Clmate Polces n a Fossl Fuel Producng Country: emand versus Supply Sde Polces Taran Fæhn, a Cathrne Hagem, b Lars Lndholt, c Ståle Mæland, d and Knut Enar Rosendahl e ABSTRACT In absence of jont global clmate acton, several jursdctons unlaterally restrct ther domestc demand for fossl fuels. Another polcy opton for fossl fuel producng countres, not much analysed, s to reduce own supply of fossl fuels. We explore analytcally and numercally how domestc demand and supply sde polces affect global emssons, contngent on market behavour. Next, n the case of Norway, we fnd the cost-effectve combnaton of the two types of polces. Our numercal results ndcate that gven a care for global emssons, and a desre for domestc acton, about 2/3 of the emsson reductons should come through supply sde measures. Keywords: Clmate polces, Carbon leakages, Ol extracton, Supply sde clmate polces, emand sde clmate polces https://do.org/10.5547/01956574.38.1.tfae 1. INTROUCTION In the context of a global clmate agreement, a cap on fossl fuel consumpton would have the same effects on global emssons as a cap on fossl fuel extracton, as consumpton must equal extracton at the global level. If fossl fuel markets were effcent, the global costs of reducng emssons would also be the same. In ths frst-best stuaton, demand and supply sde polces concde wth respect to effcency. However, wth lmted partcpaton n a clmate agreement, or wth unlateral acton by a sngle country or coalton of countres, demand sde versus supply sde polces matters. Many jursdctons show wllngness to reduce CO 2 -emssons by restrctng domestc demand for fossl fuels. omestc supply sde polces are less frequently dscussed, let alone pursued. The purpose of ths paper s to deduce the cost-effectve combnaton of the two types of polces, gven a target for a country s (or coalton s) contrbuton to global CO 2 abatement. The result hnges crtcally on how domestc demand sde and supply sde polces affect global emssons through nternatonal markets. We explore analytcally and numercally how the optmal domestc clmate polces depend on market behavour n the fossl fuel markets, the emssons from extracton, and the costs of downscalng domestc fossl fuel demand and supply. omestc polcy measures that reduce fossl fuel demand lead to lower nternatonal energy prces, and may also reduce the compettveness of domestc frms n the world markets for energy- a Correspondng author. Statstcs Norway, Research epartment. E-mal: tfn@ssb.no, Tel: ( + 47)95867999. b Statstcs Norway, Research epartment. E-mal: cah@ssb.no. c Statstcs Norway, Research epartment. E-mal: ll@ssb.no. d Statstcs Norway, Research epartment. E-mal: stm@ssb.no. e Norwegan Unversty of Lfe Scences, School of Economcs and Busness. E-mal: knut.enar.rosendahl@nmbu.no. The Energy Journal, Vol. 38, No. 1. 77
78 / The Energy Journal ntensve goods. Both effects cause so-called carbon leakages,.e. ncreased consumpton of and emssons from fossl fuels among free-rders; see, among others, Markusen et al. (1993; 1995), Rauscher (1997), and Böhrnger et al. (2010). Leakages occur also through supply sde polces,.e., polces that reduce fossl fuel extracton; see Erckson and Lazarus (2014). Such supply sde leakages result from ncreased supply by countres outsde a clmate coalton as nternatonal fuel prces rse. Harstad (2012) shows that supply sde leakages can be completely avoded f the coalton buys margnal foregn fossl fuel deposts and conserves them. Ths renders the non-coalton s supply curve locally nelastc. Although ths s a promsng result, buyng deposts may face several practcal problems such as asymmetrc nformaton, contract ncompleteness, and barganng falures. In our paper, we focus on the trade-off between domestc demand and supply sde measures. We, thus, reserve a gven unlateral contrbuton to global abatement to domestc acton; the optons of purchasng foregn fossl fuel deposts or nternatonal emsson quotas are excluded. Our case n the numercal analyss s Norway, whch has an ambtous target for domestc demand sde measures for 2020, but has so far not consdered usng supply sde measures. The Norwegan lack of focus on supply sde polces has been questoned by meda, analysts and NGOs at home, and has also attracted nternatonal attenton; see, e.g., The Economst (2009), Scence Nordc (2013), Forager Funds (2014) and Greenpeace (2014). Whle the country accounts for around 2 per cent of global ol producton, t contrbutes to less than 0.3 per cent of global ol consumpton (BP, 2013). The global combuston of fossl fuels extracted n Norway leads to CO 2 emssons that are about ten tmes hgher than total emssons of CO 2 wthn Norway. Even though leakages are lkely to be larger wth supply sde measures than demand sde measures, we conclude that t s cost-effectve for Norway to let most of the contrbuton to global emsson reductons be acheved through supply sde measures. In our benchmark scenaro, only one thrd of a gven global reducton should be realsed through demand sde measures; the remanng two thrds should come through supply sde measures, that s, by reducng ol extracton. 1 Prevous lterature on optmal (second-best) clmate polcy n the presence of carbon leakages through the nternatonal fuel markets has derved the optmal combnaton of producer and consumer taxes n a clmate coalton, gven a target for global emsson reductons. Hoel (1994) models an aggregate fossl fuel market, and derves analytcal expressons for optmal tax levels. Golombek et al. (1995) extend Hoel s analyss by modellng three fossl fuel markets (ol, coal and gas) and provde a numercal llustraton of optmal producer and consumer taxaton for a coalton of OEC countres, gven compettve fossl fuel markets. They fnd that the optmal producer tax of ol should be negatve, due to terms-of-trade effects domnatng the leakage effects (OEC s a net mporter of ol). Hagem (1994) compares numercally the costs of pure demand sde polcy wth pure supply sde polcy n the case of Norway, gven a target for ts contrbuton to global emsson reductons n 2000. The calculatons assume compettve fuel markets and conclude that t would be less costly to reduce ol producton than to ntroduce unform taxes on fossl fuel consumpton. 1. In practcal polcy, the domestc acton to meet a gven global target wll have to concur wth other exstng clmate ambtons and commtments. For nstance, Norway has demand sde commtments n the EU Emssons Tradng System and n the Kyoto agreements. Chapter 3 explans how these are accounted for n the computatons. Note that to the extent that these commtments are not met by the domestc actons studed here, ther fulfllment may mply extra costs. However, Norway has already shown wllngness to do more than smply complyng wth nternatonal commtments, e.g., through over-fulfllng the Kyoto oblgatons n 2008 12, fnancng technology transfer and engagng n ranforest preservaton n Reducng Emssons from eforestaton and Forest egradaton (RE + ). Ths s an ntatve supported by the UN- RE program and several other nternatonal organzatons; see http://www.un-redd.org/aboutredd.
