The Sources of Emission Reductions: Evidence from U.S. SO 2 Emissions from 1985 through by A. Denny Ellerman and Florence Dubroeucq

Transcription

1 The Sources of Emsson Reductons: Evdence from U.S. SO 2 Emssons from 1985 through 22 by A. Denny Ellerman and Florence Dubroeucq 4-1 January 24

2 ABSTRACT An endurng ssue n envronmental regulaton s whether to clean up exstng old plants or n some manner to brng n new clean plants to replace the old. In ths paper, a unt-level data base of emssons by nearly 2 electrc generatng unts from 1985 through 22 s used to analyze the contrbuton of these two factors n accomplshng the sgnfcant reducton of sulfur doxde emssons from these sources n the Unted States. The effect on SO 2 emssons of the new natural-gas-fred, combned-cycle capacty that has been ntroduced snce 1998 s also examned. The results ndcate that cleanng up the old plants has made by far the greatest contrbuton to reducng SO 2 emssons, and that ths contrbuton has been especally large snce the ntroducton of the SO 2 cap-and-trade program n The new natural-gas-fred, combned cycle unts have dsplaced conventonal generaton that would have emtted about 8, tons of SO 2 ; however, the effect has not been to reduce total SO 2 emssons snce the 9. mllon ton cap s unchanged, but to reduce the quantty of abatement requred of other unts n meetng the cap and thereby the cost of dong so.

3 SOURCES OF SO 2 EMISSION REDUCTIONS 2 The Sources of Emsson Reductons: Evdence from U.S. SO 2 Emssons from 1985 through 22 A. Denny Ellerman Florence Dubroeucq 1 INTRODUCTION Emssons can be reduced by emsson rate reductons at exstng plants or by dsplacng those plants by other plants, frequently new unts, wth lower emsson rates. Accordngly, one of the endurng questons underlyng polces amed at reducng ar emssons s the role of these two ways of reducng emssons. A good case study for analyzng ther relatve contrbutons s provded by the experence of the Unted States n reducng sulfur doxde (SO 2 ) emssons from the combuston of fossl fuels for the generaton of electrcty. These emssons have been reduced by about 45%, from a peak of about mllon tons n 1975 when the Clean Ar Act Amendments of 197 became effectve to 1.1 mllon tons n 22, the last year for whch data s avalable. 2 Snce 197, SO 2 emssons have been subject to two dstnctly dfferent regulatory regmes establshed respectvely by the Clean Ar Act Amendments of 197 and 199. The Clean Ar Act Amendments of 197. These amendments nsttuted a coherent and effectve regulatory system for reducng SO 2 emssons whereby 1 Ellerman s the executve drector of the Center for Energy and Envronmental Polcy Research (CEEPR) and senor lecturer at the Sloan School of Management at the Massachusetts Insttute of Technology (MIT). Dubroeucq s a canddate for the master s degree n Technology and Polcy at MIT. Fundng for ths research from the Envronmental Protecton Agency (STAR grant #R82863) and from CEEPR s gratefully acknowledged. 2 The decrease n SO 2 emssons from all sectors of the economy was slghtly larger due to the dsappearance of metals processng, mostly copper, wthn the Unted States. For the economy as a whole, peak SO 2 emssons were 31.8 mllon tons n 1973 and they had declned to 15.8 mllon tons n 21, or by 5%. (US EPA, 23).

