sectos ae elated to food and agicultue poducts, metals oes and poducts, mineal poducts, and enegy poducts. (Wood, Hawkins, van Bee, & Poliakov, 2010). In some of ou esults we distinguish fou final demand categoies; household consumption expenditue, govenment consumption expenditue, goss fixed capital fomation, changes in inventoies and valuables plus expots to the est of the wold. Matices ae denoted by bold capitals lettes, vectos by bold lowe case lettes, and scalas by italicized lowe case lettes. A pime indicates tansposed matices and vectos. A hat indicates a diagonalized vecto. The vecto i is a summation vecto with ones, and the identity matix is denoted by I. Indices and paametes used ae: i sectos, fom 1 to I, q final demand categoies, fom 1 to Q, counties, fom 1 to R, summation ove the index at hand. 2.2 The intenational input-output model In an intenational input-output (IO) model, all bilateal intemediate tade flows ae specified by fou indices: the county of oigin and destination, and the industy of oigin and destination. A matix epesentation of the intenational IO model is given in equation (1). x A A A x f x = A A A x + f x A A A x f R R 2 2 R R1 R2 RR R R (1) The vectos x 1 to x R epesent the total output vectos of counties 1 to R. The matix A s shows the impots pe unit of output of county s with egad to county. The matices A 11 to A RR, on the diagonal, show the domestic input coefficient matices. The final demand vectos f eflect the county of oigin, i.e. they show the final demand of all counties fo poducts poduced in. 4

5 In the EXIOPOL database the categoical and geogaphical destination of this final demand is also given. The categoies distinguished ae: household consumption expenditue, govenment consumption expenditue, goss fixed capital fomation, and expots to the est of the wold combined with changes in inventoies and valuables. In the case of the EXIOPOL database, the expots to the est of the wold elate to expots to counties that ae not included in the database. Each of the R final demand vectos shown in equation (1) thus esult fom the following aggegation. 1 2 f = F i + F i + + F R i (2) If the county supescipts ae omitted, system (1) can simply be epesented as: x = Ax + f (3) Its well-known solution is staightfowad (see inte alia (Mille & Blai, 2009)): ( ) 1 x = I A f = Lf (4) The matix (I A) -1 is commonly efeed to as the Leontief-invese and denoted by L. 2.3 An envionmentally extended intenational IO model Any input-output table can be extended with satellite accounts with additional infomation. These ae often elated to the column totals of the IO table, epesenting fo example total labo used, total wate used, o total CO 2 emitted secto by each secto. Dividing, e.g. sectoal emissions, by total sectoal output gives a ow d with emission coefficients, indicating the total sectoal emissions pe unit of output of each secto, in each county. Multiplication with the total output by secto, by county epoduces the diect emissions by secto, by county: e = d x ˆ (5) Whee e is the ow vecto of pollutants emitted o mateial esouces used by each secto, in each county. Total emissions o total mateial use at the wold level is then given by: e = d Lf (6) 5

6 This total e can be disaggegated fom diffeent points of view. To see how, we wite (6) in its fullest possible patitioned fom: R ˆ R R E E E d L L L F F F R ˆ R R E E E 0 d 0 L L L = F F F R1 R2 RR R1 R2 RR R1 R2 RR 0 0 ˆ R E E E d L L L F F F (7) Note that the submatices s ˆ k ks E = d k L F have dimension I x Q, and indicate which pat of the diect emissions by secto i in county is explained by the final demand of categoy q of county s. Hence, matix E diectly combines on all possible causes (types of final demand by county) with all possible impacts (diect emissions o esouce use by secto by county). Aggegation along a ow of E gives the total emissions by secto by county, i.e. an aggegation accoding to the poduce esponsibility pinciple. Aggegation along the columns of E gives the total woldwide emissions caused by the final demand of each categoy in each county, i.e. an aggegation accoding to an extended consume esponsibility pinciple. (G. P. Petes, 2008, G. Petes & Hetwich, 2008, Seano & Dietzenbache, 2010) We use the wod extended as pat of the woldwide emissions has to be allocated to govenment expenditues and pivate investments. 2.4 Specification eos We compae the outcomes of the classical, in ou case extended, consume esponsibility study with the outcome based on the use of the full envionmentally extended intenational IO model of equation (7), thus aggegated by column. In the classical study, typically (e.g. Wyckoff & Roop, 1994 o Lenzen, 1998), the foeign impot coefficient matix (i.e. A s s Leontief-invese (i.e. ) is added to the domestic input coefficient matix (i.e. A ), and the associated ( ) s 1 I A ) is pe-multiplied with the ow domestic emission s coefficients d s. Compaed with the full intenational IO model two specification eos ae made. The fist specification eo is the use of domestic emission coefficients whee foeign emission coefficients should have been used. The second is the use of the single-county Leontief-invese, with impots coefficients included, whee the full intenational Leontiefinvese should have been used. We study these eos consecutively such that thei sum equals to total specification eo. 6

