GTAP-E: An Energy-Environmental Version of the GTAP Model

Transcription

1 GTAP-E: An Enegy-Envionmental Vesion of the GTAP Model Jean-Mac Buniaux* and Tuong P. Tuong ** GTAP Technical Pape No. 16 Revised Januay 2002 Jean-Mac Buniaux was on leave fom the OECD as visiting Associate Pofesso with the Cente fo Global Tade Analysis, Pudue Univesity, West Lafayette, Indiana when this pape was witten. Tuong P. Tuong was visiting Associate Pofesso with the Cente fo Global Tade Analysis, Pudue Univesity, West Lafayette, Indiana, on leave fom the School of Economics, Univesity of New South Wales, Sydney NSW 2052, Austalia. This technical pape is a evised vesion of an ealie GTAP Technical Pape witten by T.P. Tuong (Tuong, 1999). GTAP stands fo the Global Tade Analysis Poject which is administeed by the Cente fo Global Tade Analysis, Pudue Univesity, West Lafayette, Indiana, USA Fo moe infomation about GTAP, please efe to ou Woldwide Web site at: o send a equest to conne@agecon.pudue.edu

2 GTAP-E: An Enegy-Envionmental Vesion of the GTAP Model Jean-Mac Buniaux and Tuong P. Tuong GTAP Technical Pape No. 16 Abstact Enegy is an impotant commodity in many economic activities. Its usage affects the envionment via CO 2 emissions and the Geenhouse Effect. Modeling the enegy-economy-envionment-tade linkages is an impotant objective in applied economic policy analysis. Peviously, howeve, the modeling of these linkages in GTAP has been incomplete. This is because enegy substitution, a key facto in this chain of linkages, is absent fom the standad model specification. This technical pape emedies this deficiency by incopoating enegy substitution into the standad GTAP model. It begins by fist eviewing some of the existing appoaches to this poblem in contempoay CGE models. It then suggests an appoach fo GTAP which incopoates some of these desiable featues of enegy substitution. The appoach is implemented as an extended vesion of the GTAP model called GTAP-E. In addition, GTAP-E incopoates cabon emissions fom the combustion of fossil fuels and this evised vesion of GTAP-E povides fo a mechanism to tade these emissions intenationally. The policy elevance of GTAP-E in the context of the existing debate about climate change is illustated by some simulations of the implementation of the Kyoto Potocol. It is hoped that the poposed model will be used by individuals in the GTAP netwok who may not be themselves enegy modeles, but who equie a bette epesentation of the enegy-economy linkages than is cuently offeed in the standad GTAP model. The authos ae indebted to Tom Hetel fo oiginally suggesting the topic of this study, and fo his continued suppot and encouagement. Thanks ae due also to Kevin Hanslow, and Mustafa Babike fo poviding many helpful comments and suggestions on an ealie daft. Ken Peason was quite indispensable in poviding the technical advice fo the illustative expeiments. Finally, note that this pape is a evised vesion of the GTAP Technical Pape No. 16 by Tuong (see Tuong, 1999).

3 Table of Contents 1. Intoduction Review of existing appoaches The CETM model - Ruthefod et al. (1997) The stuctue of CETM The Linkage of ETA to MACRO Comments on the stuctue of CETM The MEGABARE Model and the Technology Bundle' Appoach Desciption of the technology bundle appoach Comments on the technology bundle appoach The OECD S GREEN model Dynamics in GREEN Inte-fuel substitution Fuel-facto substitution Comments on the GREEN model The Babike-Maskus-Ruthefod (BMR) model Boges and Goulde (1984) model Towads a GTAP model with enegy substitution Top-down vesus bottom-up appoach The issue of enegy-capital substitutability o complementaity Impotance of the issue Empiical estimates of F EK The stuctue of inte-fuel and fuel-facto substitution in GTAP-E Poduction stuctue with enegy substitution Consumption stuctue Illustative scenaios Altenative implementations of the Kyoto Potocol ii

4 4.2 Macoeconomic esults Conclusion Refeences Annexes Annex 1 : Geneal equilibium elasticities in GTAP-E and GTAP. Annex 2 : Specifying county-specific cabon eductions with no emission tading in GTAP-E Annex 3 : Specifying emission tading in GTAP-E Figues Figue 1: Stuctue of CETM... 3 Figue 2: MACRO poduction nest... 4 Figue 3: MACRO consumption nest... 4 Figue 4: ETA - MACRO linkage... 8 Figue 5: Technology bundle appoach... 9 Figue 6: Composition of the technology bundle fo the electicity industy Figue 7: Composition of the technology bundle fo the steel industy Figue 8: The stuctue of poduction in GREEN Figue 9: Enegy and backstop technologies in GREEN Figue 10: The stuctue of household demand in GREEN Figue 11: Substitution elasticity when total output is held constant Figue 12: Stuctue of poduction in the Babike-Maskus-Ruthefod (1997) model Figue 13: Stuctue of final demand in the Babike-Maskus-Ruthefod (1997) model Figue 14: Stuctue of poduction in Boges and Goulde (1984) model Figue 15: Standad GTAP poduction stuctue Figue 16: GTAP E poduction stuctue Figue 17: GTAP E capital-enegy composite stuctue Figue 18: GTAP-E govenment puchases Figue 19: GTAP-E household pivate puchases Figue 20: Emission tading among Annex 1 counties Figue 21: Woldwide emission tading iii

5 Figue 22: Welfae decomposition of implementing the Kyoto Potocol with no use of the flexibility mechanisms Figue 23: Welfae decomposition of implementing the Kyoto Potocol with tading among Annex 1 counties Figue 24 : Welfae decomposition of implementing the Kyoto Potocol with woldwide emission tading Tables Table 17.1: List of Technologies in ETA... 6 Table 17.2: List of Impotant Equations in ETA... 8 Table 3 Summay Chaacteistics of CETM Table 4: Summay Chaacteistics of MEGABARE Table 5: Summay Chaacteistics of GREEN Table 6: Summay Chaacteistics of BMR Model Table 7: Summay Chaacteistics of Boges and Goulde Model Table 8: Estimates of the patial Hicks-Allen elasticities of substitution (F) and facto shaes (F) Table 9: Enegy substitution elasticities in GTAP-E and othe models Table 10: Elasticities of substitution between diffeent factos of poduction Table 11: The elationship between inne (F KE-inne ) and oute (F KE-oute ) elasticities of substitution fo the cases of Japan and the US Table 12: Elasticities of substitution between domestic and foeign souces (F D ) Table 13: Elasticities of substitution between diffeent egions (F M ) Table 14: Maginal costs of achieving the Kyoto tagets with and without using the flexibility mechanisms Table 15: Macoeconomic impacts of implementing the Kyoto Potocol iv

6 GTAP-E: Incopoating Enegy Substitution into GTAP Model 1. Intoduction Enegy is an impotant commodity in many economic activities. Its usage affects the envionment via CO 2 emissions and the Geenhouse Effect. Modeling the enegy-economy-envionment-tade linkages is an impotant objective in applied economic policy analysis. Up to now, howeve, the modeling of these linkages in GTAP has been incomplete. This is because enegy substitution, a key facto in this chain of linkages, is absent fom the standad model specification. This pape emedies this deficiency by incopoating enegy substitution into the standad GTAP model. It begins by fist eviewing some of the existing appoaches to this poblem in contempoay CGE models. It then suggests an appoach fo GTAP which incopoates some of these desiable featues of enegy substitution. The appoach is implemented as an extended vesion of the GTAP model called GTAP-E. In addition, GTAP-E incopoates cabon emissions fom the combustion of fossil fuels as well as a mechanism to tade these emissions intenationally. The policy elevance of GTAP-E in the context of the existing debate about climate change is illustated by some illustative simulations of the implementation of the Kyoto Potocol. This technical pape is a evised vesion of a ealie pape witten by T.P. Tuong (Tuong, 1999). Compaed with this vesion, the model used hee is deived fom the vesion 6.1 of the GTAP model based on 1997 data (vesion 5 of the GTAP data base). In addition to inte-fuel substitution, this model incopoates some futhe impovements, such as the computation of a Social Account Matice (SAM) which povides a full account of the cabon tax evenues and expenditues and a moe specific teatment of cabon emission tading. 2. Review of Existing Appoaches In this section, we eview some of the existing appoaches to incopoating enegy substitution into AGE models. The pupose of this section is not to undetake an exhaustive eview of the liteatue, but athe, to select some typical appoaches and examine thei impotant featues fo possible incopoation into the GTAP model. Thee ae thee main models to be consideed in this section, and these ae: (1) the CETM model by Ruthefod et al. (1997), (2) the MEGABARE model by ABARE (1996), and (3) the OECD s GREEN model by Buniaux et al. (1992). Some othe models ae also consideed in sub-section

7 2.1 The CETM Model - Ruthefod et al. (1997) This model epesents an attempt to bidge the gap between the (top down) economic models often used by economists, and the (bottom-up) pocess models used by enginees and envionmentalists in studying the effect of enegy policies on the envionment. Recognizing that full integation of these two types of models is methodologically and computationally difficul the authos of CETM attempted a patial link. This means, fistly, the constuction of a patial equilibium pocess model of the enegy secto (ETA) (which is based on the MERGE model of Manne and Richels (1996)). The model is then linked to a geneal equilibium model called MACRO. The pocess of linking the two sub-models is though the pocess of passing the enegy pice and quantity vaiables between the two sub-models and iteation until the input efeence quantities fom ETA ae close to the solutions of the MACRO model (Ruthefod et al (1997, p6)). In light of the fact that the enegy secto makes up only a small faction (less than 5%) of the goss output of most economies, convegence of the two sets of esults fom ETA and MACRO is consideed most likely. This is because if enegy is only a small pat of the industy cost stuctue then the changes in the pices and quantities of enegy demand within ETA will affect only maginally the oveall esults of industy costs and pices within MACRO. This means convegence of the two sets of esults fom ETA and MACRO can be achieved though an iteation pocess as descibed above, athe than by having to solve the optimization poblems of the two sub-models simultaneously The Stuctue of CETM The stuctue of CETM is descibed in Figue 1. Within this stuctue, the MACRO sub-model is a conventional computable geneal equilibium (CGE) model, which has 5 intenationally taded commodities and five industies: Y - Othe manufactues and sevices, NFM = Non-feous metals, PPP = Pulp and pape, TRN = Tanspot industies, OTH = Othe enegy intensive sectos. The fist industy is an aggegate of non-enegy intensive industies, and the othe fou epesent enegy-intensive industies. Factos of poduction include: land, labo, capital, electicity, and non-electic enegy. The latte two enegy inputs ae linked to ETA. Thee ae nine egions in MACRO: USA, JAPAN, CANZ (Canada, Austalia, New Zealand), OECDE (Othe OECD), CHINA, INDIA, EFFSU (Easten Euope and Fome Soviet Union), MOPEC (Mexico and OPEC counties), and ROW (The est of the wold). With eleven ten-yea time peiods, this model begins the peiod of simulation fom 1990 (benchmak yea) and ends in The stuctue of industy poduction in MACRO is as descibed in Figue 2. Fis capital and labo ae combined via a Cobb-Douglas poduction function 1. So ae electic and non-electic enegy inputs. The composite of non-enegy mateial inputs, howeve, is combined using Leontief technology. The oveall aggegation of composite pimay factos, enegy inputs, and non-enegy mateials is CES with an elasticity of substitution of Figue 3 in Ruthefod et al (1997, p. 15) did not show land but the text (p. 9) mentioned land as one of the factos of poduction. 2

8 Figue 1 Stuctue of CETM ENERGY RESERVES AND NATURAL RESOURCES LABOR, LAND ENERGY CONVERSION TECHNOLOGIES ETA ELECTRIC, NON-ELECTRIC ENERGY (E, N) MACRO SECTORS ENERGY COSTS (EC) Souce: Ruthefod et al. (1997), Figue 1, p. 7. INTERMEDIATE AND FINAL DEMAND (D) INVESTMENT (I) CAPITAL (K) 3

9 Figue 2 MACRO Poduction Nest Y it σ=0.5 σ=1 σ=1 σ=0 K L ELEC NON- ELEC Y NFM PPP TRN OTH Souce: Ruthefod et al. (1997), Figue 3, p. 15. Figue 3 MACRO Consumption Nest C t σ=0.5 σ=1 σ=1 ELEC NON- ELEC Y NFM PPP TRN OTH Souce: Ruthefod et al. (1997), Figue 2, p. 14. Consumption in MACRO is descibed as CES-nested aggegate of enegy and nonenegy composite goods. Composite enegy is a Cobb-Douglas aggegate of electic and nonelectic inputs, while composite non-enegy is a Cobb-Douglas aggegate of the five industial goods. Consumes substitute composite enegy and non-enegy inputs with an elasticity of substitution of F end = 0.5, which is chosen to appoximate the own-pice elasticity of demand fo enegy. 1

