1 CO Cos Pass Through and Windfall Profis in he Power Seor Jos Sijm, Karsen Neuhoff and Yihsu Chen May 006 CWPE 0639 and EPRG 067 These working papers presen preliminary researh findings, and you are advised o ie wih auion unless you firs ona he auhor regarding possible amendmens.
2 CO os pass hrough and windfall profis in he power seor Jos Sijm *, Karsen Neuhoff and Yihsu Chen 3 Energy researh Cenre of he Neherlands (ECN), P.O. Box 3754, 030 AD, Amserdam, The Neherlands Universiy of Cambridge, Fauly of Eonomis, Sidgwik Avenue, Cambridge CB3 9DE, UK 3 The Johns Hopkins Universiy, Deparmen of Geography and Environmenal Engineering, 3400 N. Charles S., Balimore, MD 8, USA 9 June 006 Absra In order o over heir CO emissions, power ompanies reeive mos of he required EU ETS allowanes for free. In line wih eonomi heory, hese ompanies pass on he oss of hese allowanes o he prie of eleriiy. This paper analyses he impliaions of he EU ETS for he power seor, noably he impa of free alloaion of CO emission allowanes on he prie of eleriiy and he profiabiliy of power generaion. Besides some heoreial refleions, he paper presens empirial and model esimaes of CO os pass hrough for Germany and he Neherlands, indiaing ha pass hrough raes vary beween 60 and 00 peren of CO oss depending on he arbon inensiy of he marginal produion uni and oher, marke or ehnology speifi faors onerned. As a resul, power ompanies realise subsanial windfall profis, indiaed by empirial and model esimaes presened in he paper. Keywords: Emissions rading; alloaion; CO os pass hrough; windfall profis; power seor JEL: C8, L,L94 * Corresponding auhor. Tel: address: For researh suppor we are graeful o Climae Sraegies, The Carbon Trus and he researh proje TSEC a he ESRC Eleriiy Poliy Researh Group, Universiy of Cambridge.
3 . Inroduion A major haraerisi of he presen EU Emissions Trading Sheme (ETS) is ha almos all CO allowanes are alloaed for free o he insallaions overed by he sheme. During he firs phase of he EU ETS ( ), more han. billion allowanes of onne eah are alloaed per year (EC, 005). During he firs phase of he EU ETS ( ), more han. billion allowanes of onne eah are being alloaed per year (EC, 005), abou 60% of whih is alloaed o he power seor. Agains his bakground, he presen paper analyses he impliaions of he EU ETS for he power seor, noably he impa of free alloaion of CO emission allowanes on he prie of eleriiy and he profiabiliy of power generaion. Firs of all, Seion disusses he effe of differen generaion ehnologies being used o generae eleriiy. How does he inernalisaion of CO allowane pries by individual generaors ino heir bids feed hrough o he power prie and how does his in urn affe profiabiliy? Subsequenly, Seions 3 o 5 presen empirial and model findings on passing hrough oss of CO emission allowanes o power pries in ounries of Norh-wesern Europe and he impliaions for he profiabiliy of power produion in hese ounries a he naional and firm level. Finally, he paper onludes wih a brief summary of he major findings and poliy impliaions.. Theory The EU ETS is a ap and rade sysem based primarily on a free alloaion of a fixed amoun of emission allowanes o a se of overed insallaions. Companies an eiher use hese allowanes o over he emissions resuling from he produion of hese insallaions or sell hem on he marke (o oher ompanies ha need addiional allowanes (Reinaud, 005)). Hene, for a ompany using an emission allowane represens an opporuniy os, regardless wheher he allowanes are alloaed for free or purhased a an auion or marke. Therefore, in priniple and in line wih eonomi heory, a ompany is expeed o add he oss of CO emission allowanes o is oher marginal (variable) oss when making (shor-erm) produion or rading deisions, even if he allowanes are graned for free (Burraw, Palmer e al., 00; Reinaud, 003; Burraw, Palmer e al., 005). Differen generaion ehnologies produe differen levels of CO emissions, and herefore he opporuniy oss of CO emissions per uni of power produed differ as well. For example, a Combined Cyle Gas Turbine produes abou 0.48 onnes of CO per MWh of eleriiy, while a ypial oal power saion emis abou 0.85 CO /MWh. A CO prie of 0 /CO, 3
4 herefore, inreases he generaion oss for he gas plan by 9.6 /MWh and for he oal plan by 7 /MWh. During a erain load period, he ompeiive eleriiy prie is only affeed by he prie inrease of he marginal produion uni. This an be illusraed by a marginal os (prie) duraion urve, as presened in Figure. On he X-axis he 8760 hours of a year are depied, sored in desending order of he marginal sysem oss. The Y-axis gives he marginal oss of he marginal generaion uni. The ompeiive eleriiy prie in any one hour is affeed by he ap and rade sysem hrough he prie inrease of he marginal uni. Hene, he amoun a whih he power prie inreases due o he passing hrough of CO oss may differ per hour or load period onsidered, depending on he marginal generaion uni onerned. As a onsequene, he CO oss pass hrough is defined as he average