7 2.4.1 Eo of using domestic emission coefficients We measue the effect of using domestic emission coefficients, instead of foeign emission coefficients, by means of the extended intenational IO model. In that way, we obtain the pue specification eo associated with assuming that foeign industies have domestic emission coefficients. We measue the absolute coefficient eos county-by-county, as follows: e R R 1 d L L L f d L L L f d L L L f d L L L f R1 R2 RR R R R1 R2 RR R d L L L f d L L L f R R 2 e = (8) The last tem of (8) indicates the tue total woldwide emission caused by county s total final demand, while the fist tem indicates the estimate of this total, made by assuming that the foeign d s equal the domestic d. Besides absolute county-by-county eos, which will of couse be lage the lage the county, we also study the elated elative coefficient eos. In shothand, these equal: ( ) = ( ) d Lf d Lf d Lf (9) e e e The total elative wold eo then equals: ( ) e e e (10) ε = We pesent it sepaately by causing secto-specific final demand, by means of: ( ) ε = e e e (11) i i i i Whee e i and ẽi ae calculated as: e = d l f and i i i e = d l f (12) i i i Finally, we pesent the eos sepaately by causing categoy-specific final demand eos, with ε q calculated analogously to (11). 7

8 2.4.2 Eo of using a single-county IO model To get a pue estimate of the eo made when using a single-county IO model, as opposed to an intenational IO model, we use the domestic emission coefficients of the county at hand in both cases. The absolute single-county eos ae thus calculated as follows: e ( ) ( ) 1 1 e = d I A f d I A f (13) The last tem of (13) equals the fist tem of (8), and in the fist tem of (13) we use: R A = A + A + + A and 1 2 R f = f + f + + f (14) Besides absolute eos, which vay mainly by county size, we again also pesent elative county-by-county eos, which ae calculated as follows: ( e e ) e ( ) 1 ( ) = d I A f d Lf d Lf (15) Note that these elative single-county eos ae expessed as a atio of the tue value e, instead of by ẽ. In this way, the fist set of elative eos of (9) can be added to the second set of elative eos of (15), to obtain the total elative eo of the classical extended singlecounty IO model compaed to the pesent extended intenational IO model. defined as: The elative single-county specification eo fo woldwide total emissions is ε = ( ) e e e (16) We pesent a disaggegation of this total by causing secto-specific final demand analogously to (11): ε = ( ) e e e (17) i i i i 1 The total coefficient matix A, used hee, deviates fom the IO data published by the individual counties due to the estimation method of the intenational input-output table (see, Bouwmeeste & Oostehaven, 2008, fo details). The county-by-county tade flows have been made consistent, and in the pocess they evalued in basic pices of the poducing (i.e. expoting instead of impoting) county. 8

9 And, the disaggegation by causing categoy-specific of final demand intoε q again poceeds analogously. Finally, in all cases, the total elative specification eo can be decomposed into the specification eos due to using domestic emissions coefficients and those due to using a single-county IO model, espectively, e.g. in the case of the woldwide total emissions: e e e e e e ε = = + e e e (18) 2.5 Aggegation eos Next to specification eos, we investigate aggegation eos. In an intenational IO model aggegation eos can be due to eithe sectoal aggegation o spatial aggegation. In ode to undestand the natue of aggegation eos, it is impotant to be awae that, fo any base yea, equation (6) will always esult in exactly the same estimate of emissions e, no matte the actual level of sectoal o spatial aggegation. This is simply the case, because base yea L times base yea f in (6) always epoduces base yea x at the aggegation level chosen. The same holds fo base yea ˆd times base yea x that always epoduces base yea e, at the chosen level of aggegation. Aggegation eos will only occu when the actual final demand weights ae diffeent fom those of total final demand in the base yea, fom which the models paametes in d and A ae deived, at the aggegation level chosen. To investigate the impact of the aggegation of sectos and counties, we again evaluate the eos fom the extended consume esponsibility pespective. Hence, we will measue the aggegation eos by using the weights of the fou diffeent types of final demand available in the EXIOPOL database, fo the 43 counties distinguished, which esults in 172 diffeent weighting options Sectoal aggegation eos To investigate the impact of the aggegation of sectos, CO 2 emissions and wate use ae calculated at thee levels of sectoal detail. The tue value of the woldwide emissions that ae embodied in final demand categoy q of county, is calculated at the most disaggegate level of 129 industies: 9