Clmate Polces n a Fossl Fuel Producng Country /79 Our paper contrbutes to the theoretcal lterature by analysng how dfferences n emssons from fossl fuel extracton across countres affect the relatve performance of demand sde polces versus supply sde polces. Furthermore, t supplements prevous numercal analyses of demand versus supply sdes polces n several ways: Frst, we analyse the mpact of varous non-compettve ol market assumptons. Second, we take nto account emssons due to extracton of fossl fuels and, partcularly, the dfferences n emsson ntensty across countres. Thrd, we ncorporate the fact that both producton costs and emsson ntenstes are relatvely hgh n the declne phase of an ol feld here we use detaled cost nformaton from Norwegan ol felds. Fourth, prevous estmates on cost and emsson effects are outdated. In our updatng of the nformaton base we have ncluded a revew of the emprcal lterature on relevant prce elastctes n order to assess lkely carbon leakage rates on the demand as well as the supply sde. The robustness of our calculatons s checked wth thorough senstvty analyses. Assumptons regardng supply and demand elastctes, as well as the compettve envronment on the fuel markets, are decsve for our results on the optmal dstrbuton of demand versus supply sde polces. There s a large lterature on OPEC behavour (see e.g. Grffn, 1985; Alhajj and Huettner, 2000; Smth, 2005; Hansen and Lndholt, 2008). Although the conclusons from ths lterature are rather mxed, one concluson s that OPEC does not behave as a compettve producer. In our man case we model OPEC as a strategc player that seeks to maxmze ts ncome from annual ol producton, whle other producers are prce takers. To check the robustness of our results, we also consder the compettve case, along wth stuatons where OPEC has prce or producton targets. As fossl fuels are non-renewable resources, there are mportant dynamc propertes of the market that our statc analyss does not capture. A fossl fuel producer s optmzaton behavour mples fndng an extracton path that maxmzes the present value of the resource, whch depend on the expected, future prce path (Hotellng, 1931). If producers expect a gradual tghtenng of clmate polces, they may accelerate ther extracton; see Snn (2008) for a dscusson of ths green paradox. Thus, leavng out dynamc consderatons may have mplcatons for the results. On the other hand, Venables (2011) shows that although decreasng prces may speed up producton on exstng felds, ths s offset by ther postponng effect on feld openngs; see also Österle (2012) for a smlar study. Furthermore, the government can control the avalable cumulatve producton through ther producton lcencng. Hoel (2013) consders supply sde polces and argues that conservng the margnal, most costly resources reduces both total and mmedate resource extracton. These studes show the relevance of analysng fossl fuel polces n a statc framework as ours even f some ntertemporal reallocaton s gnored. We restrct our carbon leakage consderatons to those stemmng from the fossl fuel markets, dsregardng carbon leakages through the market for energy-ntensve goods. These leakages can be mtgated or completely abolshed by compensaton schemes for exposed ndustres (e.g. free allocaton of permts) or by border tax adjustments (Böhrnger et al., 2012a, and Hoel, 1996). We therefore gnore ths channel of carbon leakages. 2. THEORETICAL ANALYSIS 2.1 Unlateral Clmate Polcy We consder a fossl fuel producng and consumng home country that ams to contrbute to a certan reducton n global greenhouse gas (GHG) emssons (Ā), through a combnaton of domestc demand sde and supply sde polces. The country s aggregate benefts from domestc
80 / The Energy Journal consumpton of fossl fuels are gven by B(y o,y c,y g), where y o, y c and y g denote domestc consumpton of ol, coal, and gas, respectvely. Wthout loss of generalty, all fuels = o,c,g are measured n unts of CO 2. We assume that the beneft functon s ncreasng n each of the fuels. Furthermore, let c (x ) denote the home country s aggregate cost of producng fossl fuel, where x denotes home producton of ths fuel. We assume that the cost functons are ncreasng and strctly convex. Fossl fuels are traded n nternatonal markets at prces P o, P c and P g.to smplfy the analytcal dervatons, we treat domestc consumpton (y ) and producton (x )asexogenous varables, set by the domestc regulator. In the numercal analyss we derve the optmal consumer and producer taxes, gven proft maxmzng domestc producers and welfare maxmzng domestc consumers. The objectve for the regulator s to maxmze welfare (W), subject to the global contrbuton target, Ā, where W s utlty of consumng fossl fuels net of producton and net mport costs: MaxW = B(y,y,y ) c (x ) P (y x ) (1) o c g y,x = o,c,g = o,c,g st. 0 E E A, where E s global emssons and E 0 s the global emssons n absence of the unlateral, domestc polces. omestc polces affect the world market energy prces and, thereby, the global emssons (carbon leakages). We wll proceed by dervng the functons for the world market prces and the mpact on global emsson of unlateral polces n a partal fossl fuel market model. In the followng secton we dsregard emssons n the fossl fuel extracton processes, but return to ths n Secton 2.3. In secton 2.4 we drve the condtons for an optmal clmate polcy, gven that the regulator takes nto account the carbon leakages effect of domestc polces. 2.2 Global Emssons from emand and Supply Sde Measures Let captal letters denote foregn producton and consumpton of the three fossl fuels (X and Y, = o,c,g). As all fuels are measured n unts of CO 2, total global emssons from combuston of fossl fuels, Ẽ, must equal global fossl fuel producton, whch agan must equal global consumpton: x + X = E = y + Y. (2) = o,c,g = o,c,g = o,c,g = o,c,g We assume that foregn consumers are prce takers, where demand for each fuel s a functon of all energy prces ( Y = (P o,p c,p g), where 0 for = j and 0 for j). For Pj Pj each fuel market, foregn producton must equal foregn consumpton plus net mport from the home country: X = (P,P,P )+y x, = o,c,g. (3) o c g