10 MACRO is linked to ETA, a patial equilibium sub-model which descibes in geate details the enegy sub-secto. ETA specifies the supply functions of electic and non-electic enegy. Electic enegy is poduced by a combination of hydo-electicity, natual gas, oil, coal, and two 'backstop' technologies: advanced high cos and advanced low cost. Non-electic enegy can be poduced eithe fom oil, gas, coal, o by non-conventional technologies (such as cabonfee backstop, enewables, synthetic fuels). The list of electic and non-electic technologies in ETA ae given in Table 1. ETA includes the following intenationally taded goods (g): 1 OIL Cude oil 2 COAL Coal 3 GAS Natual gas 4 CRT Cabon emission ights ETA is fomulated as a non-linea mathematical pogam. The decision vaiables in ETA include the following: SURPLUS The non-linea pogamming maxim and defined as the sum of consume and poduce suplus EC,t Enegy cost (in egion and time peiod t) - tillion dollas EN,t Composite enegy demand E,t Electic enegy (total) N,t Non-electic enegy (total) PE e, Poduction of electic enegy (by souce e) - tkwh PN n, Poduction of non-electic enegy (by souce n) - exaj GASNON Gas consumed to meet non-electic demands OILNON Oil consumed to meet non-electic demands RSC,x,t Undiscoveed esouces (by type x) RSV,x,t Poven eseves RA,x,t Reseve additions CLEV Cabon emissions level billion tons CRLX Cabon limit elaxation billion tons EXPRT g, Expots (of goods g) IMPRT g, Impots To undestand the intenal wokings of ETA, a list of some of the impotant equations in ETA is given in Table 2. ETA solves fo the aggegate shaes of electic and non-electic enegy. The solution is aived at by MACRO fist passing on to ETA the following vaiables and thei time paths: e,t Refeence path of electic enegy demand (TKW) n,t Refeence path of non-electic enegy demand (EJ) pvcen,t Pesent value unit cost of enegy secto inputs pvpe,t Pesent value pice of electic enegy pvpn,t Pesent value pice of non-electic enegy 2

11 Table 1 List of Technologies in ETA No. Shot Name Long Name Restictions Electicity supply technologies (e): 1 HYDRO Hydo electic 2 GAS-R Existing gas-fied 3 OIL-R Existing oil-fied 4 COAL-R Existing coal-fied 5 NUC-R Existing nuclea 6 GAS-N New vintage gas-fied DLE(e) 7 COAL-N New vintage coal-fied DLE(e) 8 ADV-HC Advanced high-cost DLE(e), XLE(e) 9 ADV-LC Advanced low-cost XLE(e) Non-electicity enegy supply technologies (n): 10 OIL-LC Low cost oil eseves X(n) 11 OIL-HC High cost oil eseves X(n) 12 GAS-LC Low cost gas eseves X(n) 13 GAS-HC High cost gas eseves X(n) 14 CLDU Coal fo diect use DLN(n) 15 NE-BAK Non-electic backstop DLN(n), XLN(n) 16 RNEW Renewables XLN(n) 17 SYNF Synthetic fuels (coal shales) DLN(n), XLN(n) Note: X(n) DLE(e) DLN(n) XLE(e) XLN(n) Fossil fuels Electicity technologies subject to decline limits, Non-electic technologies subject to decline limits Electicity technologies subject to expansion limits Non-electic technologies subject to expansion limits ETA then uses the efeence time path of enegy demand to calculate othe vaiables and paametes such as the efeence pesent value of enegy demand en,t (equation (1)), the distibutive shae paamete of electic enegy evls (equation (2)) which is then used to calculate the composite enegy demand (in volume tems) EN,t (equation (4)), and the total of consumes and poduces suplus (equation (3)). Note that the total suplus is nomally calculated as the aea between the consumes (egional) enegy demand cuve and the maginal cost cuve. Howeve, it can also be calculated as the total aea unde each egion s enegy demand cuve, then subtacting the total cost of enegy supply. The demand function is assumed to have a constant own-pice elasticity of F and the function is calibated to MACRO (i.e. using the efeence pesent value of enegy demand en,t as calculated fom MACRO - see equation (3)). The total cost to poduce enegy is a linea combination of the diect costs to poduce electic and non-electic enegy, with an allowance fo oil-gas pice diffeential of OGPD = $1.25/GJ fo all egions, an allowance fo inteegional tade tanspotation costs of $2/GJ fo gas, $1/GJ fo coal, $0.33/GJ fo oil, and $10/tonne fo cabon emission ights (see equation (21)). ETA then optimizes the mix of electic and non-electic technologies by maximizing the value of the total suplus subject to all the technological and institutional constaints (as descibed in equations (7-21) of Table 2). These constaints include things like: (a) maket cleaing conditions (supply of fuels and enegy souces must at least meet the demand, total impots must equal total expots, etc.) (equations (7-9,20)), (b) side constaints which contol 3

12 the availability of diffeent technologies, though expansion limits on new technologies, decline limits on old (and new) technologies, and exhaustion limits on non-enewable esouces, etc. (equations (10-17)). In addition, equation (18) detemines the cabon emission level and equation (19) specifies the limits on cabon emission ights which ae given exogenously fo each egion and time peiod. Equation (22) defines the invese demand function fo composite enegy in ETA, which is linked to the efeence level in MACRO as explained in the next section below The Linkage of ETA to MACRO In MACRO, the demand fo composite (electic and non-electic) enegy is stuctued as a CES function. This means the demand level fo composite enegy EN j in secto j is elated to the secto output Q j, the secto unit cost C j, and the composite enegy pice PEN,j by the elation: σ C j j = kq j PEN j EN (i) whee k is some constant and F is the own-pice elasticity of demand fo composite enegy. Let EN j, C j, and PEN j be the efeence level fo these vaiables, i.e. the level as detemined in the MACRO module. The linkage of ETA to MACRO is then defined by the following equation: PEN j j = EN j PEN j σ C j C j EN (ii) which follows fom the pevious elation, and a j 1 a j E E E N N N P t j j P t ( 1 + ) + µ (1 + j ) + µ j PEN j = E N P j P j (iii) whee: E j N j t, t ae ad-valoem tax ates on electic and non-electic enegy demand in secto j. E N µ j, µ j ae distibution magins on electic and non-electic enegy (cost indices). E j N j P, P ae the efeence pices (use costs) of electic and non-electic enegy. The last equation is based on the assumption that the stuctue of the electic and non-electic enegy composition is Cobb-Douglas. If enegy cost is only a small popotion of the oveall secto cos i.e.: PEN j EN j C j PEN j ( C j = C j PEN j ) << 1, 4

13 then equation (b) can be appoximated by: o σ PEN j j = EN j PEN t EN (iv) 1 σ EN j PEN j = PEN j (v) EN t Equation (v) can be used to epesent the invese demand function fo composite enegy in ETA which will come out to be close to that modeled in MACRO. This is added to the list of equations fo ETA (shown as equation (22) in Table 2). Table 2 List of Impotant Equations in ETA, +. = (1) en t pvpe, t e, t pvpn, t n, t elvs e, t = (2) en, t, t pvpe, t σ 1 σ EN, t σ SURPLUS = en, t, t σ 1 en, t pvcen, t EC, t (3) elvs, t 1 elvs, t E, t N, t = EN, t (4), = PE, (5) E t e, t e N, t = OILNON + PN synf, + GASNON + PN new, + PN + PN cldu, ne bak, (6) GASNON = PN gas lc, + PN gas hc, + IMPRTgas, EXPRTgas, chgas, PEgas, chgas n, PEgas n, (7) GASNON 0. 5 N, t (7b) OILNON = PN ch oil lc, oil, ht + PN PE oil hc, oil, + IMPRT oil, EXPRT oil, (8) 5

14 Table 2 List of Impotant Equations in ETA PN coal, = EXPRT + ch coal, coal n, ht IMPRT PE coal, coal n, ch + PN coal, ht cldu, PE coal, + ( 1 + syntpe) PE synf, (9) PE 10 dle, ty 1, PEdle, ty, decf + (10) PE 10 dln, tp 1, PEdln, tp, decf + (11) 10 xln, nxpf + nshfn N, t+ 1 PN xln, t+ 1 (12) PN, 10 ( f ) + nshf E ( PE ) PE xle tp, g RG, t+ 1 xle, t 1, xle + xle, exp (13) RSC (14), x, t+ 1 = RSC, x, t 5 RA, x, t 5 RA, x, t+ 1 RSV, x, t+ 1 = RSV, x, t + 5 ( RA, x, t PN x, ) + 5 ( RA, x, t+ 1 PN x, t+ 1, ) (15) df RA, (16) x, RSC, x, t x, RSV, x, t, x t pv PN, (17) x, t CLEV = e cecee, PEe, + n e n cecn n, PN ( EXPRTgas, IMPRTgas, ) cecn ( EXPRToil, IMPRToil, ) cecnoil, gas, n, (18) CLEV EXPRTc IMPRTc calim = (19) ( IMPRT ) 0 EXPRT (20) q, q, = EC = ( PE e + ( cstcexp n e, ecst q e, EXPRT ) + ( PN n q, ) n, ncst n, ) + ogpd GASNON (21) 1 σ EN j PEN j = PEN j (22) EN t 6

15 2.1.3 Comments on the Stuctue of CETM The Stuctue of Poduction and Inte-fuel and Fuel-facto Substitution. The stuctue of poduction in the MACRO module of the CETM model goups labo and capital togethe, and these factos ae sepaated fom the enegy banch (see Figue 2). This means that enegy-capital and enegy-labo will have the same substitution elasticity and this implies a sevee estiction (see the discussion on the issue of capital - enegy substitutability o complementaity in section 3.2 below). On the othe hand, the intenal stuctue of the inte-fuel substitution in the MACRO module makes a useful distinction between electic and non-electic enegy inputs. Although econometic evidence is scace with espect to the substitution between electic and non-electic enegy inputs, this distinction is useful at least fom a theoetical viewpoint. This is because the choice of the electicity geneation technologies may have an impotant impact on the envionment (such as the emission of CO 2 ), and hence the focus on electic enegy consumption level may help focus attention on the choice of these technologies 2. Diffeent foms of non-electic enegy such as oil, gas, coal (diect use), synthetic fuels, enewable fuels o the non-electic backstop technologies, ae teated as pefect substitutes in the ETA module (see equation (6) in Table 2). This assumption is pehaps athe estictive especially fom the end-use s point of view. Natual gas, fo example, is known to command a pemium ove coal because of its ease of handling. It may also come into conflict with othe assumptions made in the model such as the fact that the maket shae fo natual gas is limited (see equation (7)). Limited maket shae often implies some difficulty of substitution athe than limitation in supply. Finally, if these non-electic enegy foms ae pefectly substitutable, then thei maginal costs (pices) must also be set equal to each othe. These ae stong assumptions The Small Influence of the Enegy Secto in Linking ETA to MACRO Relying on the fact that the enegy secto makes up less than 5% of the goss output of most economies, it is anticipated that any changes in the pices and quantities of enegy demand within ETA will have only a small influence on the oveall industy cost (and hence pices and demand within MACRO). This means that convegence of the esults of ETA and MACRO can be achieved faily apidly. But this is likely to depend also on the assumptions egading supply and demand elasticities. If the supply elasticity is much geate than the absolute value of the demand elasticity then convegence can be assued. Howeve, if the convese is tue, then even if enegy is only a small popotion of the oveall industy costs, it can still act as a constaint on consumption activities, and can give ise to significant fluctuations in enegy pices and demand, and theefoe, will not help fo convegence (see Figue 4). Since ETA is a pocess model athe than a conventional econometic model, the concept of supply elasticity cannot be clealy 2 Futhemoe, as Hogan (1989, p. 54) noted, the gouping of all enegy foms togethe in an aggegate enegy demand function may mask the histoically impotant tend of electification in an enegy economy (such as that obseved in the US economy duing the peiod fom 1960 to 1982). 7

16 Figue 4 ETA - MACRO Linkage ENERGY COSTS ELASTIC SUPPLY ENERGY COSTS INELASTIC SUPPLY SUPPLY (ETA) SUPPLY (ETA) DEMAND (MACRO) DEMAND (MACRO) ENERGY QUANTITY ENERGY QUANTITY defined and tested. Howeve, the geneal concept of supply esponsiveness to pice and demand changes may still be an impotant facto to conside when looking at the issue of convegence Dynamic Adjustment Constaint on Technologies could be Linked to Endogenous Factos within the MACRO economy. Equations (10-13) epesent the dynamic adjustment constaints on new and existing technologies. They define the limits to which existing technologies can be etied (because of sunk capital costs) o new technologies to be intoduced (because of the difficulty of maket penetation). These constaints eflect economic as well as institutional factos within the cuent and futue makets, and theefoe, they could also be detemined endogenously within the model athe than being set exogenously. Fo example, the ate of maket penetation fo new technologies may be dependent on the diffeences in poduction costs between existing and new technologies. The ate of etiement fo existing technology can also be specified as a function of the expected incease in futue demand and supply and the cost of capital. In othe wods, the dynamic adjustment constaints could be linked to the investment decisions within the model, athe than being specified as exogenous. Since the absence of such a linkage is lagely due to pactical consideations, this is pobably an aea fo futhe eseach. Table 3 Summay Chaacteistics of CETM Model Chaacteistics Top-down vesus bottom-up Dynamic Inte-fuel substitution Fuel-facto Substitution Capital Enegy complementaity/substitutability CETM Bottom-up in CETM, top-down in MACRO Simultaneous Yes Yes Enegy and capital ae substitutes in the MACRO poduction stuctue, but can be complements within the enegy sub-module CETM. 8