10 e q R 1 d L L L f q R 2 d L L L fq = = d Lf R R1 R2 RR R d L L L fq q (19) This esults in Q x R = 172 diffeent tue values. Besides, (19) is calculated at the level of the 59 sectos of the EU27 input-output tables, and at the level of 10 vey aggegate industies, which esults in e (59) and e (10), espectively (see Appendix 3, fo details on q q the sectos that ae combined). We will compae only elative aggegation eos, and do that sequentially, such that the total eo fom aggegating 127 detailed industies to 10 boad sectos equals the sum of the two patial aggegation eos: eq(10) eq eq = eq(10) eq(59) eq + eq(59) eq eq (20) Note that the two patial eos do not necessaily have the same sign. They may compensate each othe, possibly making the total eo even absolutely smalle that than each of the two patial eos. The esults of compaison (20) will be combined and pesented at two levels. Fist, we pesent the effects of aggegation fo each of the 4 final demand categoies, and do that at the level of the 10 boad sectos to see which sectos ae most heavily impacted. Fo this pupose we post-multiply with the diagonal matix of the final demand column used in (19): ˆ e = d Lf (21) q q Whee e q has 10 elements showing the impact of the woldwide final demand categoy q on the value of the wate used o CO 2 emitted by each of the 10 boad sectos. The 3 x 40 outcomes of (21) ae compaed as indicated in (20). Second, we pesent the effects of sectoal aggegation fo each of the 43 counties to see which counties ae most heavily impacted. Fo this pupose we post-multiply with total final demand by county, instead of with the sepaate categoies as was done in (19): e = d Lf (22) The 3 x 43 outcomes of (22) ae compaed as indicated in (20). 10

11 2.5.2 Spatial aggegation eos As we want to compae the impact of spatial aggegation fo each individual county, the spatial aggegation elates to the each time diffeent set of the 42 emaining counties s. Again we pesent only the esults fo total final demand and not those fo each of the 4 categoies sepaately. Hence, the tue values of the woldwide wate use and CO 2 emissions ae calculated with the extended intenational IO model with all 43 individual counties included, as shown in (22). These 43 tue values ae compaed with calculations at two highe levels of aggegation fo the emaining 42 counties: 1. The emaining counties ae split into EU counties and non-eu counties, and both goups of counties ae futhe split into developed and developing ones. The latte subdivision is based on GDP/capita data (see Appendix 4 fo details). This esults in an aggegation into 5 egions, which ae each time a little diffeent. 2. The emaining counties ae aggegated into one Rest of the Wold egion. This esults in an aggegation into 2 egions, again each time a little diffeent. The 3 x 43 outcomes of (22) ae sequentially compaed analogously to (20): e (2) e e = e (2) e (5) e + e (5) e e (23) Whee e is the woldwide wate use o woldwide CO 2 emission allocated to the total final demand of county, as calculated with the full 43 county model, wheeas e (2) and e (5) ae calculated with the aggegated model, with 2 and 5 egions, espectively. 3 Results In this section we compae the diffeences in CO 2 emissions and wate use estimations due to altenative specifications and vaying secto and spatial detail. The analysis is based on the EXIOPOL database consisting of 43 counties. The estimates fo the full model ae consideed to be the tue values. We fist discuss the specification eos made when (1) assuming that domestic envionmental coefficients can be used to calculate the emissions o esouces embodied in intenational tade, and (2) using a single-county famewok vesus using an inte-county famewok to calculate the envionmental impacts. Next, we tun to the diffeences in the esults stemming fom secto and spatial aggegation. 11