4 SOURCES OF SO 2 EMISSION REDUCTIONS 3 a) exstng facltes would be subject to emsson rate lmts mposed by State Implementaton Plans that were to ensure attanment of the Natonal Ambent Ar Qualty Standard for SO 2, and b) new plants would be subject to strngent New Source Performance Standards that would requre the adopton of best avalable control technology. These provsons had become effectve by the md-197s when natonal SO 2 emssons peaked and they have remaned n effect to ths day. The Clean Ar Act Amendments of 199. Ttle IV of these amendments created a natonwde lmt on aggregate SO 2 emssons of approxmately 9 mllon tons to be acheved n two phases by an nnovatve cap-and-trade program that ssued allowances n an amount equal to the cap and requred all electrc utlty generatng unts to surrender allowances equal to the unt s emssons. Snce no specfc command concernng abatement s gven at the unt level, the operators of affected unts are free to decde whether they wll reduce emssons by lowerng the sulfur content of the fuel used to generate electrcty (ether by swtchng or retrofttng scrubbers) or by shftng generaton to lower emttng unts ncludng new unts. However, Ttle IV dd not replace the source-specfc lmts and technology mandates of the earler 197 Amendments. The cap and the assocated oblgaton to surrender allowances equal to the tons of SO 2 emtted s an addtonal requrement mposed on top of the pre-exstng structure of prescrptve regulaton. 3 The reducton n electrc utlty SO 2 emssons has been the more remarkable n that the demand for fossl-fuel-fred generaton of electrcty has grown substantally snce 197 as shown n Fgure 1. 3 The super-mposng of Ttle IV on the pre-exstng prescrptve rate lmts, whch are amed prmarly at preventng adverse local health effects means that some plants are not free to ncrease emssons (and purchase allowances). In practce, these pre-exstng constrants have not posed a serous mpedment to tradng under Ttle IV snce the cap requres a sgnfcantly greater reducton of aggregate emssons than what s requred to meet the Natonal Ambent Ar Qualty Standard for SO 2. Whle generatng unts can trade only wthn the prescrptve lmts mposed by the 197 Amendments, these lmts have become non-bndng for most unts.

5 SOURCES OF SO 2 EMISSION REDUCTIONS 4 2 3, ,5 mllon short tons , 1,5 1, Terawatt-hours Electrc Utlty Generaton Fgure 1. U.S. Fossl-fuel-fred electrcty generaton and SO 2 emssons, In general, generaton of electrcty from fossl-fuel-fred power plants has ncreased steadly whle SO 2 emssons have regularly declned. Snce the year of peak emssons, 1977, fossl-fuel-fred generaton has ncreased at an average annual rate of 2.% whle SO 2 emssons from these sources have decreased at an annual rate of 2.4%. 4 The mpled annual rate of reducton n aggregate SO 2 ntensty for fossl-fuel-fred generaton s 4.3%, from 23 pounds of SO 2 per megawatt-hour n 1977 to 7.76 pounds n 22. In broad terms, ths reducton n aggregate ntensty results from two effects: the reducton n emsson ntensty or rates at ndvdual unts and the dsplacement of hgher emttng unts by exstng sources wth lower emsson rates or new sources wth mandated lower emsson rates. Whle the trend n SO 2 emssons snce the md-197s s nstructve, the past fve years offer an especally good opportunty to examne the effect on emssons of the ntroducton of low-emttng new generatng unts. Several factors the need for new capacty to meet contnually growng demand, the avalablty of more effcent, 4 Over ths same 25-year perod, total electrcty generaton, ncludng nuclear, hydro, and renewables, has ncreased at an annual rate of 2.4%.

6 SOURCES OF SO 2 EMISSION REDUCTIONS 5 combned-cycle generatng technology, and the expectaton of relatvely low natural gas prces coalesced n the late 199 s to create a boom n the constructon of new naturalgas-fred generatng capacty. Snce natural gas emts only trace amounts of SO 2, the deployment of these new unts could be expected to reduce SO 2 emssons consderably as pre-exstng, hgher emttng generatng unts are dsplaced n meetng the demand for electrcty. As of the end of 22, the new gas-fred capacty s estmated to be 133 GWe, an approxmately 2% ncrease n generatng capacty, and another 56 GWe s under constructon and expected to be completed n the next few years, mostly n 23 (EVA, 23). About half of ths capacty conssts of sngle-cycle combuston turbnes that are used mostly for meetng peak demand and offer few f any operatng effcences compared to exstng capacty. The remanng half of the new capacty utlzes combned cycle technology that offers marked operatng effcences that would be expected to lead to greater utlzaton for these unts and greater dsplacement of exstng unts. 5 Accordngly, we focus mostly on the combned-cycle unts. Our purpose n ths paper s to analyze the sources of the reducton n SO 2 emssons and, n partcular, to dstngush between the effects of lower emsson rates at exstng unts and the dsplacement of hgher emttng generatng unts by lower emttng ones, regardless of whether these are new unts or exstng unts wth lower emssons. In dong so, we gve partcular attenton to the reducton n SO 2 emssons attrbutable to the large ncrease n new natural-gas-fred capacty n the Unted States snce The methodology we employ n ths paper does not dscrmnate between emsson rate reductons and dsplacements that respond to polcy measures and those that would have occurred anyway because of other non-polcy-related factors affectng the electrc utlty generatng sector of the economy. Accordngly, the results we report should not be nterpreted as beng entrely due to regulatory measures, although a large fracton surely s. Where approprate, menton wll be made of the non-regulatory factors. 5 For nstance, n the thrd quarter of 22, combned cycle unts consttuted 52% of the new capacty and 79% of the generaton from the new gas-fred unts.