17 2.2 The MEGABARE Model and the Technology Bundle' Appoach 3 In building the MEGABARE model on top of the GTAP famewok, the authos of that model made a delibeate decision...not to adopt the nested CES (constant elasticity of substitution) poduction function appoach to enegy substitution. This was because: It was believed that it was possible to impove on the nested CES appoach in tems of both accuacy and tanspaency by intoducing what has been temed the 'technology bundle' appoach. Using this appoach, a level of detail about diffeent technologies is intoduced into MEGABARE that is nomally found only in so-called 'bottom up' models. An attempt is made to intoduce the ealism in modelling substitution options that is a featue of 'bottom up' models while etaining extensive inteactions between the enegy and othe sectos of the economy that is a featue of 'top down' models. (MEGABARE, 1996: 4) Desciption of the Technology Bundle Appoach The technology bundle appoach is descibed below in figues 5-7. Fis the intemediate inputs into poduction ae divided into technology bundle inputs typically pimay factos and pimay enegy inputs - and non-technology bundle inputs (Figue 5). The technologies fo an industy (fo example, coal-fied electicity, gas-fied electicity etc.) ae Leontief (fixed inputoutput coefficient) combinations of technology bundle inputs. The technology bundle fo an industy is a conventional smooth poduction function (such as CRESH) combination of the output of each technology. Industy output is a Leontief combination of the technology bundle and the non-technology bundle inputs The technology bundle appoach is used in the MEGABARE model to descibe the input use of the electicity geneation industy (Figue 6) and the steel industy, which epesent typical examples of enegy intensive industies. The appoach, howeve, can also be used to descibe othe enegy intensive industies. With the steel industy, the input stuctue diffes slightly fom the electicity industy: electicity and mineals ae added to the input lis along with the pimay factos and the pimay enegy inputs (Figue 7). Figue 5 Technology Bundle Appoach Goss output by industy Leontief Commodity 1... Commodity Technology bundle Souce: ABARE (1996), Figue 6, p ABARE (1996), The MEGABARE model: inteim documentation, Febuay. 9

18 Figue 6 Composition of the Technology Bundle fo the Electicity Industy Technology bundle CRESH Technology 1... Technology Leontief Leontief K L Pimay enegy 1... Pimay K L Pimay... Pimay enegy e enegy 1 enegy e Souce: ABARE (1996), Figue 9, p. 32. Figue 7 Composition of the Technology Bundle fo the Steel Industy Technology CRES EAF... BOF Leontief Leontief L K Electicity Mineals K L Pimay enegy 1... Pimay enegy e Electicity Mineals Souce: ABARE (1996), Figue 10, p. 32. EAF and BOF stand fo electic ac funace and basic oxygen funace espectively. 10

19 2.2.2 Comments on the Technology Bundle Appoach The technology bundle appoach is inteesting and innovative. It ties to intoduce the concept of substitution between altenative technologies to give a moe ealistic desciption of the natue and ange of substitution occuing within the enegy poducing and enegy-using industies, in contast to the moe taditional concept of substitution between altenative enegy and nonenegy inputs. In doing so, the appoach can claim the following advantages: 1. it ensues that the patten of input use is consistent with known technologies which usually exhibit what may be descibed as lumpy o indivisibility constaints on cetain inputs such as capital o labo, 2. it is highly tanspaent in the sense that it allows an assessment of how some policy change can lead to elative changes in the use of diffeent technologies athe than a mee obsevation of the deived changes in inputs use (ABARE, 1996: 35). 3. the elasticity of substitution paametes in the technology bundle appoach can be estimated by efeence to the esults fom 'bottom up' models and theefoe, can cove a wide ange of data values that might occu in a simulation (ABARE, 1996: 36). While in theoy, it is tue that the technology bundle appoach can povide a moe ealistic desciption of the constaints facing the enegy poducing and enegy-using industies than a conventional econometic appoach, in pactice, howeve, it is not clea how some of these potential advantages can always be implemented. In MEGABARE, fo example, inputs into the technology bundles ae still being specified as Leontief with no explicit indivisibility o lumpy constaints imposed 4. On point 3, it is not evident how the CRESH substitution paamete used in the MEGABARE model had been actually deived fom some simulation expeiment of a bottom-up natue. On a moe impotant poin the technology bundle appoach is not dissimila to the conventional appoach in econometics whee a nested poduction stuctue is used to descibe complex substitution possibilities among the inputs 5. As Powell and Rimme (1998) note: Models in which output is poduced accoding to a technology in which capital (K), labo (L) and enegy (E) ae substitutable un into the difficulty of how to allow pasimoniously fo the highe likely substitutability between K and E than between L and E. In fac the issue of substitutability o complementaity between K and E is a long-standing issue in the enegy debate (see section 3.2 below). To handle this issue, most models allow fo K and E to be sepaated fom L. In the technology-bundle appoach, although E and K ae complements within a given technology stuctue, they ae substitutes at the highe level, whee technologies ae substitutable fo each othe. Thus, given an enegy pice incease, although K cannot be used to eplace E immediately in any given technology, a less enegy-intensive but moe capitalintensive technology can be put in place, to counte the enegy pice ise, thus fulfilling the 4 The MEGABARE documentation (ABARE, 1996) does not efe to any of these indivisibility constaints but in a diffeent documentation (Hanslow et al. (1994:28)), a efeence is made to capacity constaint in the context of the discussion of the picing fomula fo a commodity which is used as input into a paticula 'technology'. Hee, it is stated that capacity constained technology eans above nomal etuns to capital which is to be epesented by a slack vaiable. 5 See fo example, Peoni and Ruthefod (1995), Powell and Rimme (1998). 11

20 Table 4 Summay Chaacteistics of MEGABARE Model Chaacteistics Top-down vesus bottom-up Dynamic Inte-fuel substitution Fuel-facto Substitution Capital Enegy complementaity/substitutability MEGABARE Bottom-up in technology bundle specification, top-down in the est of the model stuctue Recusive Indiectly though technology substitution Indiectly though technology substitution Enegy-capital ae complements within a given technology, but can be substitutable though technology substitution. function of substitutability between K and E in the longe un. In this espec the technology bundle appoach is quite innovative and flexible. 2.3 The OECD S GREEN Model 6 GREEN is a global, dynamic AGE model which highlights the elationships between depletion of fossil fuels, enegy poduction and use, and CO 2 emissions. The main focus is on the enegy secto and its linkage to the economy. Thee ae thee types of fossil fuels in the model - oil, natual gas, and coal - and one souce of non-fossil enegy - the electicity secto. Each of these can be eplaced at some futue date by "backstop" technologies. These ae assumed to become available at an identical time peiod in all egions. Thei pices ae detemined exogenously and identically acoss all egions 7. This implies an infinite elasticity of supply. Fo each of the thee fossil fuels, thee ae two altenative backstop technologies: one cabon-fee (e.g. biomass) and one cabon-based (synthetic fuel deived fom shale o coal, with highe cabon content than conventional technology). Fo electicity, the backstop technology is cabon-fee (nuclea fusion, sola o wind powe, but excluding hydo, o nuclea fission). Thee ae eight enegy-poducing sectos in GREEN: Coal mining, Cude oil, Natual gas, Refined oil, Electicity-gas-wate distibution, Cabon-based back-stop, Cabon-fee backstop, Cabon-fee electic back-stop. The thee non-enegy poducing sectos ae Agicultue, Enegy-intensive industies, and Othe industies and sevices. Thee ae fou consumption goods: Food beveages and tobacco, Fuel and powe, Tanspot and communication, and Othe goods and sevices. These ae chosen to be diffeent fom the outputs of the poduction sectos to highlight the pincipal components of final demand 6 Buniaux, J. M., Nicoletti, G., and J. Oliveia-Matins (1992), GREEN: A Global Model fo Quantifying the Costs of Policies to Cub CO 2 Emissions, OECD Economic Studies No. 19, Winte, 49-92; Lee, Hio, Joaquim Oliveia- Matins, and Dominique van de Mensbugghe (1994), The OECD GREEN Model: An Updated Oveview, OECD Development Cente Technical Pape No Thei maginal costs, howeve, ae not identical, and theefoe, thee is a etun attibuted to the fixed facto. Backstops ae not taded. Thei ole is pimaily to limit the ise in pices, and theefoe in cabon taxes. 12

21 fo enegy. Consumes ae assumed to be deciding on the optimal allocation of thei given disposable income on saving and the fou consumption goods. The demands fo these consumption goods ae then tanslated into the demands fo poduce goods (and enegy) via a tansition o make matix. Thee ae twelve egions in the GREEN model: United States, Japan, EC, Othe OECD, Cental and Easten Euope, The fome Soviet Union, Enegy-expoting LDCs, China, India, Dynamic Asian Economies (Hong Kong, Philippines, Singapoe, South Koea, Taiwan and Thailand), Bazil, Rest of the Wold (RoW). Finally, thee ae five diffeent types of pimay factos: labo, secto-specific "old" capital, "new" capital, secto-specific fixed factos (fo each fossil fuel type, and fo the cabonfee backstop), and land in agicultue Dynamics in GREEN One special featue of the GREEN model is in its dynamic teatment of the enegy-capital complementaity / substitutability issue and also in the handling of the esouce depletion issue. The dynamics in GREEN in fact come mainly fom these two issues: depletion of exhaustible esouces, and capital accumulation. In the esouce depletion sub-model, the total (poven plus unpoven) eseves ae assumed to be detemined exogenously. Howeve, the ate at which 'unpoven' eseves ae conveted into 'poven' eseves (ate of discovey o ate of convesion) is made sensitive to the pices of oil and gas. This affects the 'potential supply', which is defined by the ate at which poven eseves ae extacted 8. Potential supply povides an uppe bound on actual supply, and if actual demand falls shot of potential supply, then the diffeence between potential and actual supply is added to the futue eseves of the fossil fuels. The esouce depletion sub-model is thus ecusively dynamic (i.e. based on cuent and past pices only) athe than fowad looking (i.e. based on some expected futue pices). Capital accumulation in the GREEN model is influenced by the putty/semi-putty assumption on the natue of capital. New capital (capital invested in cuent peiod) is putty, i.e. it is highly substitutable fo othe factos (elasticity of substitution is 2). Secto-specific old capital (capital invested in pevious peiods), on the othe hand, is semi-putty and much less substitutable fo othe factos (elasticity of substitution can be as low as 0.25). Secto-specific old capital is also much less mobile between sectos (implying small and secto-specific supply elasticities). This can esult in equilibium ental values of old and new capital being significantly diffeent fom each othe, and the atio of these ental values is used in GREEN to stimulate 'disinvestment' of old capital (see Buniaux et al. (1992: 57)). Once disinvested, old capital becomes available fo use in new investment. At any point in time, the stock of capital will consist of old and new capital, and the ate of substitution between the stock of capital as a whole and othe factos will theefoe depend on the vintage stuctue of capital. Apat fom this dynamic vintage stuctue, GREEN does not include any othe explicit investment behavio by fims. The total aggegate level of investment is defined as a esidual fom the aggegate level of 8 Though the extaction ate is assumed constant ovetime, enegy pices affect the potential supply of oil and gas though the pice sensitive convesion ate (Buniaux etal. 1992, vand de Mensbugghe, 1994). 13

22 savings minus govenment secto balance and plus net capital inflows. Once the aggegate level of investment is detemined, this is then distibuted optimally to the vaious sectos in ode to equate ates of etun on new investment Inte-fuel Substitution Inte-fuel Substitution in Poduction In estimating the inte-fuel elasticities of substitution, the geneal assumption is that enegy and capital ae weakly sepaable in poduction. This means that fims choose the cost-minimizing enegy-mix given an enegy-capital bundle. But this makes sense only if thee ae dual-fied o multi-enegy technologies available, othewise, inte-fuel substitution will involve the installation of new capital and theefoe, the assumption of sepaability between enegy and capital beaks down (Buniaux et al. (1992, p. 75)). Thus, in choosing to epesent the potential fo inte-fuel substitution, the GREEN model assumes that shot un to medium un elasticities of substitution between altenative foms of enegy ae small, between 0.5 and 1.0 in the medium tem, and only 0.25 in the shot tem. Long-un 9 elasticities of inte-fuel substitution, howeve, ae set as high as 2.0. This latte value is said to be based on empiical estimates of elasticities based on samples which have multiple powe-geneating facilities (Buniaux et al., loc. cit.). These inte-fuel substitution elasticities apply only to the non-enegy poducing sectos and the electicity geneation secto. Fo the est of the enegy poducing sectos (coal mining, cude oil, natual gas, efined oil), thee is no inte-fuel substitution (see Buniaux et al. (1992, Table 3, p. 76)) The stuctue of inte-fuel substitution in poduction in the 1992 vesion of the GREEN model is as shown in Figue 9. In a subsequent vesion 10, the stuctue is alteed significantly to allow fo thee levels of nested substitution: (i) substitution between electicity and a 'nonelectic' composite fuel, (ii) substitution between coal and a 'non-coal' composite within the nonelectic banch, and finally, (iii) substitution between oil, gas, and efined fuels within the noncoal banch. All substitution elasticities ae set within the ange 0.25 < F < 2, depending on whethe it is shot-un, medium-un, o long-un Inte-fuel Substitution in Household Demand Given the enegy intensity of each consume good, household demand fo aggegate enegy is deived fom its demand fo the fou categoies of consume goods (see Figue 10). Once the demand fo aggegate enegy is known, this demand is then allocated optimally between the diffeent fuels with the same stuctue of inte-fuel substitution as in the case of poduces demand fo enegy (Figue 9). 9 This long un is defined as the peiod ove which new capital can be installed. 10 See Lee et al. (1994, Figue 1b, p. 49) 14

23 Figue 8 The Stuctue of Poduction in GREEN Goss output Leontief Labo Domestic Impoted Enegy Capital-Fixed facto Composite Leontief Region 1... Region Capital Fixed facto 15