inrease in power prie over a erain period due o he inrease in he CO prie of an emission allowane. Prie/MWh 9 /MWh 9 /MWh CO Add On CO Work on Oil CCGT Gas Coal 8760 Hours/year Average E leriiy prie Figure Pass hrough of CO opporuniy oss for differen load periods (a a prie of 0 /CO ) We represen he differene beween he behaviour of individual generaors and he impa on he sysem prie by defining he add-on and he work-on rae. In a ompeiive environmen, generaors add-on he opporuniy oss of CO allowanes o he power prie. The inrease of he bid of he marginal uni will hen deermine how muh of he CO allowane pries are worked-on he eleriiy prie. However, in a liberalised marke, pries are ulimaely deermined by a omplex se of marke fores. As a resul, he work-on rae may be lower han he add-on rae. One reason why he work-on rae may be lower han he add-on rae is marke demand response. If higher power pries redue eleriiy demand, hen an expensive power saion 4
5 migh no need o operae and a heaper generaor will se he marginal prie. The hange in power prie is smaller han he hange in marginal oss due o emissions rading. Hene, while he add-on rae will remain 00 peren, he work-on rae will be lower han 00 peren. However, prie response is ypially raher low for households and oher small-sale onsumers of eleriiy, bu may be more signifian for major end-users suh as he powerinensive indusries. Power-inensive indusry would subsiue eleriiy purhases wih selfgeneraion of eleriiy. This pahway is less araive, as he EU ETS also overs largesale self-generaion by indusry and, herefore, faes similar os inreases, hus reduing demand response of power-inensive indusry. Neverheless, hrough self-generaion powerinensive indusry would benefi from he eonomi ren due o he ransfer of valuable, freely alloaed asses. The exen o whih arbon oss are passed hrough o power pries depends also on hanges in he meri order of he supply urve due o emissions rading. This an be illusraed by Figure, where he supply urve is haraerised by a sep funion wih wo ypes of ehnologies - A and B. The verial dash line indiaes he fixed demand. In he lef par of Figure, when here is no hange in he meri order, he hange in he power prie (Δp ) will always be equal o he marginal CO allowanes oss of he marginal generaion ehnology B. The resuling pass-hrough rae will always be uniy (in erms of boh he add-on rae and he work-on rae). However, when here is a swih in he meri order - as displayed in he righ par of Figure, he siuaion beomes differen. In his ase, he marginal ehnology is A wih CO allowanes oss equal o Δp 3 while he hange in he power prie is Δp 4. Therefore, while he add-on rae for he marginal produion ehnology A is 00 peren, he work-on rae Δp 4 /Δp 3, will be less han sine Δp 4 < Δp 3. In markes wih surplus apaiy, ompeiive pressures from exess generaion apaiy also impa he meri order and in urn, he work-on rae (Reinaud, 003). Model analyses show ha when CO oss exeeds 0 /onne, emissions rading would indue subsanial hanges in he produion meri order (Sijm e al., 005). 5
6 a b /MWh /MWh Δp Δp Δp 4 Δp 3 A B B A Figure Pass-hrough raes under hanges in he meri order In addiion, here may be several reasons why generaors do no add on he full CO oss o heir power bid pries: The expeaion of power produers ha heir urren emissions or oupu will be used as an inpu faor for he deerminaion of he alloaion of allowanes in fuure periods, mainly afer 0 bu possibly even This reaes an inenive o inrease oday s oupu and hus indues generaors o no add on he full allowane prie o heir energy bids. Volunary agreemens or he regulaory hrea of governmens o inervene in he marke if generaors make exessive windfall profis from he free alloaion migh indue generaors o limi he add-on. Oher reasons suh as he inidene of non-opimal behaviour among power produers, marke imperfeions, ime lags or oher onsrains, inluding he inidene of risks, unerainies, lak of informaion, and he immauriy or lak of ranspareny of he arbon marke. The impa on generaors profis An imporan quesion is how he pass hrough of CO opporuniy oss affes he profiabiliy of power saions. A main purpose of free alloaion of emissions allowanes under he US ap and rade programmes for SO and NOx as well as under he EU ETS for CO is o obain he poliial suppor of large emiers. Thus, he free alloaion aims o ensure ha he inroduion of he ETS does no redue profiabiliy of he eligible ompanies. The impa of emissions rading in general and free alloaion of emission allowanes in pariular an be illusraed by means of Figure. This figure illusraes he impliaions of For a full disussion and illusraion of hese reasons, see Chaper 4 of Sijm e al. (005). 6