Clmate Polces n a Fossl Fuel Producng Country /81 We further assume compettve behavour by foregn coal and gas producers. Ther aggregate supply functons are gven by: S X = S (P ), 0, = c,g. (4) P The ol market s charactersed by a domnant producer (OPEC) wth a compettve frnge (Non-OPEC): X = Z + S (P ), (5) o o o where Z s output from the domnant ol producer, and S o(p o) s aggregate supply from the compettve frnge. From (3) (5), we wrte the equlbrum fuel prces as functons of net mport from the home country and supply of ol from the domnant ol producer: P = P (y x Z,y x,y x ), = o,c,g. (6) o o c c g g Our default assumpton s that the domnant ol producer maxmzes net ncome. However, we also consder other objectve functons n the numercal analyss. 2 If the domnant ol producer seeks to maxmze net ncome, Z s found from: Max[P Z C(Z)], (7) o z where C(Z) s the producton cost. The frst order condton s gven by: P o+ P oz Z C (Z) = 0. (8) From (6) and (8), we can wrte all prces as functons of net mport from the home country: P = f (y x,y x,y x ), = o,c,g. (9) o o c c g g Equaton (2) can thus be wrtten as x + X = E = y + [f ( ),f ( ),f ( )]. (10) o c g = o,c,g = o,c,g = o,c,g = o,c,g As nternatonal fossl fuel prces are functons of net mport from the home country, domestc clmate polces wll affect emssons abroad. We defne the margnal demand sde CO 2 leakage rate of fuel, denoted L, as the ncrease n consumpton abroad (measured n carbon unts) followng from a unt decrease n domestc consumpton of fuel : 3 2. In Fæhn et al. (2013), Appendx A, we derve the equlbrum prce functons gven that the domnant ol producer a) operates as a compettve prce taker, b) keeps the ol prce constant, and c) keeps ts producton constant. 3. The leakage rate s postve and less than one ( 0 L 1) when the followng three condtons hold for each of the fuels: 1) Increased net demand of one of the fuels leads to hgher prces of all fossl fuels, 2) An ncrease n the prce reduces the sum of demand of all fuels, measured n carbon content, and 3) Hgher net demand ncreases total producton of fossl fuels from abroad, measured n carbon content (see Golombek et al., 1995).
82 / The Energy Journal Y j = o,c,g f k L = = jk. (11) y (y x ) j = o,c,gk= o,c,g S We defne margnal supply sde leakage rate of fuel ( L ) as the ncrease n total fossl fuel producton abroad (measured n CO 2 unts) followng from a unt decrease n domestc producton of fuel. As total consumpton must equal total producton, and y and x are exogenous, we fnd that: Xj ( ) x + y S j = o,c,g = o,c,g = o,c,g = o,c,g L = = x x = (12) f k 1+ jk =1 L. (y x ) j = o,c,gk= o,c,g Hence, we can express the margnal mpact on total emssons of domestc clmate polces as functons of the demand sde carbon leakage rate: Ẽ y =1 L, Ẽ x = L. (13) We see from (13) that demand sde polces are more (less) effectve n terms of global emsson reducton than supply sde polces when the demand sde leakage rate s smaller (bgger) than 0.5 ( E for ). We also notce that y E x 0 L 0.5 E y + E x =1. If both domestc consumpton and domestc producton decrease by one unt, there s no mpact on fossl fuel prces, and the fnal global mpact s one unt less emtted. So far we have dsregarded emssons due to extracton of fossl fuels. Fossl fuels are used as nput factors n the extracton process, and emsson ntenstes vary qute a lot across sources. Hence, the global mpact of domestc polces should be adjusted accordngly. 2.3 Includng Emssons from Fossl Fuel Extracton Let E denote total emssons (fossl fuel consumpton ncludng emssons from extracton): E = E + α (x )+ β (X ), (14) = o,c,g = o,c,g where α (x ) and β (X ) are emssons as functons of extracton of fossl fuel n the home country and abroad, respectvely. We fnd (see Appendx A): j j β (X )) j = o,c,g E y = E y + =1 L + B, (15) y j j β (X ) j = o,c,g E x = E x + α x + =L + α x B, x
Clmate Polces n a Fossl Fuel Producng Country /83 where B expresses the ncrease n emssons from extractons followng from ncreased fossl fuel producton abroad due to a unt ncrease n domestc consumpton, and s gven by: B β X(1 l ) β X l h, h = o,c,g = o,c,g. (16) h h where l j s the demand sde leakage rate for fuel j (ncreased consumpton of fuel j abroad due to reduced consumpton of fuel at home): f l = k j jk, j = o,c,g = o,c,g. (17) (y x ) k = o,c,g We cannot n general sgn B, as t depends on the sgns of lj, and the magntudes of β Xj. Under perfect competton, an ncrease n any y x wll ncrease all prces. 4 Wth ncreas- ng supply functons, ths means that all X wll ncrease. In ths paper, the ol market s not perfectly compettve. However, also under such market structures t s lkely that the supply of all fossl fuels from abroad ncreases when domestc demand for one of the fuels ncreases. If ths s the case, we must have that: (1 l ) 1, and lh 0, h, and B s postve and ncreasng n all of the β Xj, j = o,c,g. Hence, we see from (15) that demand sde polces become more effectve n reducng global emssons the larger emsson ntenstes abroad, whereas supply sde polces become more effectve the smaller β X and the larger domestc emsson ntensty ( α x ). A cut n j domestc demand drves down the fossl fuel prces and decreases producton abroad, and thereby also emssons from extracton. The opposte occurs under supply sde polces. Reduced domestc producton leads to ncreased emssons from extracton abroad, but less emsson from domestc extracton. From (15) we also see that E y + E x =1+α x. If both domestc consumpton and producton decrease by one unt, there are no carbon leakages, but as domestc fuel producton causes emssons from extracton, global emssons decrease by more than one unt. 2.4 Optmal Unlateral Clmate Polcy The objectve for the regulator s to maxmze welfare (W), subject to the global contrbuton target and the prce functons. That s, the regulator solves (1), where s gven by (9) and E = y + [f ( ),f ( ),f ( )] + α (x )+ β (X ). (18) o c g = o,c,g = o,c,g = o,c,g = o,c,g P From the frst-order condtons for ths maxmzaton problem, we fnd that: f k y k k y k = o,c,g (y x ) B = P + (y x )+ke (19) 4. See Golombek et al, 1995, Appendx A.