24 Figue 9 Enegy and Backstop Technologies in GREEN Enegy CES (a) Gas (e) Cude oil (f) Coal Petoleum Poducts (g) Electicity (h) CES (b) Conventional Cabon-based backstop fuel Cabon-fee backstop fuel CES (c) Domestic Impoted CES (d). Region 1... Region (a) With elasticity of substitution (σ=0.25) fo old capital, and (σ=2) fo new capital, in all sectos except coal mining, cude oil, natual gas, and efined oil (see Buniaux et. al., 1992, Figue 1b, p. 56, and Table 3, p. 76). In Lee et al. (1994), thee is some futhe nesting (all with 0.25<σ<2): between electic and non-electic composite, then between coal and non-coal composite within the non-electic banch, and finally between oil, gas, and efined fuel in the non-coal banch (b) Elasticity of substitution between conventional and backstop technologies is (σ=10) fo agicultue, efined oil, electicity, enegy-intensive industies, and othe industies, as well as fo consume goods and govenment demand, and in the poduction of investment goods and inventoies. (c) Elasticity of substitution between domestic and impoted fuels is (σ=4) fo all fuels, except electicity (σ=0.3), and cude oil (σ= ). (d) Elasticity of substitution fo fuels fom diffeent egions (wold tade elasticities) is (σ= ) fo cude oil, (σ=5) fo coal mining and natual gas, and (σ=3.0) fo efined oil. (e) Same as fo coal. (f) Same as fo coal except with (σ= ) fo domestic-impoted and inte-egional substitutions. (g) Same as fo coal except thee ae no backstop fuels and wold tade elasticities is (σ=3). (h) Same as fo coal except thee is only one cabon-fee backstop option and wold tade elasticities is (σ=0.5). 16

25 Figue 10 The Stuctue of Household Demand in GREEN Disposable income ELES Food and Beveages Fuels & powe Tanspot & communication Othe goods and sevices Saving (a) (a) Leontief (a) Othe poduce goods Enegy (b) CES Domestic Impoted (a) Same as fo tanspot & communication. (b) See Figue 9. 17

26 2.3.3 Fuel-facto Substitution The GREEN model assumes that capital-labo and enegy-labo have the same (positive) elasticities of substitution. This assumption accods with empiical econometic evidence which suppots substantial shot-un and long-un substitutability between labo and capital on the one hand, and also between labo and enegy on the othe hand. On the issue of enegy-capital substitutability o complementaity, howeve, empiical estimates seem to be moe of a poblem. A widely held opinion in this aea is that pehaps enegy and capital ae complements in the shot-un, but substitutes in the long-un. To incopoate this featue into the model, the appoach in GREEN is to utilize a vintage capital stuctue. Thus, shot un substitution between old capital and enegy can be low, while long-un substitution between new capital and enegy can be high. The net effect will then depend on the capital vintage stuctue. Ove time, the shot-un elasticities will convege to the long-un elasticities (see Figue 5 in Buniaux et al. (1992, p. 66)). The gap between shot- and long-un elasticities and the speed of the convegence depends on the dynamics of the capital stock adjustment pocess which in tun depends on assumptions made about depeciation ate and ate of new capital fomation. The lage the net eplacement ate, the smalle the gap between shot- and long-un elasticities and the faste the convegence of the fome to the latte. In GREEN, capital is combined with a fixed facto though a Leontief stuctue befoe being combined with enegy though a CES stuctue. The ole of the fixed facto is to limit the substitution away fom/towads capital fomation in the enegy-poducing sectos so as to avoid an unealistic situation whee, fo example, following an incease in the elative pice of enegy, 'too much' investment will occu in these sectos even in the shot un. The ole of the fixed facto in pimay-enegy poducing sectos is thus to impose limits on the supply elasticities of these pimay enegies. These supply elasticities have a citical ole to play, especially in enegyenvionmental policy simulation studies. Substitution between enegy and the fixed facto-capital composite is set at zeo fo all enegy-poducing sectos, except electicity. Fo electicity and othe non enegy-poducing sectos, it is set at zeo fo 'old' capital, and at a low value of 0.8 fo new capital. Substitution between labo and capital-enegy-fixed facto composite is also set at zeo fo all enegypoducing sectos including electicity. Fo othe sectos, it is set at a low value of 0.12 fo old capital and a high value of 1.0 fo new capital (Buniaux at al. (1992, Table 3, p. 76). Accoding to Boges and Goulde (1984, p. 340), to ensue that the capital-enegy complementaity condition can be achieved, it is sufficient that the elasticity of substitution between K and E within the KE nest be given a substantially smalle (even if positive) value as compaed to the elasticity of substitution between the KE composite and labo (o othe factos) in the oute nest. To be moe pecise, we can use the following fomula established fo the case of a nested CES stuctue by Kelle (1980, p. 83): σ + KE oute = [ σ KE inne σ VA ] / S KE σ VA In this fomula, S KE is the shae of the KE-composite in the oute (value-added) nes and σ KE inne and σ KE oute stand fo the inne and oute substitution elasticities between K and E espectively. If σ KE inne is less than σ VA, then the fist tem on the ight hand side is negative. But whethe σ KE oute is negative (implying complementaity between K and E in the oute nest) depends on the size of S KE as well. If S KE is small, then this is likely even if σ VA is lage. Fo 18

27 example, using the uppe limit values of 0.8 and 1.0 fo σ KE inne and σ VA espectively as used in the GREEN model fo the case of new capital, this equies S KE < 0.2 fo σ KE oute < 0 (complementaity between K and E in the oute nest). Using the lowe limit values of 0.0 and 0.12 espectively fo σ KE inne and σ VA fo the case of old capital, this equies S KE < 1.0 fo σ KE oute < 0. The condition is always satisfied since S KE is always less than 1. Oveall, thus, old capital and enegy will always come out as complements in the value added nest of the GREEN model poduction stuctue. Fo new capital, this will also be the case if the shae of capital-enegy-fixed facto component in the value-added nest is less than 20 pecent. Note that all these discussions apply to the non enegy-poducing sectos only. Fo the enegy-poducing sectos (except electicity) thee is no fuel-facto substitution. The electicity secto is chaacteized by an inne substitution elasticity of σ = 0.8 (fo new capital only), and a KE inne zeo oute substitution elasticity of σ VA = 0 in the value-added nest. This implies new capitalfixed facto bundle and enegy ae always substitutes in the electicity secto Comments on the GREEN Model One innovative featue of the GREEN model is in the handling of the enegy-capital complementaity / substitutability issue though the use of a dynamic capital vintage stuctue. Though this stuctue, the issue of long-un substitutability vesus shot-un complementay between capital and enegy is handled quite flexibly (see the illustative numeical calculations caied out in the pevious section). This is a significant impovement ove many othe models which do not handle this issue explicitly. The specification of the capital vintage stuctue is an impotant fist step. Howeve, the next step can pehaps focus attention also on the issue of capital investment. Cuently, the aggegate level of investment in the GREEN model is specified as a esidual fom the level of aggegate saving minus govenment secto balance plus net capital inflows. Once the aggegate level of investment is detemined, the aggegate level of new investment is then distibuted optimally among the sectos. Following fom this, the atio of the new- to old-capital ates of etun is also detemined, and this will then influence the ate of old-capital disinvestment (i.e. the ate at which old capital is tansfomed back into the pool of new investment in the next peiod). All of this will affect the capital vintage stuctue. Thoughout this pocess, enegy pices play an impotant ole, in influencing the ate of etun on (old and new) capital, and hence on aggegate investment. Howeve, this influence is still indiect via the aggegate etun on capital. A moe diect ole fo enegy pices may be in influencing the capital vintage stuctue diectly, fo example, in binging about a ate of investment which will equalize the ates of etun on old and new capital ove the long un. This, howeve, implies a moe fowad looking investo than is cuently assumed fo the GREEN model. 19

28 Table 5 Summay Chaacteistics of GREEN Model Chaacteistics Top-down vesus bottom-up Dynamic Inte-fuel substitution Fuel-facto Substitution Capital Enegy complementaity/substitutability GREEN Top-down with some bottom-up details in backstop technologies specifications. Recusive Yes Yes Given the vintage stuctue of poduction, capital and enegy tend to be compliments in the shot tem and substitution ove the longe tem. 2.4 The Babike-Maskus-Ruthefod (BMR) Model Babike, Maskus, and Ruthefod (1997) utilize a model fo studying the economic impact of intenational tade and envionmental policies on the wold economy. The model includes a detailed stuctue of the inte-fuel and enegy-facto substitution possibilities fo the fim and fo the household secto (see Figues 12 and 13). The stuctue of poduction in the BMR model goups labo and capital togethe. This means that one cannot give to the enegy-capital components a diffeent elasticity of substitution as compaed to the enegy-labo o capital-labo components, and this is a sevee estiction. On the othe hand, the intenal stuctue of the inte-fuel substitution in the BMR model does contain a ich stuctue, fistly with a distinction between electicity and non-electicity inputs, and then futhe disaggegation of the non-electic inputs into vaious types of fuels using a nested-ces stuctue (see Figue 12) with 5 levels: oil and natual gas at level 0 (bottom level); coal at level 1; electicity, land, labo, and capital ae at level 2; aggegate enegy and aggegate pimay facto is at level 3; intemediate input and the combined enegy-pimay facto is at level 4; and finally output is at level 5. To calculate the elasticity of substitution between any two inputs n and m at a paticula level L in the nested-ces stuctue, we can efe to the fomula deived by Kelle (1980, p. 83): σ nm L = σ n, K S n, K σ n, l[ S n, l 1 S n, l ] l= K + 1 whee K epesents the lowest level in the nested-ces stuctue at which a component exists, associated with both the n and the m inputs (the lowest common level) and L is the highest level in the nested stuctue at which the elasticity σ nm is calculated, and the cost shae S n,l is defined by: S n, l = i n S i 20

29 i.e. the sum of all the cost shaes associated with the aggegate input n at level l, o, in othe wods, the cost shae of the input component n. Using this fomula, and consideing the poduction stuctue of Figue 12, we can conclude that: (1) enegy-capital 11 substitution elasticity σ EK (consideed at the top level, i.e. holding output constan L=5) is simply equal to 0.5/S EF whee S EF is the cost shae of aggegate enegypimay factos (land, labo, capital) in the poduction stuctue. Since this value is less than 1.0, σ EK is geate than the CES substitution elasticity at level K=4. (2) Fo inte-fuel substitution, electicity and non-electicity have an elasticity of substitution of: 1/S E 0.5*[1/S E -1/S EF ] = 0.5/S EF + 0.5/S E whee S E is the cost shae of aggegate enegy in the poduction stuctue. Since S E is athe small, the elasticity of substitution between electicity and non-electicity can theefoe be vey lage. Fo example, with S E = 0.05, S EF = 0.70, the oveall, output- constan elasticity of substitution between electicity and non-electicity is (3) The elasticity of substitution between oil and gas is given by: 1/S OG 0.5*[1/S OG -1/S COG ] 1*[1/S COG -1/S E ] 0.5 [1/S E -1/S EF ] = 0.5/S OG 0.5*/S COG + [0.5/S EF + 0.5/S E ] whee S OG o S COG is the cost shae of inputs (oil, gas) o inputs (coal, oil, gas) in the total poduction stuctue. Again, assuming that S OG = and S COG = 0.015, the oveall elasticity of substitution between oil and gas is then = This is a vey lage figue. The lage magnitude of these output-constant (uppe level) elasticities of substitution as compaed to the composite input-constant (lowe-level) elasticities of substitution can be explained as follows. When a composite input (such as aggegate enegy E) is held constan thee is only a limited oppotunity fo the vaious components (fuels) of this composite enegy to be substituted fo one anothe. When the level of output is held constan howeve, thee ae also substitutions between diffeent types of aggegate inputs (e.g. aggegate enegy E fo capital K, o composite K-E fo labo L, etc). This inceases the ange of substitution (o complementaity) between the lowe-level inputs (fuels). Refe to Figue 11, fo example, whee it is assumed fo simplicity that aggegate enegy consists of only oil and gas. When the level of aggegate enegy is held constan an incease in the pice of oil (elative to gas) will induce a substitution of gas fo oil (movement fom A to B). When the level of output is held constan aggegate enegy consumption may fall because aggegate enegy pice has inceased elative to othe factos: B may now move towads C. The total movement is now fom A to C, which shows a lage eduction in oil consumption following an oil pice incease, and theefoe, it seems as though the degee of substitutability between oil and gas is now much lage. Futhemoe, as we go up the poduction stuctue, the shae of the enegy inputs will get smalle, and since the elasticity of 11 O enegy-labo, o enegy-land: since labo, land, and capital ae gouped togethe, thei substitution elasticity with espect to enegy will be the same fo all thee pimay factos. 21

30 substitution is pice elasticity nomalized by the cost shae, it will get even lage as the cost shae gets smalle. The pupose of these uppe- o oute-level elasticity calculations is to show that the oveall level of substitution between any two input components within a paticula nest may be much lage than the magnitude of the substitution elasticities. This point is impotant to keep in mind when we compae diffeent models which may have simila elasticities, but diffeent nested stuctues. Figue 11 Substitution Elasticity when Total Output is Held Constant. Composite enegy input held constant Gas C B A Output held constant Oil 22