12 3.1 Specification eos Table 3-1 epots the specification eos fo each of the 43 EXIOPOL counties. The column with heading e e epesents eo of using domestic emissions coefficients fo all industies both domestic and foeign. The column with ε shows the eo of using domestic emissions coefficients as pecentage of the tue estimates. The column epesenting e e shows the absolute diffeence between the estimate of the single-county model and the domestic coefficients model. The final column of ε shows the eo of the single-county model compaed to an intenational model as pecentage of the tue estimates. Note that fo each county the sum of the values in the columns e e and e e esults in a value fo the total eo e e, which compaable studies geneally focus on, fo example (Andew et al., 2009). Fo the CO 2 estimations a stiking esult is that the eos of using a single-county model (patially) cancel out the eos made fom using domestic emission coefficients fo 21 of the 42 counties. India, the United States, Finland and Geece ae especially notewothy. Using the domestic coefficients in an intenational model, which epesents all the intenational inte-industy linkages, lagely oveestimates the CO 2 emissions fo these counties. This indicates these counties have elatively high CO 2 emissions pe unit of poduction. Howeve, when a single-county model is used, the domestic technology stuctue cancels out the effects of the high CO 2 coefficients. Looking at the esults fo wate use, the impact of using domestic emission coefficients in an intenational model is emakable in tem of pecentage eo fo Noway. Wate is abundant in Noway and wate use is abundantly used by the Nowegian sectos. The esult fo Russia stands out in tems of absolute eo, albeit the pecentage eo is also consideable. Wate use in Japan, South Koea, the Nethelands and the United Kingdom lagely educes when estimated with domestic emission coefficients compaed to the esults of the full model. Table 3-2 shows the esults fo the same specification eos but now by industy. The esults have been calculated at the 129 secto level, but ae epesented at the 59 secto level. Also hee we obseve geneally positive values fo e e and negative values fo e e with egads to the CO 2 emission estimates. Fo wate use the sectos i27 basic metals, i28 fabicated metal poducts, i29 machiney and equipment n.e.c., i65 financial intemediation and i67 activities auxiliay to financial intemediation ae associated with lage pecentage eos of using domestic emissions coefficients in the intenational model. 12

13 Table 3-1: Absolute and elative specification eos, by final demand, by county CO 2 emissions wate use e e ε e e ε e e ε e e ε Austalia Austia Belgium Bazil Bulgaia Canada China Cypus Czech Republic Denmak Estonia Finland Fance Gemany Geece Hungay India Indonesia Ieland Italy Japan Latvia Lithuania Luxemboug Malta Mexico Nethelands Noway Poland Potugal Romania Russian Fedeation Slovak Republic Slovenia South Afica South Koea Spain Sweden Switzeland Taiwan Tukey United Kingdom United States

14 Table 3-2: Absolute and elative specification eos, by final demand, by industy CO 2 emissions wate use e e ε e e ε e e ε e e i01 Agicultue, hunting i02 Foesty, logging i05 Fishing i10 Mining of coal and lignite - peat i11 Cude petoleum and natual gas i12 Mining of uanium and thoium oes i13 Mining of metal oes i14 Othe mining and quaying i15 Food poducts and beveages i16 Tobacco poducts i17 Textiles i18 Weaing appael, fu i19 Leathe poducts i20 Wood, cok and staw poducts i21 Pulp, pape and pape poducts i22 Publishing, pinting, ecoded media i23 Coke, efined petoleum poducts and nuclea fuels i24 Chemicals and chemical poducts i25 Rubbe and plastic poducts i26 Non-metallic mineal poducts i27 Basic metals i28 Fabicated metal poducts i29 Machiney and equipment n.e.c i30 Office machiney and computes i31 Electical machiney and appaatus n.e.c i32 Radio, television and communication equipment i33 Medical, pecision and optical instuments i34 Moto vehicles, tailes and semi-tailes i35 Othe tanspot equipment i36 Funitue; manufactuing n.e.c i37 Recycling i40 Electicity, gas, steam and hot wate supply i41 Collection, puification and distibution of wate i45 Constuction i50 Sale, maintenance and epai of moto vehicles i51 Wholesale tade and commission tade i52 Retail tade; epai pesonal and household goods i55 Hotels and estauants i60 Land tanspot; tanspot via pipelines i61 Wate tanspot i62 Ai tanspot i63 Suppoting tanspot activities; tavel agencies i64 Post and telecommunications i65 Financial intemediation i66 Insuance and pension funding i67 Activities auxiliay to financial intemediation i70 Real estate activities i71 Renting of machiney and equipment i72 Compute and elated activities i73 Reseach and development i74 Othe business activities i75 Public administation and defense i80 Education i85 Health and social wok i90 Sewage and efuse disposal, sanitation i91 Activities of membeship oganisation n.e.c i92 Receational, cultual and spoting activities i93 Othe sevice activities i95 Pivate households with employed pesons ε 14