7 emissions rading for generaors profis in ase he supply urve onsiss of differen ypes of ehnology. In ase emissions rading does no lead o a hange in he meri order of he supply urve (and in oal demand; see lef hand side of Figure ), he hange in he power prie (Δp ) is jus equal o he CO oss per MWh of he marginal produion uni (B). For his uni, his implies ha profis do no hange in ase all he allowanes have o be bough, while i resuls in windfall profis in ase of full grandfahering (equal o Δp imes volume produed). For he infra-marginal uni, however, he impa of emissions rading on operaional profis does no only depend on he degree of grandfahering bu also on wheher i is more or less arbon-inensive han he marginal uni. If i is less arbon-inensive, i benefis from he fa ha he ET-indued inrease in power prie is higher han he inrease in is arbon oss per MWh. However, if he infra-marginal uni is more arbon-inensive han he marginal uni, i suffers from a loss, as he inrease in power prie is lower han he inrease in is arbon oss per MWh, noably if allowanes have o be bough on he marke. Therefore, in he laer ase, some grandfahering o his infra-marginal uni may be jusified o break even, depending on he relaive arbon-inensiy of his uni. On he oher hand, if emissions rading leads o a hange in he meri order (while oal demand remains he same; see he righ par of Figure ), he hange in he power prie (Δp 4 ) is lower han he hange in he CO oss per MWh of he marginal produion uni (A). For his uni, emissions rading resuls in a windfall profi per MWh (equal o Δp 4 ) in ase of free alloaion, bu in a loss (equal o Δp 3 Δp 4 ) if all he allowanes have o be bough. Therefore, for his uni, some grandfahering may be jusified o break even. 3 For he inframarginal uni (B), he inrease in power prie is higher han he inrease in CO oss, regardless of wheher allowanes have o be bough or no. Therefore, even if his uni has o buy all is allowanes, i will benefi from a windfall profi and, hene, here is no need for any grandfahering o his uni o break-even. 4 If eleriiy demand response o ET-indued prie inreases is suffiienly large o sop he operaion of a se of power generaors wih higher variable oss and hus he marke learing prie of eleriiy is redued o he variable oss of a ehnology wih lower variable oss, 3 4 I should be observed, however, ha he hange in he meri order migh our only during a erain load period. This has o be aouned for when analysing he impa of emissions rading on firms profis and he impliaions for assessing he exen of grandfahering o break even. Similar findings an be derived by means of Figure, showing differen ypes of ehnology along he load duraion urve. By omparing he revenues (prie/mwh * hours loaded) and he orresponding real/opporuniy oss wih and wihou emissions rading, hanges in profis an be derived for differen ypes of ehnology, inluding a hange in he meri order of hese ehnologies. 7
8 his will redue he profis of all unis operaing during his period, as all of hem will reeive revenues orresponding o he lower marke learing prie. 3. Empirial esimaes of passing hrough CO oss This seion presens some empirially esimaed raes of passing hrough CO opporuniy oss of EU emissions rading o power pries in Germany and he Neherlands. We use wo differen approahes o esimae hese raes. Firs, we look a he forward power marke, pariularly he year ahead marke where, for insane, eleriiy delivered in 006 is raded during every day of he year 005. In his approah, we assess he exen o whih hanges in forward power pries an be explained by hanges in underlying forward pries for fuel and CO allowanes. Seondly, we sudy he spo marke, noably he German power exhange (EEX), by omparing hourly spo eleriiy pries for he period January 005 ill Marh 006 wih he orresponding hourly eleriiy pries in he year 004. More speifially, we look o wha exen a hange in he spo power prie, for example a 9am on he firs Monday in January 006 relaive o he firs Monday in January 004, an be explained by a hange in he prie of a CO allowane on he EUA marke. Firs of all, however, some bakground informaion will be provided on rends in pries for fuel and CO allowanes as wekk as dark and spark spreads in he power seor of Germany and he Neherlands during he years Trends in forward pries and oss For he years , Figures 3 and 4 presen power pries versus fuel and CO oss o generae one MWh of power (assuming a fuel effiieny of 40 peren for oal and 4 peren for gas, a relaed emission faor of 0.85 and 0.48 CO /MWh for oal and gas, respeively, and full opporuniy oss for generaing eleriiy by eiher oal or gas). While Figure 3 overs he ase of oal-generaed off-peak power in Germany, Figure 4 presens he ase of gas-generaed peak power in he Neherlands. 5 5 In his seion, unless oherwise saed, oal refers o he inernaionally raded ommodiy lassified as oal ARA CIF AP#, while gas refers o he high alori gas (wih a onversion faor 35,7 GJ/m 3 ) from he Duh Gas Union Trade & Supply (GUTS). Moreover, pries for power, fuels and CO refer o forward markes (i.e. year-ahead pries). 8