84 / The Energy Journal f k k k x (y x ) k = o,c,g c (x )=P (y x )+ke (20) k s the shadow cost of the emsson constrant. ( E y and E x are the margnal effects on global emssons of ncreased consumpton and producton of fuel n the home country, respectvely, see (15)). f k (y k x k ) s the terms-of-trade effect. If the country s a net exporter k = o,c,g (y x ) of a fuel, a hgher prce mproves terms of trade. Hence, the terms-of-trade effects for a fuel exporter wll tend to favour supply sde polces,.e., to reduce producton rather than consumpton. Note that ths effect occurs also n the absence of clmate polcy. In the followng we wll dsregard terms-of-trade effects, as the prce changes and consequently the welfare mpacts of ths can be consdered mnor for the small home country relatve to the other terms n (19) (20). Small prce changes do not mply, however, that global emsson effects of these prce changes can be gnored consumpton effects abroad may well be of the same order of magntude as consumpton effects n the home country; as analysed n the prevous sectons. From (19) (20), we then fnd: B y P P c (x ) = =k. (21) E E y x Hence, optmal clmate polcy mples that the margnal cost of global emsson reductons B y P through domestc demand sde polcy should equal the margnal cost of global emsson E y P c (x ) reductons through domestc supply sde polcy across all fuels. Gven that domestc E x consumers and producers are prce takers and maxmze ther net beneft and proft, t s shown n Golombek et al. (1995) that the optmal outcome can be acheved by ntroducng fuel-specfc c p consumer taxes, t = ke, and producer taxes, t = ke. 3. NUMERICAL ANALYSIS y x We now turn to the comparson of demand and supply sde polces n the case of Norway. Our focus s on the year 2020, as the Norwegan government has specfc clmate goals for that year (see below). In Secton 3.1 we estmate margnal costs of Norwegan unlateral reductons n fossl fuel demand and supply. Ths means quantfyng B y P and P c (x ), respectvely; see Eq. (21). emand sde abatement s assessed by means of a computable general equlbrum (CGE) model for Norway, whch s smulated for the year 2020 (Fæhn and Isaksen, 2015). Supply sde measures are quantfed by dentfyng representatve, margnal cuts n Norwegan ol producton based on hstorc data for the perod 2009 2011. We then dscuss the relevance of these data for the year 2020. Norway s also a sgnfcant producer of gas, accountng for around 3 per cent of global gas producton (BP, 2013). Gas s, however, a fossl fuel wth relatvely low emssons and wth larger substtutablty aganst the hgh-emttng coal. Hence, t s not clear whether reduced Norwegan gas extracton would decrease or ncrease global emssons, and we do not consder ths supply sde opton n our analyss. 5 5. We abstract from the techncal challenges of separatng ol and gas extracton, but return to ths ssue n Secton 3.3.2.
Clmate Polces n a Fossl Fuel Producng Country /85 In Secton 3.2 we analyse the effects on global emssons by explotng a partal model of the global fossl fuel market effects, where we also take nto account emssons from extracton of fossl fuels. The model s calbrated based on hstorc data (2011), and agan the relevance for the year 2020 s brefly dscussed. These computatons wll provde the values of the denomnators n Eq. (21), E y and E x. In Secton 3.3 we combne the fndngs n the two precedng sectons to derve the optmal combnaton of demand and supply sde polces for Norway as expressed n Eq. (21). 3.1 Unlateral Clmate Polcy 3.1.1 emand sde polces The Norwegan parlament has announced hgh ambtons for ts contrbuton to global (demand sde) emsson reductons, correspondng to a 30 per cent reducton from Norwegan 1990 emssons by 2020. Moreover, t has emphassed that the lon s share of the reductons s to result from domestc acton. To obtan a margnal cost functon for demand sde measures n Norway, we use Statstcs Norway s technology-rch CGE model for the Norwegan economy, MSG-TECH (see Fæhn and Isaksen, 2015). We smulate costs of unform emssons prcng gven dfferent demand sde abatement levels. The effects are measured from a reference scenaro that ncorporates clmate polces already mplemented, approved, or promsed for the years up to 2020. From 2008, ths ncludes the partcpaton n the EU ETS. 6 Snce we assume that the demand sde abatement ams to contrbute to global emsson reductons, we only consder emssons prcng n sectors outsde the EU ETS. Wth the cap on total emssons n the EU ETS, addtonal cuts n Norwegan ETS sectors wll merely dsplace emssons to ETS-regulated nstallatons n other European countres. Based on a number of smulatons, we fnd a margnal cost curve for Norwegan demand sde measures as expressed by Eq. (22) and depcted n Fgure 1. (See the numerator of the frst fracton of Eq. (21)). For all the smulated emsson targets, all abatement takes place as reduced ol consumpton, mostly wthn the transport sector. These measures are a mxture of nvestments n new vehcle technologes, substtutng publc for prvate transport and reducng transportaton demand. 2 B y P o= 2.5A + 86.6A + 23.4. (22) o A denotes the level of domestc emsson reductons (measured n mllon tonnes of CO 2 ). As Fgure 1 suggests, the margnal abatement cost s ncreasng and farly lnear. 7 The nterpretaton s that you wll have to resort to yet more expensve cuts as you add to the abatement ambtons, but the cost ncrease between the last CO 2 unt abated and the next cost-effcent margnal measure s farly constant. The abatement cost at the margn s a result of a vast number of smultaneous responses n the large-scale CGE model that depend on a varety of elastctes and other parameters. 3.1.2 Supply sde polces The costs of supply sde measures n our statc framework are the forgone profts by not extractng the ol, correspondng to P c (x ); see the numerator of the second fracton of Eq. (21). o o o 6. The same smulated scenaro s used n Clmate Cure 2020 (2010), the report of an offcally apponted commsson tasked wth preparng the ground for evaluatng Norway s clmate polcy. 7. Ths s consstent wth the assumpton made by e.g., Tol (2014).