31 Figue 12 Stuctue of Poduction in the Babike-Maskus-Ruthefod (1997) Model OUTPUT Leontief Non-enegy intemediaies (plus cude inputs to efining) CES (σ=0.5) Cobb-Douglas Cobb-Douglas Land Labo Capital Electicity CES (σ=0.5) Cobb-Douglas Coal Oil Natual Gas 23

32 Table 6 Summay Chaacteistics of the BMR Model Model Chaacteistics Top-down vesus bottom-up Dynamic Inte-fuel substitution Fuel-facto Substitution Capital Enegy complementaity/substitutability BMR Model Top-Down Recusive Yes Yes Enegy is athe a compliment to capital (as is land and labo. 2.5 Boges and Goulde (1984) Model Boges and Goulde (1984, p. 340) assume a much simple stuctue fo the inte-fuel and fuel-facto substitution possibilities. Howeve, the model allows fo labo to be sepaated fom capital, and enegy and capital ae to be gouped togethe in one nest. This is consistent with the appoach taken in the GREEN model. To allow fo the possibility of significant complementaity between K and E, Boges and Goulde assumed a fixed-coefficient stuctue fo the KE composite. Using the Kelle fomula as descibed in the pevious section, the substitution elasticity between enegy and capital at the top level would then be given by F EK = 1*[1/S EK 1], whee S EK is the cost shae of capital and enegy inputs. Since S EK < 1, then F EK < 0, i.e. capital and enegy ae significant complements at the top level of the poduction stuctue. On Figue 13 Stuctue of Final Demand in the Babike-Maskus-Ruthefod (1997) Model Final Demand CES (σ=0.5) All non-enegy goods Cobb-Douglas Cobb-Douglas Oil Natual Gas Coal Electicity 24

33 Figue 14 Stuctue of Poduction in Boges and Goulde (1984) Model OUTPUT Cobb-Douglas Non-enegy intemediaies including impots Cobb-Douglas Labo Leontief Enegy Capital Cobb-Douglas Leontief Electicity Coal mining Cude & natual gas Petoleum efining Gas utilities the issue of inte-fuel substitution, Boges and Goulde assume a Cobb-Douglas stuctue, but ecognize that pehaps with the petoleum poduct and gas sectos, a fixed coefficient technology would be moe appopiate (see Figue 14). On the household consumption side, the utility stuctue allows fo substitution between cuent consumption and futue consumption, as well as between goods and sevices and leisue. The goods and sevices secto is Cobb-Douglas with thee diffeent types of enegy commodities: electicity, gas and gasoline and othe fuels. 25

34 Table 7 Summay Chaacteistics of the Boges and Goulde Model Model Chaacteistics Top-down vesus bottom-up Dynamic Inte-fuel substitution Fuel-facto Substitution Capital Enegy complementaity/substitutability Boges and Goulde Model Top-Down Simultaneous Yes Yes Stict complementaity between capital and enegy. 3. Towads a GTAP Model with Enegy Substitution In this section we discuss the issue of how to incopoate the impotant featues of enegy substitution as eviewed in the pevious section into the GTAP model. Cuently, in the standad GTAP model 12, thee is no inte-fuel, no fuel-facto (enegy - pimay facto) substitution, even though ecent vesion of the model allows fo a non-zeo constant elasticity of substitution between all intemediate inputs. This latte featue is an impovement ove pevious vesions. Howeve, it still does not go fa enough to allow fo an adequate teatment of the issue of enegy substitution, hence a moe substantial appoach needs to be taken hee. Thee ae two impotant issues which must be addessed when consideing extending the GTAP model to include enegy substitution in its stuctue. The fist elates to the question of a choice between a top-down vesus a bottom-up appoach. The second elates to the question about complementaity / substitutability between enegy and capital inputs ove time. 3.1 Top-Down Vesus Bottom-Up Appoach In selecting an appoach fo incopoating enegy-substitution into the GTAP model, thee ae geneally two diffeent appoaches 13. The bottom-up (engineeing) appoach often stats with a detailed teatment of the enegy-poducing pocesses o technologies, and then asks the questions: given a paticula level of demand fo enegy sevices (which may be defined in tems of the level of outputs of cetain activities, such as tavel, heating, ai conditioning, lighting, o even steel making, etc.), what is the most efficient way of going about meeting these demands in tems of the enegy technologies employed and the level of inputs. The top-down (economic) appoach, on the othe hand, stats with a detailed desciption of the maco (and intenational) economy and then deives fom thee the demand fo enegy inputs in tems of the demand fo vaious sectos outputs though highly aggegate poduction o cost functions. The advantage of a bottom-up appoach is in the detailed specification of the enegy technologies, though which newly developed o futue technologies can be incopoated into the 12 As documented in Hetel, T.W. and M.E. Tsigas "Stuctue of GTAP", Chapte 2 in Hetel (1997). 13 See, fo example, Wilson and Swishe (1993). 26

35 analysis. This povides it with much moe ealism than in the econometically-specified poduction function of the top-down appoach. On the othe hand, the latte can claim advantage in the fact that thee is histoical evidence in suppot of the assumed behavioal esponse implied in the poduction function specification, wheeas the bottom-up technology specifications may lack this behavioal content 14. To utilize the advantages of both appoaches, a top-down (maco-econometic o computable geneal equilibium) model can be linked to a bottom-up pocess model and the two models ae solved simultaneously. Howeve, thee ae many theoetical and computational difficulties associated with such a linkage. As a esul in some cases, a patial link is pusued (such as the ETA-MACRO link in the CETM model discussed in section 2) o a simulated appoach to a pocess model is used (such as the specification of the enegy-secto poduction possibilities in tems of technology bundles in the MEGABARE model, see also section 2). While thee ae cetain advantages associated with these patial appoaches, the pice to pay fo such an appoach is in the added complexity in model specification, and also the additional data and paamete equiements. Fo example, in the MEGABARE model, thee is the question of what paametes ae to be used fo the substitution between the technology bundles to ensue some consistency with obseved behavio based on histoical data. As a esult of these difficulties, and the desie to offe a widely-accessible enegy model, these appoaches ae not pusued hee. Instead, it is suggested that a simple top-down appoach be used, which can incopoate most of the impotant featues of the existing top-down models in this aea, such as the GREEN o BMR models. 3.2 The Issue of Enegy-Capital Substitutability o Complementaity Having settled on a top-down appoach to epesent enegy-substitution, the next question to conside is: which paticula stuctue should be used to epesent the substitution possibilities between altenative fuels (inte-fuel substitution) and between the enegy aggegate as a whole and othe pimay factos, such as labo and capital (fuel-facto substitution). In paticula, the question of enegy-capital complementaity o substitutability is a majo issue in this liteatue. In this section, we look at this issue fom a theoetical viewpoint and then go on to eview some of the empiical estimates of the paametes fo enegy and capital substitution /complementaity in the liteatue Impotance of the Issue Accoding to Vinals (1984), the issue of enegy-capital complementaity o substitutability may tun out to be a cucial one in detemining the diection of the adjustment of aggegate output following enegy pice changes:...the key paamete that detemines whethe output poduced goes up o down afte an enegy pice incease is the degee of complementaity/substitutability between enegy and capital, measued by σ EK [the substitution elasticity between enegy and capital] (Vinals, 1984: ). 14 As a esul thee would be some difficulties in guessing what would be the futue ates of penetation of new technologies into the maket. 27

36 In Vinals' simple one-secto model with no distotions, when the capital stock is given, and the wage level is flexible, enegy-capital substitutability is a sufficient condition fo output poduced to decline following an enegy pice incease. Altenatively, enegy-capital complementaity is a necessay condition fo output poduced to ise following an enegy pice incease. These esults point out how cucial it is fo macoeconomic analysis to detemine whethe enegy and capital ae complements o substitutes (Vinals, 1984, p 238, italics oiginal) Empiical Estimates of σ EK Despite the theoetical impotance of the σ EK paamete, empiical estimates of this paamete must ovecome many difficulties. Table 4 gives some indicative values of σ EK as estimated fom vaious empiical studies. It can be seen fom this Table that both the sign and magnitude of this paamete vaies significantly between diffeent studies. The poblem aises patly because enegy-capital substitutability is a long-tem adjustment pocess, and theefoe, empiical estimates of σ EK must take into account the issue of how shot-tem enegy usage can be dynamically adjusted to a theoetically optimal level in the long un, based on the level of investment. Convesely, capital must also adjust to the expected level of enegy pices in the long tem. Hogan (1989) has shown that whee a coect specification of a dynamic capital-enegy usage stuctue is specified, moe meaningful and accuate estimates of the inte-fuel and enegy-pimay facto substitution elasticities can be achieved. The key to the poblem of specification is that a model must be able to epesent the flexibility (in enegy usage) in the long un but also allow fo igidity o inflexibility in the shot to medium tem due to capital constaint:...esponses to pice changes take time. Although thee is ovewhelming evidence of geat flexibility in the use of enegy and othe inputs, the most impotant changes in enegy utilization depend upon changes in enegy-using equipment. If this equipment changes slowly, then the full esponse to enegy pice changes will take many yeas to unfold... Initially, the pice shocks have little effect on demand pe unit of output; often the effects ae so small as to suggest little esponse at all. But the new pices unleash foces that eventually poduce damatic changes in total enegy demand...this demand esponse can be both a substantial beak fom tend and a confusing mixtue of fuel substitutions. Analysis of this shot-un ecod, in the seach fo insights into long-un possibilities, places geat emphasis on the need fo a desciption of the dynamics of enegy demand adjustment 15. Inflexibility in capital adjustment comes fom technological factos (such as discete o lumpy investments), as well as adjustment costs. To descibe this inflexibility, one appoach is to use a technology o pocess model. Altenatively, the long-tem adjustment pocess of capital can also be specified diectly in an economic model (such as in GREEN). Howeve, it is not always easy to find empiical estimates fo the paametes of these models, hence the uncetainty suounding the extent of enegy-capital substitutability o complementaity. 15 Hogan (1989, p. 54) 28

37 Table 8 Estimates of the Patial Hicks-Allen Elasticities of Substitution (σ) and Facto Shaes (σ). US Bendt-Wood (1975) US Kulatilaka (1980) US Pindyck (1979) Euope Pyndyck (1979) Austalia Tuong (1985) FKK FLL FEE FMM FKL FKE FLE FKM FLM FEM "L "E "K "M K = Capital, L= Labo, E = Enegy, M= Mateial. Souce: Vinals (1984), Table 3, p. 242, and Tuong (1985). 3.3 The Stuctue of Inte-fuel and Fuel-facto Substitution in GTAP-E Poduction Stuctue with Enegy Substitution Based on the vaious stuctues of inte-fuel and fuel-facto substitutions adopted in othe models as descibed in section 2, the following is suggested as a good option fo GTAP-E. On the poduction side, enegy 16 must be taken out of the intemediate input nest to be incopoated into the value-added nest (see Figues 15 and 16). The incopoation of enegy into the value-added nest is in two steps. Fis following the stuctue in the CETM model as well as the Babike-Maskus-Ruthefod (1997) model, enegy commodities ae fist sepaated into electicity and non-electicity goups. Some degee of substitution is allowed within the non-electicity goup (σ NELY ) as well as between the electicity and the non-electicity goups (σ ENER ). The values of these substitution elasticities ae shown in Table 5. These ae chosen to be in the middle ange of the values adopted in othe models. 16 Pimay enegy (such as coal, gas, cude oil) can be used, not only as a souce of enegy input fo vaious industial and household activities (e.g. natual gas to povide the enegy souce fo electicity poduction, coal as enegy souce fo steel making), they can also be used as a feedstock. In this latte use, the chemical content of the enegy input (such as natual gas) is simply tansfomed to become pat of the output commodity (such as fetilize) athe than being used up as an enegy souce. Simila examples ae cude oil used as feedstock in the petoleum efiney industy, coke used as a feedstock in steel poduction, etc. 29

38 Nex the enegy composite is then combined with capital to poduce an enegy-capital composite 17, which is in tun combined with othe pimay factos in a value-added-enegy (VAE) 18 nest though a CES stuctue (See Figue 17). The substitution elasticity between capital and the enegy composite (σ KE ) is still assumed to be positive (indicating enegy and capital ae substitutes in the inne nest ). Howeve, povided the value of σ KE is set at a level lowe than σ VAE, the oveall substitution elasticity (as viewed fom the oute nest ) between capital and enegy may still be negative (Boge and Goulde (1984, p. 340)). To be moe pecise, we can use the fomula deived by Kelle (1980, p. 83) which specifies the elationship between the inne and oute elasticity of substitution between K and E as follows: KE oute = [ σ KE inne σ VAE ] / S KE σ VAE / SVAE σ + whee S KE is the cost shae of the KE-composite in the oute (value-added) nes and σ KE-inne and σ KE-oute indicate the inne and oute substitution elasticities between K and E espectively. Figue 15 Standad GTAP Poduction Stuctue Output σ = 0 Value-added σ VA All othe inputs (including enegy inputs) σ D Land Labo Capital Domestic Foeign σ M Region 1 Region 17 The eason fo a focus on the enegy-capital composite was given in section 3.2. See also the discussion in section egading the diffeences between enegy-capital and enegy-labo substitution. 18 The tem value-added-enegy is used to emphasize the fact that enegy is now pesent in this nest. 21 Fo details on the industy secto aggegation, see Table A1 of the Appendix. 30