9 [ /MWhe] Germany 40 Offpeak 0 Coal oss CO oss Figure 3 Off-peak power pries versus fuel/co oss in Germany (year ahead, ) [ /MWhe] 80 he Neherlands Offpeak Coal Gas oss 0 CO oss Figure 4 Peak power pries versus fuel/co oss in The Neherlands (year ahead, ) The German ase shows ha he fuel (i.e. oal) oss o generae power have been more or less sable a a level of abou 6 /MWh during he years In addiion, CO oss of oal-generaed power have been sable during he seond par of 004 bu have approximaely ripled during he firs par of 005 from abou 6 /MWh in January o some 8 /MWh in July. This suggess ha he inreasing off-peak pries in Germany over his period may have been aused primarily by he rising CO pries (and no by higher fuel pries). However, during he seond par of 005 (Augus-Deember 005) CO oss per oalgeneraed MWh have been raher sable while off-peak pries have oninued o rise. This indiaes faors oher han fuel and CO oss influene power pries. The Duh ase illusraes ha he fuel (i.e. gas) oss o produe eleriiy have risen subsanially from some 33 /MWh in early January 005 o abou 56 /MWh in early 9
10 Sepember 005. CO oss of gas-generaed power have also inreased over his period, bu less dramaially, i.e. from 4 o /MWh (parly due o he relaively low - bu onsan - emission faor of gas-generaed eleriiy). This suggess, hene, ha besides he CO os pass hrough he rising peak load pries in he Neherlands over his period - from abou 5 o 80 /MWh - are largely due o oher faors, espeially he rising gas pries. However, omparable o he German ase, whereas boh gas and CO oss have been more or less sable during he las quarer of 005 (or even delined a bi as far as gas oss are onerned), peak power pries oninued o inrease o 84 /MWh in lae Deember 005. Trends in dark and spark spreads on forward markes Figures 5 and 6 presen rends in dark/spark spreads and CO oss per MWh over he years in Germany and he Neherlands, based on forward (i.e. year ahead) pries for power, fuels and CO emission allowanes. For he presen analysis, a dark spread is simply defined as he differene beween he power prie and he os of oal o generae a MWh of eleriiy, while a spark spread refers o he differene beween he power prie and he oss of gas o produe a MWh of eleriiy. If he oss of CO are inluded, hese indiaors are alled lean dark/spark spreads or arbon ompensaed dark/spark spreads. 6 [ /MWh] 48 DE peak/dark spread Peak [ /MWh] 3 DE off-peak/dark spread Offpeak 3 6 CO oss 6 CO oss Figure 5 Trends in dark spreads and CO oss per oal-generaed MWh in Germany during peak and off-peak hours (year ahead, ) 6 These spreads are indiaors for he overage of oher (non-fuel/co ) oss of generaing eleriiy, inluding profis. For he presen analysis, however, hese oher oss - for insane apial oss, mainenane or operaing oss - are ignored as, for eah speifi ase, hey are assumed o be onsan for he (shor-erm) period onsidered - alhough hey may vary per ase onsidered - and, hene, hey do no affe he esimaed pass-hrough raes. 0
11 [ /MWh] 48 NL peak/spark spread Peak (spark spread) [ /MWh] 3 NL off-peak/dark spread Offpeak (dark spread) CO oss CO oss Figure 6 Trends in spark/dark spreads and CO oss per gas/oal-generaed MWh in he Neherlands during peak and off-peak hours (year ahead, ) For Germany, Figure 5 depis rends in dark spreads in boh peak and off-peak hours, based on he assumpion ha a oal generaor is he prie-seing uni during hese periods. 7 In addiion, i shows he oss of CO allowanes required o over he emissions per MWh generaed by a oal-fired power plan (wih an emission faor of 0.85 CO /MWh). The figure suggess ha up o July 005 hanges in he dark spread an be largely explained by hanges in he CO oss per MWh. Sine Augus 005, however, his relaionship is less lear as he CO oss have remained more or less sable, while he dark spreads have oninued o inrease rapidly. For he Neherlands, Figure 6 depis rends in he spark spread during he peak hours and he dark spread during he off-peak hours, based on he assumpion ha a gas- versus oal-fired insallaion is he prie-seing uni during hese periods, respeively. In addiion, i presens he oss of CO allowanes o over he emissions per MWh produed by a gas- and oalfired power saion, wih an emission faor of 0.48 and 0.85 CO /MWh, respeively. Similar o he German ase, Figure 7 suggess ha, during he period January-July 005, hanges in he dark/spark spreads in he Neherlands an be largely aribued o hanges in he CO oss per MWh, bu ha aferwards his relaionship is less lear. Saisial esimaes of CO os pass hrough raes on forward markes Below, we provide empirial esimaes of pass hrough raes of CO emissions rading oss o forward power pries in Germany and he Neherlands for he period January-Deember 005. The basi assumpion of esimaing CO os pass hrough raes is ha during he observaion period he dynamis of he power pries in Germany and he Neherlands an be fully explained by he variaions in he fuel and CO oss over his period (see Figures 3 and 4). Hene, i is assumed ha during his period oher oss, for insane operaional or 7 I is aknowledged, however, ha during erain periods of he peak hours - he super peak - a gas generaor is he marginal (prie-seing) uni, bu due o lak of daa, i is no possible o analyse he super peak period in Germany separaely.