86 / The Energy Journal Fgure 1: Margnal Costs of Foregone Fossl Fuel Consumpton n Norway We sngle out ol felds whch can be charactersed as margnal, n the sense that termnatng extracton nvolves small proft losses per unt. Ol felds n the declne phase generally have hgher costs than felds n the plateau phase. Explanatons are that margnal operatng costs, ncludng energy nput, are ncreasng as remanng ol n the reservor declnes. In addton, IOR (Improved Ol Recovery) actvtes to prolong the lfetme of maturng felds can nvolve new costly nvestments, mplyng that the proft losses of not undertakng an IOR project may be modest (though not always). Typcally, these felds also have hgher emsson ntensty. Unfortunately, we have lmted nformaton about IOR costs (see below). For the years 2009 2011 we have sngled out nne Norwegan felds where ol consttuted a major part of total petroleum producton (several were pure ol felds). In addton, these felds were n, or close to, the declne phase. We have feld data from Statstcs Norway on producton volumes and varable costs, costs that would not accrue f ol producton were termnated. Based on these data we have constructed a margnal producton cost curve; see Appendx B. To calculate margnal forgone profts by reduced ol producton, we apply the average ol prce over the perod 2009 2011 (US 84.5 per barrel of Brent Blend), and subtract the margnal producton costs. The results can be consdered as the margnal costs of forgone ol extracton n Norway, and are shown n Fgure 2. The resultng curve of margnal forgone profts s concave, snce the margnal producton cost curve has the usual convex shape. Note that we dsregard the fact that reduced supply of Norwegan ol would ncrease the ol prce margnally, slghtly ncreasng the revenue of the remanng Norwegan ol export; see the senstvty analyss. The supply sde margnal cost curve, where A S s reduced extracton measured n mllon tonnes of CO 2, s: 2 P o c (x o o)= 0.7A S+ 19.6A S 6.1. (23) We see that t s actually proftable to reduce a small amount (0.3 Mt of CO 2 ) rrespectve of clmate benefts, due to hgh producton costs of some of the smaller felds. 8 8. Ths s not explaned by the averagng of the ol prce; even f we use the actual yearly ol prces some of the felds come out wth negatve proft n some of the years. They stll produced all the years.
Clmate Polces n a Fossl Fuel Producng Country /87 Fgure 2: Margnal Costs of Foregone Ol Extracton n Norway In our study we are nterested n abatement optons n the near future such as 2020. Thus, the relevant queston s to what extent the margnal cost functon depcted n Fgure 2 s representatve for comng years. Several of the felds we have studed for the years 2009 2011 wll stop producng before 2020. On the other hand, some felds that are now n ther plateau phase wll be n ther declne phase around 2020, suggestng that ther costs per unt producton wll ncrease. It s dffcult to know whether the net effect of these consderatons wll push the cost curve n Fgure 2 up or down. However, there are several reasons why we may have underestmated the total costs of producton,.e., overestmated the costs of reducng producton. Frst, we do not have specfc nformaton about the costs of IOR projects, whch are often projects wth lmted profts per unt of extracton. Second, we have only consdered advanced termnaton of maturng felds. A costeffectve downscalng of ol producton may also mply that some felds wth lmted proftablty are not developed at all. To help assessng the relevance of Fgure 2 for the year 2020, we have also gathered nformaton on an ol feld named Ivar Aasen that s decded to be developed. Here we have access to nformaton about both expected annual development and operatng costs, as well as producton (The Norwegan Ol Company, 2012). Based on the reported data we calculate a break-even ol prce of US 60 per barrel for ths feld, usng a dscount rate of 6 per cent whch the ol company uses. The forgone profts of not developng ths feld are comparable to the data behnd Fgure 2. 9 In addton, Rystad (2013) ponts to several Norwegan (undeveloped) ol felds wth break-even 9. An ol prce of US 84.5 per barrel, and a break-even prce of US 60 per barrel, mples a cost of US 24.5 per barrel forgone ol producton, correspondng to US 58 per tonne CO 2. An average producton of 1.4 mllon Sm 3 over the perod 2016 2028 leads to 3.7 mllon tonnes of CO 2 per year when t s combusted.
88 / The Energy Journal prces above US 72 per barrel. These observatons support our belef that the costs of reducng ol producton are lower than what we assume n our analyss. The ol prce around 2020 may be dfferent from what t was n 2009 2011. The steep declne n the ol prce lately llustrates ths qute clearly. If the ol prce n 2020 becomes lower than expected, then forgone profts of reduced ol extracton wll also be lower f extracton costs are unchanged (and vce versa f the ol prce becomes hgher than expected). However, extracton costs have tended to be postvely correlated wth the ol prce (see e.g., Osmundsen et al. (2015), who fnd that rg rates n the Gulf of Mexco ncrease sgnfcantly wth ol and gas prces), meanng that the effect of a dfferent ol prce on forgone profts could be moderated. In addton, a lower (hgher) ol prce could ental lower (hgher) ol producton n the reference case (.e., n a stuaton wthout supply sde polces), at least f the lower (hgher) ol prce s expected. We are nterested n the margnal costs of reducng ol producton compared to the reference case. Hence, t s not certan that a lower (hgher) ol prce wll lead to lower (hgher) costs of supply sde measures f the ol prce change s expected by the Norwegan ol producers. To sum up, although the uncertantes are rather large, t seems more lkely that the margnal costs of supply sde measures around 2020 le below than above the curve shown n Fgure 2. 