15 3.2 Aggegation eos Secto aggegation Table 3-3 shows the diffeences in estimated CO 2 emissions and wate use when educing secto detail fom 129 sectos to 59 sectos and to 10 sectos. The esults ae pesented by secto (at the 10 secto level) and by final demand categoy. The esults fo CO 2 emissions ae athe mixed showing no distinctive cases. The lage pecentage eos found fo the aggegation fom 129 to 59 sectos fo final demand categoy 4: changes in inventoies and valuables and expots to the est of the wold seem to indicate a data constuction issue. Exteme pecentage eos ae found fo wate use, especially in the aggegation fom 59 to 10 sectos. Secto E Electicity, gas and wate supply shows the lagest pecentage eos. Secto F Constuction and secto I Tanspot, stoage and communication ae also associated with lage pecentage eos fo the aggegation to 10 sectos fo household consumption expenditue, govenment consumption expenditue and goss fixed capital fomation. Wate use pecentage eos esulting fom the aggegation to 10 sectos ae elatively lage fo all sectos satisfying the final demand categoy goss fixed capital fomation Table 3-4 epots the pecentage eos esulting fom secto aggegation fo the 43 EXIOPOL counties. The lagest eos fo CO 2 emissions ae elated to the aggegation to 10 sectos. The oveestimation of CO 2 emissions at the 10 secto level is highest fo Geece, Noway and the Russian Fedeation. Fo wate use the pecentage eos obtained fom aggegating to 59 sectos patially cancel when aggegation to 10 sectos. 15

16 Table 3-3: Sectoal aggegation eos, by final demand, by impacted secto CO 2 emissions wate use ε (129 59) ε (59 10) ε (129 10) ε (129 59) ε (59 10) ε (129 10) Final demand 1: household consumption ABC D E F GH I JK L MN OPQ Final demand 2: govenment consumption ABC D E F GH I JK L MN OPQ Final demand 3: goss fixed capital fomation ABC D E F GH I JK L MN OPQ Final demand 4: changes in inventoies and valuables; expot to the est of the wold ABC D E F GH I JK L MN OPQ

17 Table 3-4 Sectoal aggegation eos, by final demand, by county CO 2 emissions wate use ε (129 59) ε (59 10) ε (129 10) ε (129 59) ε (59 10) ε (129 10) Austalia Austia Belgium Bazil Bulgaia Canada China Cypus Czech Republic Denmak Estonia Finland Fance Gemany Geece Hungay India Indonesia Ieland Italy Japan Latvia Lithuania Luxemboug Malta Mexico Nethelands Noway Poland Potugal Romania Russian Fedeation Slovak Republic Slovenia South Afica South Koea Spain Sweden Switzeland Taiwan Tukey United Kingdom United States

18 3.2.2 Spatial aggegation Table 3-5 shows the pecentage eos esulting fom spatial aggegation. Each counties CO 2 emissions and wate use fom the full model ae compaed to the estimations of each county in a setting with fou othe egions, and in a setting with one lage est of wold table. The egions and est of wold table both esult fom the aggegation of the emaining 42 patnes of the county in focus. The egions fomed in the setting combining the county table with fou sets of counties can be found in Appendix 4. The CO 2 emissions which ae oveestimated when educing the intenational model to one county and a est of wold occu when the fou egions ae combined into a est of wold table. In contast, the CO 2 emissions that ae undeestimated at the most aggegated spatial level show the lage eo at the aggegation of the individual counties to the egions. The eos in the estimation of wate use ae moe exteme compaed to the spatial aggegation eos of the CO 2 estimates. Noway, Luxemboug en Switzeland stand out with egads to the oveestimation of the wate use as the tading patnes of the focus county become moe aggegated. Again, undeestimations made occu at the aggegation of the individual counties to the fou geneal egions, wheeas the oveestimation occus mostly at the aggegation of the fou egions into a est of the wold egion. 18