12 mainenane oss, are onsan and ha he marke sruure did no aler over his period (i.e. hanges in power pries an no be aribued o hanges in ehnology, marke power or oher supply-demand relaionships). Based on hese assumpions, he relaionship beween power pries (P), fuel oss (F) and CO oss is expressed by equaion (), where supersrips and g indiae oal and gas, respeively. Likewise, he erm CO is he CO os assoiaed wih oal and gas a ime. Thus, i is equal o he produ of he CO allowanes prie a ime and he ime-invarian CO emission rae of oal or gas generaors. In our analysis, fuel oss are assumed o be fully passed on o power pries. 8 This is equivalen o fixing he oeffiien β a uniy. P = α + βco + β F + ε () g, g, By defining Y as he differene beween power prie and fuel os, equaion () beomes he enral regression equaion of whih he oeffiien β has been esimaed. In fa, Y represens he dark spread for oal-generaed power and he spark spread for gas-generaed power. Y = ( P F ) = α + βco + ε () g, g, Like mos prie series, power prie daa exhibi serial orrelaion. Hene, he error erm ε is haraerised by a so-alled I(0) proess (inegraed of order zero). 9 ε = ε + u, (3) where u is a purely random variable wih an expeed value of zero, i.e. E(u ) = 0, and a onsan variane over ime, i.e., Var(u ) = σ. 8 9 In Sijm e al. (006), his assumpion was dropped, bu i urned ou ha he esimaed pass hrough raes for fuel and CO oss were unreliable due o he observaion ha fuel and CO oss are highly orrelaed. An I(0) (inegraed of order zero) is an auoregressive proess wih one period of lag, i.e., AR() and wih a propensiy faor < (see equaion (3)) (Sewar and Wallis, 98). I indiaes a proess of orrelaion frequenly experiened in every day s life. For insane, if he ambien emperaure was high yeserday and here are no major hanges in he weaher ondiions, he emperaure oday should be more or less similar. In a ase, he emperaure oday provides a prior belief from whih omorrow s emperaure an be inferred. Saisially, when assuming ε an I(0) proess (i.e., <) in equaion (3), he series is a leas weakly independen. Therefore, boh PW and OLS will be adequae o esimae pass-hrough raes given he orre speifiaion. However, we are aware of he possibiliy of a non-oinegraion proess sine hree series power pries, fuel oss and CO oss follow an I() proess based on Dikey-Fuller Tes. Thus, in his paper, we inend o provide a preliminary assessmen of he empirial CO pass hrough raes.
13 In 005, eleriiy forward onras were raded a he German power exhange EEX only a limied number of days. For he remaining days a selemen prie was repored based on he hief rader priniple. This requires all hief raders o daily submi a spreadshee wih heir evaluaion of pries for more han 40 differen onra ypes. I is unlikely ha all onra ypes would be updaed on a daily basis in ommensurae wih CO pries. Sine he differen proools used by various ompanies reporing power pries are proprieary informaion, we do no posses suh informaion and are unable o onsider i in he esimaion proedure. Thus, o illusrae he effe, we assume in he appendix ha he repored pries are a weighed average over he pries during he previous days or weeks. Esimaing () wihou onsidering his reaes an error on he lef hand side of he equaion ha we are esimaing. This error reaes a bias in he esimaion of β. This bias exiss if we esimae β using an ordinary leas square esimaion, bu an inrease signifianly if we esimae a non-oinegraed proess based on (3) using oher approahes ha ieraively deermine boh β and. The alernaive approah we would usually apply in suh a siuaion is an esimaion based on he firs differenes. Bu one again we show in he appendix ha he error on he lef hand side of he equaion an reae a very srong bias in his esimaion. Hene, we onlude ha he leas affeed alernaive is a simple OLS esimaor. We aep ha we have an esimaor ha migh slighly underesimae he CO pass hrough. We are somewha onerned abou he fa ha prie series are ypially auo-orrelaed. In fa, boh power pries and CO oss series are I() proesses. Thus, if CO and eleriiy prie series are no also oinegraed, hen he error erms follow an I() proess and will fail o onverge o zero. However, sine boh forward eleriiy pries and CO pries are bounded, i urns ou o be less of an issue in our analyses. Finally, we know ha he ypial onfidene inervals repored by our esimaion will no longer appropriaely represen he unerainy in he esimaion. Therefore, we apply boosrapping o illusrae he auray of our esimaion. In pariular, we esimae β using he daa from a resried observaion period, hus we an examine he robusness of he esimaion. More speifially, we firs onsru a subse of daa for boosrapping (e.g. January-Oober). We repea his proess by sliding he wo-monh window (e.g., January-February merged wih Marh-April, May-Deember, e.), resuling in a oal of six regressions wih boosrapped daa. Table summarizes he esimaed CO pass hrough raes in Germany and he Neherlands and also gives he maximum and minimum of he OLS esimaor assoiaed wih various boosrapping esimaions. Wih onfidene of abou 80 peren, we an say ha hese raes are wihin he inerval of 60 and 7 peren in Germany, and beween 64 and 8 peren in he Neherlands. In ligh of he aforemenioned mehodologial diffiulies, he resuls 3