39 Figue 16 GTAP E Poduction Stuctue Output σ = 0 Value-added-Enegy (Including enegy inputs) All othe inputs (Excluding enegy inputs but including enegy feedstock) σ VAE σ D Natual Resouce Lan Labo Capital-Enegy Composite Domestic Foeign F LAB σ M Skille Uskilled Region 1 Region Figue 17 GTAP E Capital-Enegy Composite Stuctue Capital-Enegy Composite σ KE Capital Enegy Composite σ ENER = 1.0 Non- Electic σ NELY =0.5 Electic σ D Coal σ D Non-Coal σ NCOL =1.0 Domesti c Foeign σ M Domestic Foeign σ M Gas Oil Petoleum poducts Region 1 Region σ D Region 1 Region Domestic Foeign σ M Region 1 Region 31

40 Table 9 Enegy Substitution Elasticities in GTAP-E and Othe Models SECTOR Capital-Enegy (K-E) Inte-Fuel SECTOR GTAP-E (a) GREEN (b) Ruthefod (c) Electic vs Non-electic GTAP-E (e) Coal vs othe non-electic between non-coal, non-electic GREEN Coal Cude Oil Gas Petoleum, coal poducts Electicity (d) Feous metals Chemical, ubbe, plastic poducts Othe manufactuing; tade, tanspot Agicultue, foesty, and fishey Commecial/public sevices, dwellings Fo details on the secto aggegation, see Table A1 of the Appendix. a To ensue capital and enegy ae complements in the shot un, while substitutes in the long un, the elasticity of substitution between K and E aggegate must be set lowe than the elasticity between K and othe pimay factos (F VAE ). b In the GREEN model, if the long-tem elasticity between K and E is equal to 2.0 and the shot un value set equal to 0.25, then the intemediate tem (appoximately 5 yeas) elasticity - which depends on the vintage stuctue of the capital - will be about 0.8 (see Figue 5, Buniaux et al. (1992), p. 66). c In Ruthefod models, a value of 0.5 is set fo substitution between enegy composite and land-labo-capital composite in the non enegy-poducing industies (Babike et al. (1997)), o between enegy composite and labo-capital composite (Ruthefod et al (1997), and Bohinge and Pahlke (1997)). d K-E substitution fo the electicity industy is detemined, not by an econometic paamete, but by the specification of altenative electicity-geneation technologies in the pocess model. This is based on the values of 1.0 chosen fo the substitution between electic and non-electic, and between non-coal fossil fuels in (Babike et al. (1997)), Ruthefod et al. (1997). Bohinge and Pahlke (1997) howeve, chose a value of 2.0 fo the substitution between non-coal fossil fuels. Fo substitution between coal and non-coal fossil fuels, Babike et al. (1997) chose a value of 0.5, wheeas Bohinge and Pahlke (1997) chose a value of 0.25 if the non-coal aggegate includes electicity. e 32

41 In GTAP-E, the (inne) value of σ KE is assumed to be 0.5 fo most industies 21 (including electicity), and is set equal to 0.0 fo coal, oil, gas, petoleum and coal poducts, and agicultue/foesty/fishey. This is based on the (low-to-middle) ange of the values adopted by othe models, such as the GREEN model, and the models used by Babike et al. (1997), Ruthefod et al. (1997), Bohinge and Pahlke (1997) (see Table 5). The value of σ VAE anges fom 0.2 to 1.45 and this seems to be slightly lage than the values adopted by othe models (see Table 6), but these ae the values cuently used in the standad GTAP model. Based on the values of S KE fo some typical egions in the GTAP- 4E data base 22, the oute values of σ KE ae deived using the above fomula and ae shown in Table 7. Fom this Table, it can be seen that most industies (with the exception of electicity in the USA, and electicity, feous metals, and chemical, ubbe, plastic poducts in Japan) ae chaacteized as having an oveall complementaity elationship between enegy and capital despite the fact that σ KE is still specified as non-negative within the enegy-capital nest. Table 10 Elasticities of Substitution Between Diffeent Factos of Poduction Secto GTAP-E a GREEN Ruthefod ( VAE ) L - KEF b E - KF c Coal Cude Oil Gas Petoleum, coal poducts Electicity Feous metals Chemical, ubbe, plastic poducts Othe manufactuing; tade, tanspot Agicultue, foesty, and fishey Commecial/public sevices, dwellings a In GTAP-E: between land, natual esouces, aggegate labo, and capital-enegy composite. b Between labo (L), and enegy-capital-fixed facto composite (EKF). c Between enegy (E) and capital-fixed facto composite (KF). d Between land, labo, and capital (see Babike et al. (1997)), o between labo and capital (Ruthefod et al (1997), and Bohinge and Pahlke (1997)). d 22 See Malcolm and Tuong (1999). 33

42 Table 11 The Relationship Between Inne (F KE-inne ) and Oute F KE-oute ) Elasticities of Substitution fo the Cases of Japan and the US Secto Japan USA KE-inne VAE S VAE S KE KE-oute S VAE S KE KE-oute Coal Cude Oil Gas Petoleum, coal poducts Electicity Feous metals Chemical, ubbe, plastic poducts Othe manufactuing; tade, tanspot Agicultue, foesty, and fishey Commecial/public sevices, dwellings Note: σ [ σ σ ] / S + σ / S, whee S KE, σ KE-inne ae the cost shae and substitution elasticity KE oute = KE inne VAE KE VAE VAE espectively fo the capital-enegy composite and S VAE, σ VAE ae the cost shae and substitution elasticity espectively fo the value-added-enegy composite. Finally, Tables 8 and 9 show the Amington elasticities fo the substitution between domestic and impoted good (σ D ), and between impoted goods fom diffeent egions (σ M ). The values of σ D and σ M fo GTAP-E ae taken fom the standad GTAP model, and ae seen to be lowe than some of the values used in othe models, such as those in Babike et al. (1997). In studies which seek to simulate the tade effect of a homogeneous enegy commodity maket (such as that fo coal) in esponse to an enegy-envionmental shock (such as the imposition of a cabon tax), these Amington elasticities may play a cucial ole. Howeve, this issue is not consideed in this pape. Table 12 Elasticities of Substitution Between Domestic and Foeign Souces (σ D ) Secto GTAP-E GREEN b Ruthefodc Low-High Coal Cude Oil 10.0 a Gas Petoleum, coal poducts Electicity Feous metals Chemical, ubbe, plastic poducts Othe manufactuing; tade, tanspot Agicultue, foesty, and fishey Commecial/public sevices, dwellings a This is highe than the standad value of 2.8 used in most GTAP applications. b Buniaux et al. (1992), p. 76. c Babike et al. (1997), 34

43 35

44 Table 13 Elasticities of Substitution Between Diffeent Regions (F M ) Secto GTAP-E GREEN b Ruthefod c Low-High Coal Cude Oil 20.0 a Gas Petoleum, coal poducts Electicity Feous metals Chemical, ubbe, plastic poducts Othe manufactuing; tade, tanspot Agicultue, foesty, and fishey Commecial/public sevices, dwellings a This is highe than the standad value of 5.6 used in most GTAP applications. b Buniaux et al. (1992), p. 76. c Babike et al. (1997) Consumption Stuctue On the consumption side, the existing stuctue of GTAP assumes a sepaation of pivate consumption fom govenment consumption (consumption by households of publicly povided goods) and pivate savings. Govenment consumption expenditue is then assumed to be Cobb- Douglas with espect to all commodities (F G = 1). In the GTAP-E model, enegy commodities ae sepaated fom the non-enegy commodities with a nested-ces stuctue as shown in Figue 18. If, howeve, the substitution elasticity F GEN given to the inne enegy nest and F GENNE given to the oute nest ae both equal to 1 (substitution elasticity F GNE in the non-enegy nest is assumed to be equal to F G and is theefoe also equal to 1), then the GTAP-E stuctue is equivalent to the oiginal GTAP stuctue. In geneal, howeve, if F GEN F GENNE 1, then the GTAP-E stuctue allows fo diffeent substitution elasticities within the enegy and non-enegy sub-goups, as well as between the two goups. Fo the cuent vesion of GTAP-E, the following values ae adopted: F GEN = 1, and F GENNE = 0.5. This stuctue is vey simila to the stuctue of household demand given in Ruthefod et al. (1997) (see Figue 3), and Bohinge and Pahlke (1997), except that in the model of Bohinge and Pahlke, a smalle value of 0.3 is used fo substitution between enegy and non-enegy aggegates, and a highe value of 2 is used fo substitution between fossil fuels (excluding coal). 36

45 Figue 18 GTAP-E Govenment Puchases Demand fo composite σ GENNE = 0.5 Enegy composite σ GEN = 1 Non-enegy composite σ GNE = 1 Coal σ D Domestic Foeign σ M Region... Region Household pivate consumption (i.e. consumption of pivate goods) is assumed to be stuctued accoding to the constant-diffeence of elasticities (CDE) functional fom in the existing GTAP model. If the enegy commodities within the CDE stuctue have the same income and substitution paametes, then accoding to the theoy of the CDE stuctue, these commodities can be aggegated into a single composite with the same paametes as that of the individual components. Cuently, in fac within the GTAP model, fou of the five enegy commodities (coal, oil, gas, and electicity) have simila paametes, which diffe only fom that of the petoleum and coal poducts. This implies we can aggegate the enegy commodities into a composite which emains in the CDE stuctue and has the same (o the aveage of the) CDE paamete values chaacteizing the individual enegy commodities. To allow fo flexible substitution between the individual enegy commodities, the enegy composite is now specified as a CES sub-stuctue, with a substitution elasticity of F PEN = 1 (see Figue 19) which is simila to the value given to F GEN (see Figue 18). This is the same as the value adopted in Ruthefod et al. (1997) (see Figue 3) and consistent with the medium tem value adopted in the GREEN model (see section 2.3.2). 37

46 Figue 19 GTAP-E Household Pivate Puchases Household demand fo Pivate Goods CDE Enegy composite σ PEN = 1 Non-enegy commodities σ D Coal σ D Domestic Foeign σ M Domestic Foeign Region... Region σ M Region... Region To bette chaacteize the behavio of GTAP-E in compaison with GTAP, it is woth calculating the oveall geneal equilibium elasticities in both models (see Annex 1). GE elasticities depend on the stuctue of the model, the value of the substitution paametes and the paticula closue assumed. They also depend on the benchmak database. The elasticities in Annex 1 have been calculated by using the vesion 4 of the GTAP data base. Thus the elasticities epoted in Annex 1 ae pimaily to illustate the behavioal implications of intoducing intefuel substitution. Since these elasticities ae also dependent on the base data, they ae diffeent in the cuent vesion of the model that is based on the vesion 5 of the GTAP data base. 4.Illustative Scenaio GTAP-E has been specifically designed to simulate policies in the context of Geenhouse Gas (GHG) mitigation. This is best illustated by using GTAP-E (based on GTAP Vesion 5 Data Base) to simulate the Kyoto Potocol. By signing this Potocol in 1997, a numbe of industialized counties efeed to heeafte as Annex 1 counties committed themselves to educe thei GHGs emissions elative to thei 1990 levels. Initially, the Potocol aimed at ambitious eductions: the total emissions of Annex 1 counties wee planned to be bought down in 2012 by 5 pe cent below thei 1990 levels. The Potocol made povision fo county specific tagets. A numbe of so-called flexibility mechanisms wee also povided in ode to allow emission eductions to be eallocated among Annex 1 counties. The Emission Tading (ET) 38

47 mechanism and the Joint Implementation (JI) mechanism aimed at eallocating the buden of the emission eductions among Annex 1 counties. In contas the Clean Development mechanism (CDM) would allow Annex 1 counties to fund emission eductions in non-annex 1 counties. Howeve, the initial impetus of the Potocol apidly faded away. While subsequent COP (Confeences of Paties) meetings stuggled with inticate methodological and implementation issues, emissions in most Annex 1 counties wee gowing well beyond the Kyoto tagets. As time passed, the Kyoto objectives inceasingly appeaed out of ange to some Annex 1 membes - - paticulaly the USA. In Mach 2001, the USA decided to withdaw fom the Potocol. Though the emaining Annex 1 counties eiteated thei commitment to implement the Potocol in Bonn, it is most likely that the US withdawal will make the Potocol aggegate constaint nealy nonbinding at the level of the emaining Annex 1 counties 23. The scenaios discussed in this section ae pimaily illustative. Specific limitations include, fistly that they efe to the initial vesion of the Potocol, including the US. Secondly, they only conside emissions of cabon dioxide while the Potocol coves a basket of GHGs and includes net emissions fom land use changes. Thidly, the use of the flexibility mechanisms is appoximated by assuming unesticted emission tading leading to complete equalization of the maginal of costs of abatement among paticipating counties, an outcome that is most unlikely given the eal-wold limitations associated with flexibility mechanisms. Finally, the Potocol is simulated in a static famewok that leaves aside all aspects elated to the timing of its implementation. 4.1 Altenative Implementations of the Kyoto Potocol Thee scenaios ae consideed. The fist one is the no tade case. Hee, Annex 1 counties meet thei commitments individually without elying on the use of the flexibility mechanisms. The applied emission constaints coespond to the eductions that Annex 1 counties ae foecast to achieve in 2012 i.e. the fist commitment peiod of the Potocol elative to thei coesponding emission levels in an unconstained baseline scenaio. Since this infomation equies using a dynamic model with an explicit time dimension, it is not eadily available in GTAP-E. The emission constaints used hee ae taken fom the OECD GREEN model (OECD, 1999, p. 29). In the second scenaio, unesticted emission tading among Annex 1 counties appoximates the use of ET and JI mechanisms ( Annex 1 tade case). The total emission constaint applied to Annex 1 counties in the second scenaio is the same as in the fist one, augmented by the amount of hot ai 24 fom the Fome Soviet Union. The thid scenaio assumes that cabon emissions ae taded woldwide without any estiction ( wold tade case). The constaint applied to wold emissions is the sum of the Annex 1 commitments and of the benchmak emission levels fo the non-annex1 counties. 23 This is because the emission suplus oiginating fom the economic ecession in the Fome Soviet Union often efeed to as hot ai suffices to compensate the eductions to be achieved in the emaining Annex 1 counties. 24 If Emission Tading is used, the emission suplus in the Fome Soviet Union can be, in pinciple, tansfeed to othe Annex 1 Paties at no cost. In this scenaio, the amount of hot ai in the Fome Soviet Union is assumed equal to 100 million tons of cabon o 13 pecent of the 1997 emission levels of the EEFSU egion. 39