3.2 Numercal Analyss of Global Fossl Fuel Markets 3.2.1 The partal fossl fuel market model Based on the exposton n Secton 2, we construct a smple numercal model that makes t easy to dentfy and adjust the basc assumptons drvng the results (the model equatons are specfed n Appendx C). The man drvers are ) prce responsveness on the demand sde (ncludng substtuton effects between ol and other fossl fuels), ) prce responsveness of Non- OPEC supply, ) OPEC s response, and v) dfferences n emsson ntensty n ol extracton. We consder so-elastc demand functons (.e., wth constant drect and cross prce elastctes), soelastc supply functons for compettve fossl fuel producers, and constant unt producton costs for OPEC (when behavng as a domnant producer). As we are focusng on a permanent cut n ol supply as a potental supply sde measure, we are mostly nterested n the long-run effects n the market,.e., we consder long-run elastctes. Fnally, we model fxed emsson ntenstes n ol extracton, but these should be nterpreted as emsson ntenstes of margnal producton. Appendx contans a detaled dscusson of the man drvers, n partcular a revew of exstng demand and supply elastcty estmates from the lterature. Here we only present the assumptons of our benchmark case, whch are motvated n Appendx. Ol prce ncreases may reduce ol consumpton n varous ways. Ol consumers may reduce ther total energy use, or they may swtch to other energy goods such as coal, gas, or renewables. Swtchng to other energy goods requres that there are vable alternatves, whch wll vary across sectors. Reducng total energy use may ether nvolve reduced use of energy servces (e.g., drvng fewer mles, producng/consumng less energy-ntensve products), or usng more energy-effcent vehcles (or transport modes), captal, or equpment. In the long run, hgher prces may also stmulate the development of more ol-effcent technologes. In prncple, long-run prce elastctes should capture all these effects. Based on the lterature revew, we apply a drect prce elastcty of 0.5 n the long run, and cross-prce elastctes for coal and gas of 0.08. However, we report the effects of other estmates as well. Hgher prces of ol ncrease the proftablty of ol exploraton, new felds developments, and IOR projects. An ncrease n the prce of ol wll mostly affect extracton of so-called margnal
Clmate Polces n a Fossl Fuel Producng Country /89 resources, such as exploraton and feld development n ultra-deep waters, developments of smaller felds and unconventonal ol, and IOR projects. Hgher ol prces may also lead to mproved technologes n the long run, smlarly to ol-effcency mprovements on the demand sde. Based on the lterature revew, we use a supply elastcty of 0.5 for Non-OPEC. Ths mples that ol demand and Non-OPEC supply are equally prce elastc. However, due to substtuton between ol and other fossl fuels, the fossl fuel demand elastcty (wth respect to the ol prce, and measured n CO 2 unts) becomes around 0.4. As dscussed n Secton 2, our default assumpton n our benchmark case s that OPEC behaves as a domnant producer. The unt producton cost of OPEC s calbrated so that our reference smulaton s consstent wth the market outcome of the year 2011. 10 In our benchmark case, the unt margnal producton cost of OPEC turns out to be 45 per cent of the ol prce n 2011, whch s wthn the range of producton costs reported by IHS CERA for OPEC countres (see e.g. Fgure 3.9 n Mnstry of Petroleum and Energy, 2011). 11 When we model OPEC as a compettve producer, we assume the same supply elastcty as for Non-OPEC. Although the lon s share of CO 2 -emssons from ol use takes place as the ol s combusted, emssons from ol extracton have to be counted, as well. Accordng to OGP (2012), the average GHG emssons per unt producton worldwde n 2011 were 159 tonnes CO 2 e (CO 2 equvalents) per 1,000 toe hydrocarbon produced. The fgure for the Mddle East s only 51 tonnes CO 2 e, but the coverage s less comprehensve for ths regon hence the real average could potentally be hgher. The European fgure s 84 tonnes CO 2 e. OGP (2012) does not report fgures for Norway, but based on data from Statstcs Norway we calculate the average Norwegan emsson ntensty n 2011 to be 60 tonnes CO 2 e per 1,000 toe. For the rest of Non-OPEC we make a rough calculaton based on the OGP (2012) fgures for the Mddle East, Europe and the world, arrvng at around 200 tonnes CO 2 e. 12 The average fgures reported above wll typcally devate from the margnal change n emssons of ncreased or reduced ol producton. Reduced ol producton n Norway could e.g., nvolve reduced IOR actvty or advanced termnaton of a feld. In both these cases, energy use per unt extracton wll tend to be hgher than average; see Fæhn et al. (2013), Appendx C. The same could be true for reduced ol exploraton or feld developments, at least n aggregate, as the margnal areas or felds wll tend to be less proftable, whch often means that more costly energy s needed per unt producton. Smlarly, ncreased supply from other Non-OPEC producers could mply hgher-thanaverage emsson ntenstes. For nstance, Canadan ol sands are consdered relatvely costly and thus margnal resources, wth average emsson ntenstes around three tmes the world average. When t comes to OPEC supply, however, ncreased producton may come from ncreased extracton of developed felds n countres lke Saud Araba and, thus to a lesser extent nvolve hgher emsson ntenstes. 10. The market n 2020 wll lkely devate somewhat from the market outcome n 2011. However, ths devaton has margnal mpacts on our results. 11. Note that the calbrated unt cost for OPEC s ncreasng n the absolute value of the resdual demand elastcty. When the resdual demand s more elastc, OPEC s less nterested n cuttng supply to ncrease the ol prce, and hence unt costs must be hgher to obtan a reference case consstent wth base year data. 12. OGP (2012) reports both emssons and producton data for seven regons of the world. We deduct emssons and producton from the Mddle East and half of those from Europe (.e., Norway), and calculate the emsson ntensty for the remanng regons, whch we then assume s representatve for Non-OPEC.