14 presened in Table need o be inerpreed and reaed wih auion. In pariular, we offer some explanaions of possible omplexiies and disuss he poenial direion of bias: Table Empirial esimaes of CO pass hrough raes in Germany and he Neherlands for he period January-Deember 005, based on year ahead pries for 006 (in %) Counry Load period Fuel (effiieny) OLS Boosrap ( monhs) min max Germany Peak Coal (40%) Off-peak Coal (40%) Neherlands Peak Gas (4%) Off-peak Coal (40%) Firs, he very high pass hrough rae for Germany migh parially be explained by inreasing gas pries during 005. Given ha gas generaors (insead of oal generaors) se he marginal prie in Germany markes during some peak hours, his ould onribue o power pries inrease in peak forward onras. As oal generaors benefi from his gas os-indued inrease in power pries, i leads o an overesimae of he pass hrough rae of CO oss for oal-generaed power. Finally, Sijm e al. (005 and 006) presen and disuss a wide variey of furher esimaions of CO pass hrough raes. In general, he esimaions based on he period January-July 005 resul in lower pass hrough raes han esimaions based on he period 005 as a whole. For insane, he pass hrough rae for he Neherlands peak hours is esimaed a 38 peren for he period January-July 005, while i is esimaed a 78 peren for 005 as a whole. This differene in esimaed pass hrough raes beween he period January-July and 005 as a whole ould possibly be aused by some delays in he marke inernalising he CO prie (i.e. marke learning), rapidly rising gas pries (noably during he firs period of 005), higher power pries due o inreasing sariy and/or marke power (pariularly during he laer par of 005), or by various oher faors affeing power pries in liberalised wholesale markes. Empirial esimae using he hourly spo markes in Germany Anoher approah o assess he impa of he CO allowane oss on he wholesale power prie is o ompare he day-ahead eleriiy pries per hour on he German power exhange (EEX for every day in 005) wih he orresponding pries in 004. The implii assumpion is ha faors oher han CO and fuel oss remain unhanged during hese wo years. Aording o equaion (4), he differene in he eleriiy prie during a erain hour afer he inroduion of he ETS and he orresponding hour in 004 is explained by he differene in oal pries during he hours onerned, he impa of he CO prie on he EUA marke and by an error erm. 4
15 p p = p p + β p + ε (4) oal oal o 005, 004, 005, 004,5 005, We se p o, o refle he oss of CO emissions a he daily allowane prie for a oal power saion wih an emission rae of 0.9 CO /MWh. As oal is a he margin during mos of he day, his an hen also be inerpreed as he work-on rae for oal power saions. Figure 7 depis β for differen hours of he day. We have spli he observaion period in hree seions, mainly o examine wheher he daily paern is onsisen over ime. While his paern did no hange during he day, he level of work on rae inreased for eah subsequen period onsidered..5 Hour 6 ( ) values for period shown in Char s quarer nd half 005 s half Hour 00-0 Hour Hour Hour Hour Hour 0 - Hour 3-3 Hour Hour Hour Hour 0 - Hour 3-3 Figure 7 Work-on rae of CO oss on he German spo power marke for differen ime periods, assuming oal generaors are a he margin The figure invies hree observaions. Firs, during off-peak periods he work-on rae seems o be lower han one. This ould be parially explained by ineremporal onsrains of power saions hey prefer o operae during off-peak periods if his saves sar up oss. As CO oss inrease he sar up oss, hey also reae addiional inenives o lower pries during off-peak periods o reain he saion running (Muesgens and Neuhoff, 006). Seond, if oal generaors se he prie during peak periods, hen hese are usually vinage saions wih higher hea raes and herefore higher emission oss. Finally, he inrease of gas pries during he year is likely o also explain some of he prie inrease during peak periods. 5
16 As open spinning yle gas urbines migh be alled during some peak periods, heir inreased oss wih higher gas pries an furher push up he power prie. Therefore we now fous on he hour 3-4 pm for whih ineremporal effes and he gas-prie impa from peaking unis running a maximum a few hours a day is leas prevalen, as indiaed by he lower value for his hour relaive o oher peak hours in Figure 7. Figure 8 depis for eah day he prie inrease of eleriiy in he hour 3-4 pm relaive o he pre-ets year 004. The urves are again orreed for oal pries and herefore de fao oal oal depi: 005, 004, 005, 004, p p ( p p ). Euro/MWh Change in dark spread CO oss for oal Jan-05 Feb-05 Apr-05 May-05 Jul-05 Sep-05 O-05 De-05 Feb-06 Figure 8 Coal prie orreed prie inrease for eleriiy (3-4 pm) depied as dos and heir 40day moving average (volaile (dark) line) and he evoluion of he CO prie (grey line) As an be noed in Figure 8, during January no he enire CO prie was passed hrough, bu subsequenly a lose link seems o exis beween he inrease of he CO os and he inrease of he eleriiy prie relaive o 004. In Sepember he publi debae in Germany evolved abou wheher he inlusion of CO opporuniy oss ino he eleriiy prie is appropriae, and indued generaion ompanies o exerise wih some auion. I seems ha evenually power firms managemen ook he posiion ha any oher behaviour han pass hrough profi maximisaion is inappropriae, and publily aknowledged suh behaviour, hus allowing raders o reurn o he habi of fully inernalising he CO opporuniy oss. By he end of he year 005 he German eleriiy pries furher inreased. We did no analyse he reasons for his developmen. The prie inrease ould be aribued o one of he 6