48 Table 10 epots the emission changes elative to the benchmak levels and the coesponding maginal abatement costs of meeting the emission limitations. In the no-tade case, emission eductions ange fom 20 to almost 40 pecent. These elatively shap eductions eflect the fast gowth ates of emissions, as obseved in many Annex 1 counties since 1990, the efeence yea of the Potocol. The GREEN model makes the assumptions that these ates will emain almost unchanged duing the fist decade of the 21 st centuy. The maginal abatement costs coesponding to these eductions ange fom $126 in the US to $233 in Japan (whee these ae 1997 US dollas). These costs ae in the ange of estimates fom othe studies (see Weyant and Hill, 1999; OECD, 1999). Maginal costs ae lowe in the US than in othe Annex 1 counties despite the highe eduction ate because the US uses elatively moe coal and taxes enegy less heavily. In moe cabon-efficient counties, such as Japan, the maginal abatement costs ise faste, othe things being equal. The fist column of Table 10 shows that while emissions ae educed in Annex 1 counties that ae subject to binding constaints, they incease in the othe counties, a phenomenon that used to be efeed to as cabon leakage. The causes of cabon leakage ae multiple and involve competitiveness effects as well as the eactions of the wold enegy makets 25. In this scenaio, the leakage ate defined as the additional emissions in counties with no binding constaint elative to the emission eductions in counties with binding constaints amounts to 7 pe cent including the EEFSU egion and 4 pecen excluding EEFSU 26 Allowing unesticted tade among Annex 1 counties shifts the buden of the eduction away fom oil poducts in the elatively cabon-efficient economies (USA, EU, JPN, and RoA1) towads coal in the Fome Soviet Union. This induces a substantial eduction of the maginal abatement costs: fom aound $150 in the no-tade case to $78 in the Annex 1 tade case). These cost savings imply that the EEFSU egion sells about 300 million tons of cabon pe yea to othe Annex 1 Paties, the lagest single shae of which is puchased by the USA (see Figue 20). This epesents a tansaction woth $24 billion pe yea. The ight-hand section of Table 10 shows the esults fom a hypothetical woldwide emission tading system. In this case, the lagest eduction takes place in the CHIND egion (China and India) while the Annex 1 counties account fo less than half of the wold eduction. The wold maginal abatement cost does not exceed $30 pe ton of cabon. At this pice, aound 650 million tons of cabon ae taded each yea, with China and India accounting fo the lagest sale shae and the USA buying moe than half of these emissions (see Figue 21). 25 See Buniaux and Oliveia-Matins (2002) fo an analytical assessment of these effects. 26 Emission tading among Annex 1 counties implies that constaint of the EEFSU egion becomes effective as pat of the Annex 1 total constaint while this constaint is not binding in the no tade scenaio. As a esul Annex 1 emissions incease ex post elative to thei levels in the no tade scenaio by an amount equal to the hot ai less the leakage that would occu in the EEFSU in the not tade case. In the same way, wold emissions in the wold tade case ae highe than in the no tade case by an amount equal to the hot ai: less the total leakage geneated in the EEFSU and in the non-annex 2 egions in the no tade case. As fo the non-annex 1 egions, this might not be ealistic as most analysts ecognize that the Clean Development Mechanisms is not going to pevent cabon leakages. 40

49 350 Figue 20 : Emission tading among Annex 1 counties Million tons of cabon hot ai EEFS U USA EU JPN RoA1 positive figues ae sales; negative figues ae puchases Figue 21 : Woldwide emission tading EEX positive figues ae sales; negative figues ae puchases Million tons of cabon CHIND EEFSU USA EU JPN RoW hot ai RoA1 41

50 Table 14 Maginal Costs of Achieving the Kyoto Tagets with and Without Using the Flexibility Mechanisms Kyoto with No Use of the Maginal Costs Kyoto with Emission Tading Kyoto with Woldwide Maginal Costs % Reduction of Emissions (1997 USD pe Ton of Cabon) % Reduction of Emissions (1997 USD pe Ton of Cabon) % Reduction of Emissions (1997 USD pe Ton of Cabon) USA EU EEFSU JPN RoA EEx CHIND RoW Annex Non-Annex Leakage ate (incl. EEFSU) 7.1% na na Leakage ate (incl. EEFSU) 4.0% 3.7% na Note that the maginal costs ae expessed in eal tems (i.e. deflated by the GDP deflato of each county/egion). Theefoe, slightly diffeent maginal costs in case of emission tading ae consistent with a common tading pice of nominal tems 42

51 4.2 Macoeconomic Results Table 11 epots the macoeconomic costs of implementing the Kyoto Potocol in tems of the pecentage change in pe capita utility of the epesentative household and the associated tems-of-tade changes. If the flexibility mechanisms ae not used, the costs fo the Annex 1 Paties (measued in tems of utility of the epesentative egional household) anges fom 0.25 pe cent in the USA to 1.3 pe cent in the RoA1 egion. The highe cost in the RoA1 egion is patly explained by the degadation of the tems-of-tade elated to the fact that many counties belonging to this egion ae net enegy expotes. In contas in the net enegy-impoting, Annex 1 economies, the costs of imposing cabon estictions ae patly mitigated by tems-of-tade impovements associated with the eduction in intenational enegy pices paticulaly fo oil. The EEFSU egion loses 0.4 % of its welfae despite the fact that it has no cabon constaint to comply with; this loss is entiely explained by the fall of the enegy expots value. Inteestingly, some non-annex 1 counties/egions might even lose moe than the Annex 1 counties following the implementation of the Kyoto Potocol. This is clealy the case fo the enegy expotes (EEx). Emission tading among Annex 1 counties (see the middle columns of Table 11) educes the losses in all Annex 1 counties while geneating substantial gains (+ 2.8 pecent) in the EEFSU egion. It also contibutes to a eduction in the losses incued by the non-annex 1 enegy expotes as it shifts the buden of the eduction fom oil towads coal and theefoe implies a lowe fall of the intenational oil pice. A woldwide emission tading system would contibute to a eduction in the economic costs fo the Annex 1 counties and enegy expotes, while geneating net gains in China, India and the EEFSU egion. Figues 22 to 23 summaize the eal income changes (in tems of equivalent vaiation) implied by the thee altenative implementations of the Kyoto Potocol and povide a decomposition of the eal income vaiations into tems-of-tade and allocative 27 effects. The most noticeable outcome is that substantial cost saving can be achieved by allowing emissions to be taded. Annex 1 tading would cut the aggegate wold eal income loss by a half ($110 billion (1997 USD) to $50 billion) and a woldwide tading system would futhe educe the cost by anothe half (fom $50 billion to less than $25 billion). It must also be noted that almost evey paty has a vested inteest in some fom of emission tading (with the noticeable exception of the RoW egion) though the Fome Soviet Union has an unambiguous inteest in esticting tading to Annex 1 counties only. 27 In Figues 22 to 24, allocative effects include pue losses fom less efficient allocations of poduction and consumption as well as the eal income benefits and losses fom the sales and puchases of cabon emissions. 43

52 Table 15 Macoeconomic Impacts of Implementing the Kyoto Potocol: Pecent change in welfae (in ) and tems of tade (tot) Kyoto With No Use of the Flexibility Mechanisms Kyoto with Emission Tading Among Annex 1 Counties only Kyoto with Woldwide Emission Tading USA EU EEFSU JPN RoA EEx CHIND RoW Figue 22 : Welfae decomposition of implementing the Kyoto Potocol with no use of the flexibility mechanisms Tems of tade eff. Allocative eff. USA EU EEFSU JPN RoA1 EEx CHIND RoW A1 N-A1 Wold Figue 23 : Welfae decomposition of implementing the Kyoto Potocol with tading among Annex 1 counties Tems of tade eff. Allocative eff. USA EU EEFSU JPN RoA1 EEx CHIND RoW A1 N-A1 Wold 44

53 Figue 24 : Welfae decomposition of implementing the Kyoto Potocol with woldwide emission tading Tems of tade eff. Allocative eff. USA EU EEFSU JPN RoA1 EEx CHIND RoW A1 N-A1 Wold 5. Conclusion This technical pape has suveyed some existing CGE models which deal with the issue of enegy substitution. Impotant featues of these models ae highlighted, and whee possible, some of these impotant featues have been adapted into the existing standad GTAP model. The esult in the model, nick-named GTAP-E is then used to conduct some altenative scenaios involving implementation of the Kyoto Potocol. The main pupose of these expeiments is to highlight the suitability of the GTAP-E model in analyzing the implications of altenative stategies to educe GHG emissions. The intoduction of the enegy-envionmental dimension in GTAP is only one step towads the elaboation of a GTAP famewok that is suitable to analyze GHG issues. It is hoped that the cuent vesion of GTAP-E could be futhe extended in ode to incopoate some othe aspects, such as the complex elationship between land uses and GHG emissions. 45

54 ! closue with exogenous tade balances exogenous pop psaveslack pfactwld pofitslack incomeslack endwslack tadslack ams atm atf ats atd aosec aoeg avasec avaeg afcom afsec afeg afecom afesec afeeg aoall afall afeall au dppiv dpgov dpsave to tp tm tms tx txs qo(endw_comm,reg) RCTAX MARKCTAX dcwfd(negycom3,prod_comm,reg) dcwfd(coals,coals,reg) dcwfd(oils,oilexs,reg) dcwfd(gass,gasexs,reg) dcwfd(oil_pcs,oil_pcexs,reg) dcwfi(negycom3,prod_comm,reg) dcwfi(coals,coals,reg) dcwfi(oils,oilexs,reg) dcwfi(gass,gasexs,reg) dcwfi(oil_pcs,oil_pcexs,reg) dcwpd(negycom3,reg) dcwpi(negycom3,reg) dcwgd(negycom3,reg) dcwgi(negycom3,reg) c_ctaxbas(reg,negycom3b)! DTBAL exogenous fo all egions except one,! and cgdslack exogenous fo that one egion (which can be any one). dtbal("usa") dtbal("eu") dtbal("eefsu") dtbal("jpn") dtbal("roa1") dtbal("eex") dtbal("chind") cgdslack("row") ; Rest Endogenous ; swap gco2t("usa")=rctax("usa"); swap gco2t("eu")=rctax("eu"); swap gco2t("jpn")=rctax("jpn"); swap gco2t("roa1")=rctax("roa1"); Shock gco2t("usa") = -35.6; Shock gco2t("eu") = -22.4; Shock gco2t("jpn") = -31.8; Shock gco2t("roa1") = -35.7; 46

55 Annex 1 Geneal Equilibium Elasticities in GTAP-E and GTAP To compae GTAP-E with GTAP, the simplest and most effective way is to compae the oveall geneal-equilibium (GE) elasticities of the GTAP-E model with those of the GTAP model. The GE elasticities ae a function of the stuctue of the model, the values of the substitution paametes assumed, the benchmak database and the paticula closue assumed 28. Fo a standad GE closue whee all the pices and quantities of non-endowment commodities ae allowed to be endogenously detemined, the GE elasticities calculated fo this closue will tuly eflect the geneal equilibium chaacte of the demand elasticities 29. Fist we look at the GE own-pice elasticities. These elasticities measue the pecentage change in the output of commodity i in egion (i.e. qo(i,)) following a 1% change in its ownpice (pm(i,)) induced by an appopiate petubation in the output tax to(i,). The change in the output level can come fom two diffeent causes: (i) changes in the geneal level of activity (we can efe to this as the output (expansion o contaction) effect ), and (ii) changes due to the substitution of one input o commodity fo anothe (the substitution effect 30 ). Fo the enegy commodities, because of the additional (enegy) input-substitution stuctue intoduced into the GTAP-E model, we expect the negative substitution effect in this model to add to the negative output effect when the pice of an enegy commodity inceases. This means the magnitude of the GE own-pice elasticities fo enegy commodities in the GTAP- E model is likely to be geate than those in the GTAP model. This is in fact confimed in Table 10: the changes in the GE elasticities fo the enegy commodities ae all negative when we go fom GTAP to GTAP-E, indicating that the magnitudes of the (negative) elasticities ae all inceasing. Fo the non-enegy commodities, on the othe hand, since both the GTAP and GTAP-E models have simila stuctues fo these commodities, we will expect that thee ae insignificant changes in the GE own-pice elasticities as we move fom GTAP to GTAP-E. Fom Table 10, this is again confimed: the small vaiations in the magnitudes of these elasticities fo the nonenegy commodities aise only fom the output (expansion/contaction) effects and which ae seen to be small. Also, the vaiation can be in eithe diection. Tables 11 and 12 give the GE coss-pice elasticities fo the US and China fo illustative puposes. Fo both of these counties, we notice that all enegy commodities ae substitutes (coss-pice elasticities being positive), with the exception of the pais: COL and ELY, and OIL and P_C. These pais of enegy commodity ae complements because COL is a significant input into ELY, and similaly OIL is a significant input into P_C. As we move fom GTAP to GTAP-E, the magnitudes of the coss-pice GE elasticities fo the enegy commodities become geate, as expected. This is in contast to the case of the GE 28 As the GE elasticities ae a function of the paticula closue assumed, in this section, we pesent the GE elasticities which ae associated with the expeiment consideed in the next section. Changing this expeiment and its closue will affect the GE elasticities. 29 See Chapte 5 of Hetel (ed.) (1997). 30 Hee substitution can occu between diffeent outputs (i.e. in final demand) as well as between diffeent inputs (intemediate demand). 47