90 / The Energy Journal Our benchmark case assumpton s that margnal emsson ntenstes are 50 per cent above the reported average fgures above. 13 For Norway and (other) Non-OPEC countres ths s related to the margnal supply most lkely beng more emsson-ntensve than average supply. For OPEC, the ncrease s partly related to less comprehensve reportng and relance on Mddle East fgures (see above) and partly to margnal supply possbly beng more emsson-ntensve than average supply. Thus, we set the emsson ntenstes n Norway, OPEC and Non-OPEC equal to respectvely 90, 76 and 300 tonnes CO 2 e per 1,000 toe. 14 For comparson, emssons from consumng (.e., combustng) 1,000 toe of ol s about 3,070 tonnes of CO 2. Although of mnor mportance here, we also account for emssons from extractng other fossl fuels, and set emsson ntenstes for coal and gas equal to the Non-OPEC emsson ntensty reported above (.e., 300 tonnes of CO 2 e per 1,000 toe). 3.2.2 Effects on global emssons of demand and supply sde polces We frst report the smulaton results of exogenously reducng Norwegan ol extracton or consumpton by one unt of CO 2. We are nterested n the net effects on global emssons,.e., the denomnators E x and E y n Eq. (21). As shown n Secton 2, the sum of E x and E y should equal one plus α x,.e., the emssons from domestc extracton (relatve to emssons from consumpton). Table 1 dsplays the net global emsson reductons when OPEC acts as ether a domnant or a compettve producer. The table also shows the varous components of the emsson reductons. Note that the leakage rate L o defned n Secton 2 s equal to mnus the sum of Ol market leakage and Coal/gas market leakage under emand sde polcy (and also equal to the sum of the three frst components under Supply sde polcy). We frst notce that leakage through the ol market s around 50 per cent for both demand sde and supply sde leakage. Ths s certanly the case f OPEC acts compettvely, and follows straghtforwardly from the assumpton of equal (absolute values of) supply and demand elastctes. If OPEC acts as a domnant producer, t s optmal for the producer group to adjust ts supply slghtly more to changes n Norwegan supply or demand compared to n the compettve case, but the dfference s not bg: Supply sde leakage through the ol market s 55 per cent, compared to 45 per cent for demand sde leakage. Next, we see from Table 1 that overall market leakage s substantally lower under demand sde polcy than under supply sde polcy, whether OPEC behaves compettvely or as a domnant producer. Ths s due to substtuton between ol and other fossl fuels, whch obvously goes n dfferent drecton dependng on whether the ol prce drops (demand sde) or ncreases (supply sde). When ol demand abroad ncreases (decreases) due to reduced Norwegan ol consumpton (extracton), coal and gas consumpton s somewhat reduced (ncreased). Ths effect alone accounts for almost 10 per cent of the gross emsson reducton. Fnally, the mportance of emssons from fossl fuel extracton s modest, accountng for respectvely 3.7 per cent and 6.1 per cent of the net global emsson effect of reduced Norwegan 13. The margnal emsson ntenstes could n fact be even hgher at least some felds have even hgher emsson ntenstes. However, t s dffcult to know f the most emsson-ntensve felds are the margnal felds, both n Norway and elsewhere. Anyway, as the results n Table 1 ndcate, assumng even hgher emsson ntenstes n producton would not alter our results substantally, as emssons related to producton are much lower than emssons from combuston. 14. It could be argued that the emsson ntensty of Norwegan ol extracton should be set to zero, as these emssons are regulated by the EU ETS, whch has a cap on overall emssons (cf. the dscusson of ETS sectors n Secton 3.1.1). As seen n the followng secton, however, these emssons are of less mportance.
Clmate Polces n a Fossl Fuel Producng Country /91 Table 1: Net Global Emsson Reducton from Reduced Norwegan Ol Extracton or Consumpton by One Unt of CO 2. Benchmark Case extracton and consumpton of ol. The effects are hghest for demand sde polcy, as under supply sde polcy ncreased emssons from ol extracton outsde Norway are modfed by reduced emssons from ol extracton n Norway. Obvously, net emsson reductons are senstve to a number of assumptons such as prce elastctes and OPEC behavour. Hence, n Secton 3.3.2 below we present a detaled senstvty analyss. We now use the fndngs n Table 1 to analyse the optmal balancng of demand and supply sde polces, focusng on the case wth OPEC as a domnant producer. 3.3 Optmal Balancng of emand and Supply Sde Polces 3.3.1 The cost-effectve soluton By combnng the demand sde and supply sde cost curves expressed n eqs. (22) and (23) wth the quantfcatons of ther net effects on global emssons, E x and E y, derved above, we can fnd the optmal composton of domestc acton; see Eq. (21). Eq. (21) expresses that the margnal cost of global emsson reducton n optmum s equal for demand sde and supply sde acton. In our benchmark case wth OPEC as a domnant producer the net global emsson reducton of demand sde polces, A, s estmated to 67.6 per cent of domestc abatement, A (see Table 1). The correspondng estmate for supply sde polces s 35.3 per cent of A S. As seen from Eq. (21), the margnal cost of global emssons reducton s equal to Eq. (22) dvded by 67.6 per cent for demand sde polces and to Eq. (23) dvded by 35.3 for supply sde polces. We explot these relatons to depct the optmum combnaton of demand sde and supply sde polces for a global contrbuton target, Ā(see Eq. (1)). We pck a target of 5 Mt of CO 2 by 2020 (equal to 1.9 mllon Sm 3 ), whch corresponds to about 10 per cent of Norway s present domestc greenhouse gas emssons. 15 Moreover, In Fgure 3 we show a bath tub dagram wth 15. The Norwegan government and parlament have hgh ambtons for domestc abatement by 2020. An agreement (The Clmate Settlement) sgned n 2008, and renforced n 2012, by the parlamentary majorty states that 2/3 of the emsson reducton from 1990 to 2020 wll take place wthn own borders. When assumng that Norway fulfls ths domestc target through abatng Norwegan non EU-ETS emssons, a 5Mt cut n global emssons n 2020 s n lne wth what can be acheved when the leakages are accounted for; see Fæhn et al. (2013).
92 / The Energy Journal Fgure 3: Combnng Leakage-adjusted emand and Supply Sde Margnal Cost Curves length equal to Ā =5Mt global emsson reducton, and where the margnal costs of global emsson reducton through demand (supply) sde measures are shown from left to rght (rght to left). The ntersecton pont between the two curves shows the optmal combnaton of demand and supply sde measures. We notce that 3.3 Mt CO 2, or about 2/3 of the global contrbuton target, should be met through reduced ol extracton. The remanng 1.7 Mt s met by demand sde polces. The correspondng margnal costs of reducng global CO 2 emssons through ths combnaton are 336 US per tonne 16. The fgure also reflects that f the global reductons were to be met through demand sde measures alone, costs would more than double. Implementng ths optmal combnaton of demand and supply sde measures would mean that domestc CO 2 emssons should be reduced by 2.5 Mt of CO 2. Gven our benchmark case, ths domestc abatement s necessary to obtan a global reducton of 1.7 Mt CO 2 snce global achevement s only 67.6 per cent of the domestc abatement. The domestc CO 2 reducton can be obtaned, e.g., by a domestc CO 2 tax on non-eu ETS emssons of 228 US per tonne CO 2 (cf. Fgure 1). Ths tax rate corresponds to the margnal cost of reducng global CO 2 emssons n optmum (336 US per tonne) multpled by the same share as explaned above (67.6 per cent). Almost 90 per cent of the demand measures that are proftable to carry out relates to transportaton, of whch reduced prvate transport accounts for 20 per cent and transton to more clmate frendly vehcles accounts for the rest. In the optmum, Norwegan ol extracton should be reduced by 3.5 mllon Sm 3 (one standard cubc meter, Sm 3, equals 6.29 barrels), whch s 3.1 per cent of total Norwegan ol producton n 2012. 3.5 mllon Sm 3 contans 9.2 Mt CO 2, however, wth benchmark assumptons only 16. 1US/tonne CO 2 = 0.42 US/barrel. 1 tonne CO 2 = 0.317 toe.