17 following hree faors, (i) sariy of generaion apaiy, (ii) higher gas pries han in previous winers, hus higher pries when gas is a he margin, and (iii) he exerise of marke power. Looking a he overall piure suggess ha marke pariipans in Germany have fully passed hrough he opporuniy oss of CO allowanes in he spo marke. 4. Model esimaes of CO os pass hrough In addiion o he empirial esimaes, CO os pass hrough raes have been esimaed for some EU ounries by means of he model COMPETES. 0 COMPETES is basially a model o simulae and analyse he impa of sraegi behaviour of large produers on he wholesale marke under differen marke sruure senarios (varying from perfe ompeiion o oligopolisi and monopolisi marke ondiions). The model has been used o analyse he impliaions of CO emissions rading for power pries, firm profis and oher issues relaed o he wholesale power marke in four ounries of oninenal Norh-wesern Europe (i.e. Belgium, Frane, Germany and he Neherlands). The major findings of he COMPETES model wih regard o CO os pass hrough are summarised in Table. They are ompared o model resuls from he Inegraed Planning Model (IPM), whih are desribed in more deail in Neuhoff e al. (006). As resuls are very sensiive o he gas/oal shif, small differenes in he assumpions abou gas pries, available gas generaion apaiy and ineronneion apaiy an explain he differenes beween boh models resuls for he Neherlands. Under all senarios onsidered, power pries urn ou o inrease signifianly due o CO emissions rading. In ase of a CO prie of 0 /onne, hese inreases are generally highes in Germany (3-9 /MWh) wih an inermediae posiion for Belgium (-4 /MWh) and he Neherlands (9- /MWh). The model predis very low prie inreases for Frane (-5 /MWh), whih refles he predominan nulear generaion basis of his ounry. 0 COMPETES sands for COmprehensive Marke Power in Eleriiy Transmission and Energy Simulaor. This model has been developed by ECN in ooperaion wih Benjamin F. Hobbs, Professor in he Whiing Shool of Engineering of The Johns Hopkins Universiy (Deparmen of Geography and Environmenal Engineering, Balimore, Maryland, USA). For more deails on his model, see Sijm e al. (005) and referenes ied here, as well as websie hp://www.eleriiymarkes.info. 7
18 Table Model esimaes of eleriiy prie inreases (in /MWh) due o CO oss a 0 / Belgium Frane Germany Neherlands Unied Kingdom COMPETES IPM Differenes in absolue amouns of CO os pass hrough beween he individual ounries onsidered an be mainly aribued o differenes in fuel mix beween hese ounries. For insane, during mos of he load hours, power pries in Germany are se by a oal-fired generaor (wih a high CO emission faor). On he oher hand, in Frane hey are ofen deermined by a nulear plan (wih zero CO emissions), while he Neherlands ake an inermediae posiion - in erms of average CO emissions and absolue os pass hrough - due o he fa ha Duh power pries are se by a gas-fired insallaion during a major par of he load duraion urve. In relaive erms, i.e. as a perenage of he full opporuniy oss of EU emissions rading, COMPETES has generaed a wide variey of pass hrough raes for various senarios and load periods analysed. While some of hese raes are low (or even zero in ase he power prie is se by a nulear plan), mos of hem vary beween 60 and 80 peren, depending on he ounry, marke sruure, demand elasiiy, load period and CO prie onsidered. In addiion, Table provides he resuls of simulaion runs by he Inegraed Planning Model (IPM), a deailed power seor model for he EU developed by ICF Consuling. A a prie of 0 /CO, he average amoun of CO os pass hrough in he UK is esimaed a 3-4 /MWh, while for Germany and he Neherlands his amoun is esimaed a 7 and 5 /MWh, respeively. 5. Esimaed windfall profis As COMPETES inludes deailed informaion a he operaional level for all (major) power ompanies in he ounries overed by he model, i an also be used o esimae he impa of emissions rading on firms profis a he aggregaed level as well as a he level of major individual ompanies. Suh quaniaive resuls are helpful o undersand he qualiaive impa, bu he numbers should only be aken as an indiaion of he order of magniude involved. We will furher disuss his aspe a he end of he seion. The high esimae for he Neherlands (ompared o a similar esimae by he COMPETES model) migh be aused by he older naure of he IPM model wih oal having a sronger influene on power pries. 8