19 Table 3-5 Spatial aggegation eos, by final demand, by county CO 2 emissions wate use ε (43 5) ε (5 2) ε (43 2) ε (43 5) ε (5 2) ε (43 2) Austalia Austia Belgium Bazil Bulgaia Canada China Cypus Czech Republic Denmak Estonia Finland Fance Gemany Geece Hungay India Indonesia Ieland Italy Japan Latvia Lithuania Luxemboug Malta Mexico Nethelands Noway Poland Potugal Romania Russian Fedeation Slovak Republic Slovenia South Afica South Koea Spain Sweden Switzeland Taiwan Tukey United Kingdom United States

20 Refeences Andew, R., Petes, G. P., & Lennox, J. (2009). Appoximation and egional aggegation in multi-egional input-output analysis fo national cabon footpint accounting. Economic Systems Reseach, 21(3), 311. Bouwmeeste, M. C., & Oostehaven, J. (2008). Methodology fo the constuction of an intenational supply-use table. Pape pesented at the IIOA intemediate confeence, Seville, Spain. Lenzen, M. (1998). Pimay enegy and geenhouse gases embodied in austalian final consumption: An input output analysis. Enegy Policy, 26(6), Lenzen, M. (2011). Aggegation vesus disaggegation in input-output analysis of the envionment. Economic Systems Reseach, 23(1), 73. Lenzen, M., Pade, L., & Munksgaad, J. (2004). CO 2 multiplies in multi-egion input-output models. Economic Systems Reseach, 16(4), Mille, R. E., & Blai, P. D. (2009). Input-output analysis: Foundations and extensions. Cambidge Univesity Pess. Petes, G. P. (2008). Fom poduction-based to consumption-based national emission inventoies. Ecological Economics, 65(1), Petes, G., & Hetwich, E. (2008). Post-kyoto geenhouse gas inventoies: Poduction vesus consumption. Climatic Change, 86(1), Seano, M., & Dietzenbache, E. (2010). Responsibility and tade emission balances: An evaluation of appoaches. Ecological Economics, 69(11), Su, B., & Ang, B. W. (2010). Input output analysis of CO2 emissions embodied in tade: The effects of spatial aggegation. Ecological Economics, 70(1), Su, B., Huang, H. C., Ang, B. W., & Zhou, P. (2010). Input output analysis of CO2 emissions embodied in tade: The effects of secto aggegation. Enegy Economics, 32(1), Tukke, A., Poliakov, E., Heijungs, R., Hawkins, T., Neuwahl, F., Rueda-Cantuche, J. M., Giljum, S., Moll, S., Oostehaven, J., & Bouwmeeste, M. (2009). Towads a global 20

21 multi-egional envionmentally extended input-output database. Ecological Economics, 68(7), Wiedmann, T. (2009). A eview of ecent multi-egion input-output models used fo consumption-based emission and esouce accounting. Ecological Economics, 69(2), Wood, R., Hawkins, T., van Bee, T., & Poliakov, E. (2010). Development of hamonized supply and use tables fo the EXIOPOL database, and, consumption activities and waste in EXIOBASE, EXIOPOL Deliveable DIII.2.a, DIII.3.a, DIII.2.c.2 and DIII.2.c.3) Wyckoff, A. W., & Roop, J. M. (1994). The embodiment of cabon in impots of manufactued poducts: Implications fo intenational ageements on geenhouse gas emissions. Enegy Policy, 22(3),

22 Appendices Appendix 1: EXIOPOL county list The 43 EXIOPOL counties the 27 EU membe counties ae maked Austalia Hungay (EU) South Koea Austia (EU) India Romania (EU) Belgium (EU) Indonesia Russian Fedeation Bulgaia (EU) Ieland (EU) Slovakia (EU) Bazil Italy (EU) Slovenia (EU) Canada Japan South Afica China Latvia (EU) Spain (EU) Cypus (EU) Lithuania (EU) Sweden (EU) Czech Republic (EU) Luxemboug (EU) Switzeland Denmak (EU) Malta (EU) Taiwan Estonia (EU) Mexico Tukey Finland (EU) Nethelands (EU) United States Fance (EU) Noway United Kingdom (EU) Gemany (EU) Poland (EU) Geece (EU) Potugal (EU) 22