7 SOURCES OF SO 2 EMISSION REDUCTIONS 6 The next secton of the paper explans the data base and methodology that s used to dentfy the source of observed SO 2 emsson reductons. Results are then reported n the next secton, and a fnal secton concludes. A techncal explanaton of the decomposton methodology and the full data results are provded n appendces. DATA AND METHODOLOGY Adopton of the 199 Clean Ar Act Amendments, and specfcally the decson to allocate allowances to generatng unts accordng to average heat nput and the 1985 SO 2 emsson rate, requred the U.S. Envronmental Protecton Agency (US EPA) to develop a more detaled and accurate data base than had exsted prevously. Ths data base lsts annual SO 2 emssons and heat nput at the unt level for over 3 generatng unts from 1985 on. The avalablty of data at the unt level s partcularly mportant snce any gven power plant wll typcally consst of several generatng unts, usually three to four but sometmes as many as a dozen, usually bult n dfferent years and typcally subject to dfferng regulatory requrements. Absent unt-level data, t would be mpossble to tell whether an observed change n emssons at a power plant s due to changes n emsson rates at all or several unts or to the changng utlzaton of the consttuent unts wth dfferng emsson rates because of dfferent regulatory requrements. Our analyss s based on ths data base from whch some 1, rarely utlzed, old, and small unts are excluded. The remanng 1,89 unts account for 99% of total SO 2 emssons from the electrc utlty sector durng the years (US EPA, 23). 6 Gven ths concentraton of SO 2 emssons n two-thrds of the total generatng unts (and about 95% of total heat nput), any perceptble change n total SO 2 emssons from the 6 A unt s ncluded n the data base f t meets one of several crtera developed to determne unts that are sgnfcant n generatng electrcty. These crtera are: 1) more than 5 trllon Btu heat nput n any year from 1995 through 21, 2) more than 1 trllon Btu n any two years out of four consecutve years between 1995 and 21. A 1 MWe unt consumng 1 trllon Btu n a year wth a heat rate of 1, Btu/kwh would generate 1 GWh of electrcty n the year, or 1, hours (about 11% of the hours n a year) at full capacty.