19 Table 3 presens a summary of he hanges in oal firms profis due o emissions rading under wo senarios, i.e. perfe ompeiion (PC) and oligopolisi ompeiion, i.e. sraegi behaviour by he major power produers (ST). These ET-indued profi hanges an be disinguished ino wo aegories:. Changes in profis due o ET-indued hanges in produion oss and power pries. This aegory of profi hanges is independen of he alloaion mehod. Aually, he esimaion of his aegory of profi hanges is based on he assumpion ha all ompanies have o buy heir allowanes and, hene, ha CO oss are real oss.. Changes in profis due o he free alloaion of emission allowanes. Aually, his aegory of profi hanges is an addiion or orreion o he firs aegory for he exen o whih allowanes are grandfahered - raher han sold - o eligible ompanies. Table 3 Changes in aggregaed power firms profis due o CO emissions rading in Belgium, Frane, Germany and he Neherlands, based on COMPETES model senarios Senario a Prie elasiiy Toal profis Change in profis due o: prie effes free alloaion Toal hange in profis due o emissions rading [M ] [M ] [M ] [M ] [%] () () (3) (4) (5) (6) (7) PC0-ze PC0-ze PC PC ST0-le ST0-le ST ST a) PC and ST refer o wo differen model senarios, i.e. perfe ompeiion (PC) and oligopolisi (or sraegi) ompeiion (ST). Numbers aahed o hese abbreviaions, suh as PC0 or PC0, indiae a senario wihou emissions rading (CO prie is 0) versus a senario wih emissions rading (a a prie of 0 /CO ). The addiions 'ze' and ' le' refer o a zero prie elasiiy and low prie elasiiy (0.), respeively, ompared o he baseline senario wih a prie elasiiy of 0. We sar wih he analysis of he impa of a perfely ompeiive environmen. In he fourh olumn of Table 3, i is assumed ha all firms have o buy all heir emissions allowanes on he marke, i.e. here are no windfall profis due o grandfahering. Even under his ondiion, oal firm profis inrease in he perfe ompeiion senarios. This resuls from he fa ha, on average, power pries are se by marginal unis wih relaively high arbon inensiies ha pass heir relaively high arbon oss hrough o hese pries. Infra-marginal unis wih relaively low arbon inensiies are no faed by hese high arbon oss bu benefi from he higher power pries on he marke. Profis inrease by 6 billion if we assume no demand 9
20 response, and by billion if we assume a very srong demand response of 0.. Thus, he high demand elasiiy senarios (i.e., 0. and 0.) provide a lower bound esimaion of windfall profis. Noe, ha in he long erm invesmen equilibrium we expe a fixed raio beween demand and he number of power saions, and hene a reduion of demand will no affe profiabiliy of individual power saions. The oal profis of power generaors are obviously furher inreased if we onsider he impa of he free alloaion (olumn 5), as is illusraed in olumn (6). The model provides addiional insighs ino he impa of sraegi behaviour of power generaors. If we assume power generaors a sraegially, hen hey will push up pries and, herefore, heir profis already double in he referene ase ha ignores emission rading. I would be subje o furher empirial work o assess o wha exen his level of profis orresponds o he siuaion before he inroduion of emission rading. If we now inrodue emission rading ino he model senarios, hen profiabiliy in he absene of free alloaion is slighly redued (by less han %). While all generaors profi from he higher pries, he effe of a smaller marke dominaes his effe and herefore slighly redues heir revenues. COMPETES is based on he assumpion of a linear demand funion whih implies a lower rae of passing hrough under oligopolisii ompeiion han ha in ompeiive markes. If onsan elasiiy of demand supply were assumed in he model, hen higher pass hrough raes han ompeiive markes would resul. These lower pass hrough raes in he ase of oligopolisi ompeiion explain why profis due o emissions rading (exluding free alloaion) are slighly redued in he ST senarios. Noe, however, ha due o sraegi behaviour, profis in he referene ST senario are signifianly higher han in he CP senario. Free alloaion (olumn 5) one again makes all senarios very profiable for power indusry (olumn 6). Under he presen EU ETS, however, ompanies do no have o buy heir emission allowanes on he marke bu reeive hem largely for free. This implies ha hey are able o realise windfall profis due o grandfahering as hey sill pass on he arbon oss of grandfahered emission allowanes. The fifh olumn of Table 3 shows esimaes of hese profis, based on esimaes of oal firms CO emissions and he assumpion ha power ompanies reeive, on average, 90 peren of he allowanes o over heir emissions for free. A a prie of an emission allowane of 0/CO, hese windfall profis vary beween 5.3 and 7.7 billion, depending on he senario onsidered. As oal produion and oal emissions are generally This implies an inrease of he wholesale prie level from 0 /MWh o 30 /MWh, and hene of he reail prie level (inluding ransmission, disribuion and markeing oss) from les say 70 /MWh o 90 /MWh, would resul in a reduion of demand by 0 peren. 0