23 Appendix 2: EXIOPOL secto classification i01.a Cultivation of paddy ice i01.b Cultivation of wheat i01.c Cultivation of ceeal gains nec i01.d Cultivation of vegetables, fuit, nuts i01.e Cultivation of oil seeds i01.f Cultivation of suga cane, suga beet i01.g Cultivation of plant-based fibes i01.h Cultivation of cops nec i01.i Cattle faming i01.j Pigs faming i01.k Poulty faming i01.l Meat animals nec i01.m Animal poducts nec i01.n Raw milk i01.o Wool, silk-wom cocoons i02 Foesty, logging and elated sevice activities (02) i05 Fishing, opeating of fish hatcheies and fish fams; sevice activities incidental to fishing (05) i10 Mining of coal and lignite; extaction of peat (10) i11.a Extaction of cude petoleum and sevices elated to cude oil extaction, excluding suveying i11.b Extaction of natual gas and sevices elated to natual gas extaction, excluding suveying i11.c Extaction, liquefaction, and egasification of othe petoleum and gaseous mateials i12 Mining of uanium and thoium oes (12) i13.1 Mining of ion oes i Mining of coppe oes and concentates i Mining of nickel oes and concentates i Mining of aluminum oes and concentates i Mining of pecious metal oes and concentates i Mining of lead, zinc and tin oes and concentates i Mining of othe non-feous metal oes and concentates i14.1 Quaying of stone i14.2 Quaying of sand and clay i14.3 Mining of chemical and fetilize mineals, poduction of salt, othe mining and quaying n.e.c. i15.a Pocessing of meat cattle i15.b Pocessing of meat pigs i15.c Pocessing of meat poulty i15.d Poduction of meat poducts nec i15.e Pocessing vegetable oils and fats i15.f Pocessing of daiy poducts i15.g Pocessed ice i15.h Suga efining i15.i Pocessing of Food poducts nec i15.j Manufactue of beveages i15.k Manufactue of fish poducts i16 Manufactue of tobacco poducts (16) i17 Manufactue of textiles (17) i18 Manufactue of weaing appael; dessing and dyeing of fu (18) i19 Tanning and dessing of leathe; manufactue of luggage, handbags, saddley, haness and footwea (19) Manufactue of wood and of poducts of wood and cok, except funitue; manufactue of aticles of i20 staw and plaiting mateials (20) i21 Manufactue of pulp, pape and pape poducts (21) i22 Publishing, pinting and epoduction of ecoded media (22) i23.1 Manufactue of coke oven poducts i23.20.a Manufactue of moto spiit (gasoline) i23.20.b Manufactue of keosene, including keosene type jet fuel i23.20.c Manufactue of gas oils i23.20.d Manufactue of fuel oils n.e.c. i23.20.e Manufactue of petoleum gases and othe gaseous hydocabons, except natual gas i23.20.f Manufactue of othe petoleum poducts i23.3 Pocessing of nuclea fuel i24 Manufactue of chemicals and chemical poducts (24) i25 Manufactue of ubbe and plastic poducts (25) i26.a Manufactue of glass and glass poducts i26.b Manufactue of ceamic goods i26.c Manufactue of bicks, tiles and constuction poducts, in baked clay 23