Clmate Polces n a Fossl Fuel Producng Country /93 35.3 per cent of ths, or 3.3 Mt, s the resultng global emsson reducton n the optmal equlbrum. Ths reducton can be acheved n dfferent ways, e.g., through a producton tax on Norwegan ol extracton. The optmal margnal cost of reducng global CO 2 emssons estmated above (336 US per tonne) corresponds to a producton tax of US 50 per barrel,.e., 60 per cent of the crude ol prce assumed. Ths tax s found by multplyng the margnal cost by the net effect on global emssons of reduced Norwegan extracton (0.353) and by the CO 2 content of a barrel of ol (0.42). As mentoned n Secton 3.1.2, the break-even prce of the Ivar Aasen ol feld, whch can be charactersed as relatvely proftable, s around US 60 per barrel. Below we dscuss the pros and cons of mplementng a producton tax. Here we want to emphasse that a producton tax of around US 50 per barrel could potentally lead to a much bgger reducton n ol extracton than the 3.1 per cent calculated above. The reason s that, as underlned n Secton 3.1.2, we most lkely overestmate the costs of reducng ol extracton. 3.3.2 Senstvty analyss There are other uncertantes n our calculatons, too, especally the effects n the fossl fuels markets of reduced ol extracton or consumpton n Norway. As our bref lterature revew n Appendx shows, both supply and demand elastctes vary a lot across emprcal studes, wth most estmates rangng between 0.1 and 1 n absolute value. For our purpose, the relatve dfference between supply and demand elastctes s the most mportant. OPEC s behavour s also somewhat uncertan. In Table 2 we present a number of senstvty analyses where we adjust assumptons from our benchmark case. The global contrbuton target s held fxed at 5 Mt CO 2. We notce from Table 2 that assumng compettve behavour by OPEC gves more or less the same results as above ths s not surprsng gven the results n Table 1. Besdes that, we see that whether demand sde polces or supply sde polces are most effectve n reducng global emssons depends qute a lot on what we assume about the ol market. If we thnk that OPEC keeps ts supply fxed, or f demand s twce as elastc as supply, cuts n ol extracton are even more effectve n reducng global emssons, and 90 per cent of the global emsson reductons results from supply sde measures. Nevertheless, the optmal producton tax does not change much. The global emsson effects of reduced ol extracton ( E x ) are ncreased, shftng the supply sde curve n Fgure 3 downwards. Lkewse, the demand sde curve n Fgure 3 shfts upwards. Stll, the ntersecton pont drops down, meanng that the shadow cost of the emsson constrant k (cf. Eqs.(19) (20)) declnes. However, the optmal tax on ol extracton s proportonal to E x (cf. Secton 2), whch has ncreased. The domestc CO 2 prce drops qute substantally, though, due to a combnaton of lower k and lower E y. If we thnk that supply s twce as elastc as demand, cuts n ol extracton s less effectve and the share of emsson reductons resultng from supply sde measures drop to 25 per cent. Agan, we see that the optmal producton tax s less affected, whle the domestc CO 2 prce has ncreased qute a lot. If supply s even more elastc relatve to demand, supply sde measures become even less attractve. The lterature revew n Appendx does not suggest so, but the large varaton n estmated elastctes mples that such a case cannot be ruled out. Related to ths, f OPEC for some reason chooses to keep the ol prce fxed, reduced ol extracton by the home country gves nsgnfcant global emsson reducton, as supply s merely shfted to other producers and the only effect wll arse from the small dfferences n emssons from extracton at the margn. The optmal choce wll be the conventonal one of only dong demand sde polces; see Table 2. As mentoned, we dsregard the fact that reduced supply of Norwegan ol would ncrease the ol prce margnally. Ths would reduce the costs of supply sde measures by roughly 10 per
94 / The Energy Journal Table 2: Senstvty Analyss. Effects of Reducng Norwegan Extracton or Consumpton of Ol by One Unt of CO 2 cent n the cost-effectve soluton depcted n Fgure 3 and slghtly ncrease the costs of demand sde measures. If we have overestmated the costs of reduced ol extracton, we should undertake even more supply sde measures than suggested by Fgure 3. Moreover, the optmal domestc CO 2 prce and the optmal producton tax for Norwegan ol extracton should then be reduced. For nstance, f we scale down the supply sde cost curve by 50 per cent, 83 per cent of total abatement should be caused by supply sde measures, wth the optmal domestc CO 2 prce and producton tax beng 126 US per tonne CO 2 and 28 US per barrel; see Table 2. On the other hand, we have gnored the challenges of separatng ol and gas extracton, whch may suggest that we have underestmated the forgone profts of reduced ol extracton. However, the share of gas n total ol and gas producton for the nne felds studed above was merely 5 per cent. Moreover, for 8 of the 13 felds currently under development on the Norwegan shelf, more than 90 per cent of recoverable reserves are ol (Mnstry of Petroleum and Energy, 2013). Hence, ths may be of lmted mportance. The hgher the ol prce, the less proftable t s to restrct extracton from a gven ol feld as the proft margn s hgher. As explaned n Secton 3.1.2, however, a hgher ol prce does not necessarly mean that supply sde measures become more costly f the hgher prce s expected by the ol producers, as more expensve resources wll then be extracted n the reference case. The costs of supply sde measures then depend on the proft margn of these more expensve resources. Most lkely, the margnal unt avalable for supply sde polces wll stll have a proft near zero and, thus, be a low-cost supply sde measure. Anyway, t s very unlkely that t s cost effectve to