8 SOURCES OF SO 2 EMISSION REDUCTIONS 7 electrc utlty sector as a whole wll be determned almost entrely by changes at these 1,89 sgnfcant unts. Snce annual SO 2 emssons are the product of heat nput, measured n mllon Btus (mmbtu) and the emsson rate measured n pounds of SO 2 per mmbtu (#/mmbtu), changes n observed emssons from one year to the next at any gven unt can be decomposed nto two components: a change n the annual emsson rate, whch would reflect the use of a hgher or lower SO 2 -emttng fuel or the nstallaton of emsson control equpment, and 2) a change n annual heat nput at the unt, whch may reflect a change n aggregate demand for electrcty or the effect of dsplacement of one unt by another n meetng any gven level of demand. In nearly all cases, both effects operate, often n off-settng drectons; however, the relatve contrbutons of each can be dentfed usng analytc technques explaned brefly below and more fully n the appendx. Whle the causes of changes n emssons at any ndvdual unt can be decomposed nto two effects, changes n observed emssons from any aggregate of generatng unts must take account of the nteracton of all the unts n the aggregate. For nstance, f one unt s utlzed less, as measured by heat nput or generaton, and the utlzaton of another unt s ncreased by the same amount, the effect on total emssons depends on the emsson rates at the two unts. If the emsson rate s lower at the unt ncreasng utlzaton than at the other unt, total emssons wll decrease wthout any change n the emsson rates at the two unts. Thus, for any aggregate, changes n total emssons can be broken down nto three components emsson rate reductons at ndvdual unts, changes n aggregate demand, and changes n the utlzaton of unts wth dfferng emsson rates as represented n the followng equaton. (1) de = der + deh agg + deh Dsp where der = the sum of the changes n emssons due to changes n emsson rates at ndvdual unts, deh agg = the change of emssons that can be attrbuted to changes n aggregate demand wthout any change n emsson rates at ndvdual unts,

9 SOURCES OF SO 2 EMISSION REDUCTIONS 8 deh Dsp = the change of emssons that can be attrbuted to the dsplacement of some unts by others n meetng aggregate demand. The left-hand-sde varable of equaton (1) s observed and the frst two rght-hand-sde terms can be easly calculated. The term der s the sum of the change n emssons due to changes n the emsson rate at all consttuent unts and the term deh agg can be found by multplyng the pror year s emssons by the percentage change n heat nput for the aggregate. Any dfference between the sum of der and deh agg and the observed change n emssons, de, s due, by defnton, to dsplacement, or the emsson effects of the changng shares n heat nput of the unts composng the aggregate. The avalablty of data ndcatng whether the fuel burned n a generatng unt s coal or ol/gas allows us to decompose the dsplacement effect nto a shft between fuels and dsplacements among the unts composng each fuel aggregate, as follows: (2) deh Dsp = deh bet + deh w/,coal + deh w/,ol/gas where deh bet = the change of emssons that can be attrbuted to changng shares of generaton between coal and ol/gas unts, deh w/ Coal = the change n emssons due to a redstrbuton of heat nput among unts usng coal, and deh w/ Ol/Gas = the change n emssons due to a redstrbuton of heat nput among unts usng ol or natural gas. One easy way to vsualze ths decomposton s to recall that the change n emssons due to changng heat nput at any ndvdual unt results from the change n aggregate demand for generaton, any change n fuel shares, and ndvdual dsplacements wthn the two fuel categores. Imagne a stuaton n whch there s no change n the emsson rates at ndvdual unts so that all changes n emssons are due to these three demand effects. If a coal-fred unt has ncreased emssons by 3% whle aggregate demand has ncreased 1% and the demand for aggregate coal-fred generaton has ncreased by 1%, one percentage pont of the observed three-percent ncrease n emssons at ths ndvdual unt can be attrbuted to each of the three effects: deh agg, deh bet, and deh w/_coal. If

10 SOURCES OF SO 2 EMISSION REDUCTIONS 9 observed emssons had not ncreased at all at ths unt whle the other condtons appled, then t could be sad that ths unt experenced a 2% reducton n utlzaton due to dsplacement by other coal-fred unts. Once these dfferences are calculated for all unts consttutng some aggregate, they can then be summed to determne all of the components n equatons (1) and (2). Choosng the approprate level of aggregaton for determnng growth n aggregate demand and changes n fuel shares n the Unted States s not obvous. Fuel shares dffer markedly by regon, as do the growth rates n the generaton of fossl-fuelfred electrcty. Usng the natonal aggregate would not provde an accurate estmate snce t would assume that generatng unts are part of one large ntegrated natonal market, whch they are not. At the other extreme, a state-level aggregaton would be smlarly msleadng snce electrcty control areas often encompass several states and electrcty flows frequently cross state boundares even when control areas follow state lnes. As a mddle ground we have used the nne census regons, the composton of whch s gven n Table 1 below and for whch regonal aggregate data s gven n Table A1 of the appendx. Accordngly, we calculate deh agg and the components of deh Dsp on a regonal bass and then sum across the nne census regons to obtan natonal fgures.