24 i26.d Manufactue of cement, lime and plaste i26.e Manufactue of othe non-metallic mineal poducts n.e.c. i27.a Manufactue of basic ion and steel and of feo-alloys and fist poducts theeof i27.41 Pecious metals poduction i27.42 Aluminum poduction i27.43 Lead, zinc and tin poduction i27.44 Coppe poduction i27.45 Othe non-feous metal poduction i27.5 Casting of metals i28 Manufactue of fabicated metal poducts, except machiney and equipment (28) i29 Manufactue of machiney and equipment n.e.c. (29) i30 Manufactue of office machiney and computes (30) i31 Manufactue of electical machiney and appaatus n.e.c. (31) i32 Manufactue of adio, television and communication equipment and appaatus (32) i33 Manufactue of medical, pecision and optical instuments, watches and clocks (33) i34 Manufactue of moto vehicles, tailes and semi-tailes (34) i35 Manufactue of othe tanspot equipment (35) i36 Manufactue of funitue; manufactuing n.e.c. (36) i37.1 Recycling of metal waste and scap i37.2 Recycling of non-metal waste and scap i40.11.a Poduction of electicity by coal i40.11.b Poduction of electicity by gas i40.11.c Poduction of electicity by nuclea i40.11.d Poduction of electicity by hydo i40.11.e Poduction of electicity by wind i40.11.f Poduction of electicity nec, including biomass and waste i40.12 Tansmission of electicity i40.13 Distibution and tade of electicity i40.2 Manufactue of gas; distibution of gaseous fuels though mains i40.3 Steam and hot wate supply i41 Collection, puification and distibution of wate (41) i45 Constuction (45) Sale, maintenance, epai of moto vehicles, moto vehicles pats, motocycles, moto cycles pats and i50.a accessoies i50.b Retail sale of automotive fuel i51 Wholesale tade and commission tade, except of moto vehicles and motocycles (51) i52 Retail tade, except of moto vehicles and motocycles; epai of pesonal and household goods (52) i55 Hotels and estauants (55) i60.1 Tanspot via ailways i60.2 Othe land tanspot i60.3 Tanspot via pipelines i61.1 Sea and coastal wate tanspot i61.2 Inland wate tanspot i62 Ai tanspot (62) i63 Suppoting and auxiliay tanspot activities; activities of tavel agencies (63) i64 Post and telecommunications (64) i65 Financial intemediation, except insuance and pension funding (65) i66 Insuance and pension funding, except compulsoy social secuity (66) i67 Activities auxiliay to financial intemediation (67) i70 Real estate activities (70) i71 Renting of machiney and equipment without opeato and of pesonal and household goods (71) i72 Compute and elated activities (72) i73 Reseach and development (73) i74 Othe business activities (74) i75 Public administation and defense; compulsoy social secuity (75) i80 Education (80) i85 Health and social wok (85) i90.01 Collection and teatment of sewage i90.02.a Collection of waste i90.02.b Incineation of waste i90.02.c Landfill of waste i90.03 Sanitation, emediation and simila activities i91 Activities of membeship oganization n.e.c. (91) i92 Receational, cultual and spoting activities (92) i93 Othe sevice activities (93) i95 Pivate households with employed pesons (95) i99 Exta-teitoial oganizations and bodies 24

25 Appendix 3: Secto aggegation NACE Rev. 1.1 classification* EXIOPOL sectos A Agicultue, hunting and foesty i01 i02 B Fishing i05 C Mining and quaying i10 i11 i12 i13 i14 D Manufactuing i15 i16 i17 i18 i19 i20 i21 i22 i23 i24 i25 i26 i27 i28 i29 i30 i31 i32 i33 i34 i35 i36 i37 E Electicity, gas and wate supply i40 i41 F Constuction i45 G Wholesale and etail tade; epai of moto vehicles, i50 i51 i52 motocycles and pesonal and H Hotels and estauants i55 I Tanspot, stoage and communication i60 i61 i62 i63 i64 J Financial intemediation 65 i66 i67 K Real estate, enting and business activities i70 i71 i72 i73 i74 L Public administation and defense; compulsoy social i75 secuity M Education i80 N Health and social wok i85 O Othe community, social and pesonal sevice activities i90 i91 i92 i93 P Activities of households i95 i96 i97 Q Exta-teitoial oganizations and bodies i99 * A desciption of the NACE Rev. 1.1 classification can be found at: Nom=NACE_1_1&StLanguageCode=EN 25

26 Appendix 4: Spatial aggegation EU-high PPP int$ EU-low PPP int$ Luxemboug Cypus Nethelands Geece Denmak Malta Austia Potugal Ieland Slovenia Sweden Czech Republic Belgium Hungay United Kingdom Slovak Republic Gemany Poland Finland Estonia 9882 Italy Lithuania 8602 Fance Latvia 8031 Spain Bulgaia 6301 Romania 5654 Othe-high PPP int$ Othe-low PPP int$ Noway Mexico 9201 United States Tukey 8867 Switzeland Bazil 7021 Canada Russian Fedeation 6824 Austalia South Afica 6773 Japan Indonesia 2417 Taiwan China 2364 Koea, Rep India 1574 Data on GDP pe capita, PPP (cuent intenational $) fo 2000 Wold Development Indicatos, accessed Souce: Wold Bank national accounts data, and OECD National Accounts data files. 26