56 coss-pice elasticities fo the non-enegy commodities. In the latte case, since both GTAP and GTAP-E assume simila stuctues fo these non-enegy commodities, thei coesponding GE coss-pice elasticities as thus also simila 31. Finally, between the enegy and non-enegy commodities, we notice a significant degee of complementaity (negative coss-pice elasticities) between P_C and ELY on the one hand, and the non-enegy commodities on the othe hand. This eflects the impotance of P_C and ELY as majo enegy inputs into the poduction of these non-enegy commodities. 31 The non-enegy commodities ae also obseved to be all substitutable fo each othe despite the fact that in the intemediate input sub-stuctue, zeo substitution was assumed between these non-enegy intemediate inputs. The substitution as eflected in the GE coss-pice elasticities, howeve, eflects mainly the output (contaction/expansion) effects, which come fom a e-allocation of esouces esulting fom a change of the elative pices among these commodities. 48

57 Table A1-1 Geneal-Equilibium Own-Pice Elasticities GE Elasticities WITH Enegy Substitution fom GTAP-E Model (A): Sectos/ Commodities JPN CHN IND USA E_U FSU NEX NEM COL OIL GAS P_C ELY I_S CRP OMN AGR SER GE Elasticities WITHOUT Enegy Substitution fom GTAP Model (B): Sectos/ Commodities JPN CHN IND USA E_U FSU NEX NEM COL OIL GAS P_C ELY I_S CRP OMN AGR SER Change in Own-Pice Elasticity fom (B) to (A) Sectos/ JPN Commodities CHN IND USA E_U FSU NEX NEM COL OIL GAS P_C ELY I_S CRP OMN AGR SER

58 Table A1.2 Geneal-Equilibium Coss-Pice Elasticities fo the USA GE Coss-pice Elasticities WITH Enegy Substitution fom GTAP-E Model (A): Sectos/ Commodities COL OIL GAS P_C ELY I_S CRP OMN AGR SER COL OIL GAS P_C ELY I_S CRP OMN AGR SER Sectos/ GE Coss-pice Elasticities WITHOUT Enegy Substitution fom GTAP Model (B): Commodities COL OIL GAS P_C ELY I_S CRP OMN AGR SER COL OIL GAS P_C ELY I_S CRP OMN AGR SER Sectos/ Absolute diffeence:(a) - (B) Commodities COL OIL GAS P_C ELY I_S CRP OMN AGR SER COL OIL GAS P_C ELY I_S CRP OMN AGR SER

59 Table A1-3 Geneal-Equilibium Coss-Pice Elasticities fo China Sectos/ GE Coss-pice Elasticities WITH Enegy Substitution fom GTAP-E Model (A): Commodities COL OIL GAS P_C ELY I_S CRP OMN AGR SER COL OIL GAS P_C ELY I_S CRP OMN AGR SER Sectos/ GE coss-pice elasticities WITHOUT enegy substitution fom GTAP model (B): Commodities COL OIL GAS P_C ELY I_S CRP OMN AGR SER COL OIL GAS P_C ELY I_S CRP OMN AGR SER Sectos/ Absolute diffeence:(a) - (B) Commodities COL OIL GAS P_C ELY I_S CRP OMN AGR SER COL OIL GAS P_C ELY I_S CRP OMN AGR SER

60 Annex 2 Specifying County-specific Cabon Reductions with no Emission Tading in GTAP-E. The following box shows the closue and shocks used to simulate the no-tade case. This scenaio assumes no change of the tade account: thus the vaiable DTBAL (a linea vaiable expessed in changes) is exogenous and equal to zeo in all counties/egions except one. Accodingly, the slack vaiable cgdslack is made endogenous (while it is exogenous in the standad closue). Thus investment is calculated as a esidue in ode to guaantee no change of the tade account. The quantitative estictions applied to cabon emissions ae intoduced by making the eal cabon tax RCTAX (i.e. the nominal cabon tax deflated by the GDP deflato) endogenous and the emission gowth ates gco2t exogenous and equal to the Kyoto commitments (expessed as a pecentage eduction elative to the coesponding emission levels in 2010 in a scenaio with no constaints). Altenatively, one might impose an exogenous eal o nominal cabon tax (RCTAX o NCTAX) and leave the emission gowth ates to be detemined endogenously. An accompanying pogam calculates the Social Account Matices (SAMs). The Table A2-1 below shows the SAM of the US afte the emission constaint has been applied. The best way to intepet the income flows associated to the estiction is to assume that the estiction is imposed though a domestic maket of emission ights. The ow CAG shows the evenues that ae peceived by some kind of centalized Cabon Agency fom selling emission pemits. The total poceeds of these sales amounts to 124 billion 1997 USD, two thids of which oiginate fom sales to the electicity secto (42 billion 1997 USD) and to the othe industies and sevices (40 billion 1997 USD). Thus, in the electicity secto, puchases of emission pemits would amount up to 15 pe cent of all electicity sales. The total poceeds fom domestic pemit sales ae then efunded to the Regional Household (see the enty of 124 billion 1997 USD paid by of the RHH). 52

61 Box A.2.1 Closue and Shocks fo No Tading Scenaio! closue with exogenous tade balances exogenous pop psaveslack pfactwld pofitslack incomeslack endwslack tadslack ams atm atf ats atd aosec aoeg avasec avaeg afcom afsec afeg afecom afesec afeeg aoall afall afeall au dppiv dpgov dpsave to tp tm tms tx txs qo(endw_comm,reg) RCTAX MARKCTAX dcwfd(negycom3,prod_comm,reg) dcwfd(coals,coals,reg) dcwfd(oils,oilexs,reg) dcwfd(gass,gasexs,reg) dcwfd(oil_pcs,oil_pcexs,reg) dcwfi(negycom3,prod_comm,reg) dcwfi(coals,coals,reg) dcwfi(oils,oilexs,reg) dcwfi(gass,gasexs,reg) dcwfi(oil_pcs,oil_pcexs,reg)! dcwpd(negycom3,reg) dcwpi(negycom3,reg) dcwgd(negycom3,reg) dcwgi(negycom3,reg) c_ctaxbas(reg,negycom3b) DTBAL exogenous fo all egions except one,! and cgdslack exogenous fo that one egion (which can be any one). dtbal("usa") dtbal("eu") dtbal("eefsu") dtbal("jpn") dtbal("roa1") dtbal("eex") dtbal("chind") cgdslack("row") ; Rest Endogenous ; swap gco2t("usa")=rctax("usa"); swap gco2t("eu")=rctax("eu"); swap gco2t("jpn")=rctax("jpn"); swap gco2t("roa1")=rctax("roa1"); Shock gco2t("usa") = -35.6; Shock gco2t("eu") = -22.4; Shock gco2t("jpn") = -31.8; Shock gco2t("roa1") = -35.7; 53

62 54

63 Annex 3 Specifying Emission Tading in GTAP-E. Setting up an emission tading system equies to identify a global emission constaint fo the goup of counties/egions involved in tading and to allocate emission quotas among these counties/egions, the sum of which is equal to the global constaint. The global constaint in GTAP-E is imposed by making exogenous the vaiable gmakco2t (see the box below), while the coesponding maginal abatement cost fo the tading aea (i.e. coesponding to the common pice at which pemits ae taded) is specified as an endogenous vaiable (see MARKCTAX in the box below). The quotas allocated to each tading patnes ae specified by making the coesponding vaiables gco2q exogenous (note that these vaiables ae endogenous and automatically equal to gco2t in the no tade scenaio) and by shocking these vaiables along with a given quota allocation. It is to the use to veify that the sum of the quotas in tems of emission levels coesponds to the total constaint imposed to the exogenous vaiable gmakco2t (in the example below, the weighted sum of the quotas gowth ates specified fo the Annex 1 counties/egions must be equal to the exogenous eduction of the Annex 1 emissions by % imposed to the vaiable gmakco2t). Failue to specify a consistent quota allocation will esult into tading flows imbalances. The closue below implies that the sum of the tade account and the net cabon flows (i.e. the poceeds of emission sales and the expenditues of emissions puchases) is set exogenous and equal to zeo. In othe wods, if a county buys emission ights, it has to compensate fo it by expoting moe goods and sevices such as to satisfy to the assumption of a constant net capital flow with the est of the wold (i.e. the net investment-saving balance emains unchanged as will be illustated late on). Altenative closue ules migh of couse, be used. The Table A3-1 shows the SAM fo the US in the Annex 1 tade case. The total evenue peceived by the Cabon Agency (CAG) is lowe than in the no tade case (76 billion of 1997 USD compaed with 124 billion of 1997 USD). The explanation is twofold. Fis extending emission tading to Annex 1 counties lowes the pice of pemits (fom USD to USD pe ton of cabon). Second, assuming that the Cabon Agency plays a centalized ole in aticulating the domestic and the intenational pemit make it has now to pay fo buying pemits to the Fome Soviet Union (see the negative enty of 11 billion 1997 USD of the CAG ow to the ROW column). The Table A3-2 epots the intenational flows including those elated to pemit tading. It shows that the total amount of pemit sales by the EEFSU egion amounts to 24 billion of 1997 USD, 11 billions of it ae sales to the USA (see the ow CTRAD). Given the closue ule, the net capital flows in each county/egion (ISBAL) emains constant and equal to thei benchmak values so that any flow associated to pemit tading has to be balanced by a compensatoy change of the tade account (BALPW). Fo instance, pemit sales in the EEFSU egion make possible a deficit of the tade account by 41 billion of 1997 USD. 55

64 To summaize, specifying a pemit-tading scheme involving a sub-goup of counties/eg equies the following steps: o o o o The counties/egions that ae involved in tading ae specified in the base data (basedata.ha) by setting the coesponding values of the D_MARK coefficients (dum vaiable fo paticipation to pemit tading, heade EMTR) equal to unity. The coesponding RCTAX vaiables ae set endogenous in the closue. The county/egion specific quotas have to be specified. This is done by making coesponding gco2q vaiables exogenous in the closue and by specifying the gowth of t quotas in the SHOCK file. The aggegate emission gowth fo the tading aea (gmakco2t) is set exogenous shocked accodingly while the equilibium pemit pice fo the aea (i.e. the pice at w pemits ae exchanged: MARKCTAX) becomes endogenous (see the coesponding SW statement below). Note that all values of the D_MARK coefficients should be equal to zeo unless a pemit-ta scheme is specified. 56

65 Box A.3.1 Closue and Shocks fo Emissions Tading Among Annex 1 Counties! basic closue exogenous pop psaveslack pfactwld pofitslack incomeslack endwslack tadslack ams atm atf ats atd aosec aoeg avasec avaeg afcom afsec afeg afecom afesec afeeg aoall afall afeall au dppiv dpgov dpsave to tp tm tms tx txs qo(endw_comm,reg) RCTAX("EEx") RCTAX("CHIND") RCTAX("RoW") MARKCTAX dcwfd(negycom3,prod_comm,reg) dcwfd(coals,coals,reg) dcwfd(oils,oilexs,reg) dcwfd(gass,gasexs,reg) dcwfd(oil_pcs,oil_pcexs,reg) dcwfi(negycom3,prod_comm,reg) dcwfi(coals,coals,reg) dcwfi(oils,oilexs,reg) dcwfi(gass,gasexs,reg) dcwfi(oil_pcs,oil_pcexs,reg) dcwpd(negycom3,reg) dcwpi(negycom3,reg) dcwgd(negycom3,reg) dcwgi(negycom3,reg) c_ctaxbas(reg,negycom3b)! DTBALCTRA (incl. pemit tading) exogenous fo all egions except one,! and SAVESLACK exogenous fo that one egion (which can be any one). dtbalcta("usa") dtbalcta("eu") dtbalcta("eefsu") dtbalcta("jpn") dtbalcta("roa1") dtbalcta("eex") dtbalcta("chind") cgdslack("row") gco2q("usa") gco2q("eu") gco2q("eefsu") gco2q("jpn") gco2q("roa1") ; Rest Endogenous ; swap gmakco2t=markctax; Shock gco2q("usa") = -35.6; Shock gco2q("eu") = -22.4; Shock gco2q("jpn") = -31.8; Shock gco2q("roa1") = -35.7; Shock gco2q("eefsu") = ; Shock gmakco2t = ; 57

66 Table A3-2 Intenational Flows in the Annex 1 Tade Scenaio (billion of 1976 USD) 1 USA 2 EU 3 EEFSU 4 JPN 5 RoA1 6 EEx 7 CHIND 8 RoW Total 1 BALPW CTRAD ISBAL Total