11 SOURCES OF SO 2 EMISSION REDUCTIONS 1 Regon New England Mddle Atlantc East North Central West North Central South Atlantc East South Central West South Central Mountan Pacfc States CT, MA, ME, NH, RI, VT NJ, NY, PA IL, IN, MI, OH, WI IA, KS, MN, MO, ND, NE, SD DC, DE, FL, GA, MD, NC, SC, VA, WV AL, KY, MS, TN AR, LA, OK, TX AZ, CO, ID, MT, NM, NV, UT, WY CA, OR, WA Table 1. U.S. census regons and consttuent states The greater effcency of the new combned cycle unts presents a problem n estmatng the SO 2 emsson reductons attrbutable to ths new capacty. The heat nput used by these new unts s fully ncorporated nto the components of equatons (1) and (2), but these unts generate more electrcty per unt of nput than conventonal unts. Snce electrcty s the fnal output, some accountng must be made for ths addtonal dsplacement and emsson reducton, whch shows up otherwse erroneously as a reducton n aggregate demand. Ths adjustment s made through a three-step process as explaned n more detal n the appendx on methodology. Frst, the heat nput savngs attrbutable to the use of combned cycle generatng plants s determned. We observe an average heat rate (Btus per kwh) of 7,4 for the combned cycle unts and we assume an average heat rate of 1, Btus/kwh for the generaton beng dsplaced. These fgures mply that the heat nput dsplaced by these new combned cycle unts s 35% greater than the heat nput use observed at these new unts. The second step s to determne whether the ncreased generaton s dsplacng coal-fred or ol/gas fred generaton, whch we do on a regonal bass. The last step s to calculate the emsson reducton by multplyng the dsplaced heat nput for each type of generaton by the respectve average regonal emsson rates.

12 SOURCES OF SO 2 EMISSION REDUCTIONS 11 DECOMPOSITION RESULTS NATIONAL RESULTS FOR THE PERIOD Fgure 2 below and Table A2 of the appendx show the natonal change n SO 2 emssons n tons by year for the perod and by the three components of equaton (1), that s, changes n the emsson rate, changes n aggregate demand, and changes n dspatch among unts from one year to the next. 2,, 1,, tons SO2 (1,,) (2,,) (3,,) (4,,) der deh agg deh dsp Fgure 2: Natonal change n SO 2 emssons n tons by factor, from 1985 to 21. The most salent feature of Fgure 2 s the very large reducton n SO 2 emssons n 1995, the frst year of Phase I of the Acd Ran Program. Ths reducton s especally remarkable n that 1) the cap appled only to a sub-set of unts n that year (albet the largest and most hghly emttng unts), 2) these unts reduced emssons far more than was requred to meet the cap n that year (or for any year of Phase I), and 3) the much larger set of generatng unts that dd not become subject to the cap untl 2 ncreased emssons by some 439, tons n 1995 compared wth The second largest annual reducton s n 2, when all of the other generatng unts were frst subject to the Ttle IV cap and therefore requred to pay the gong prce of

13 SOURCES OF SO 2 EMISSION REDUCTIONS 12 allowances (about $15/short ton n ths year) for all SO 2 emssons. The reducton n 2 occurred despte the large accumulaton of banked allowances from the Phase I unts (11.6 mllon tons) that would have easly covered the abated emssons n 2, had the owners been wllng to pay the prce of an allowance. That they dd not do so suggests that the cost of reducng emssons at these unts was less than $15/short ton. The broader pont that emerges from the emsson reductons observed n these two years s that, when a prce must be pad for otherwse permtted emssons, further reductons of emssons can be acheved. Settng these two years asde, SO 2 emssons typcally declned each year (11 out of 15), but by much smaller amounts than were observed n 1995 and 2. Table 2 summarzes the emsson reductons shown on Fgure 2 by component and perod, preand post-ttle IV. ( tons SO 2 ) der - 2,343-6,1-8,345 deh agg + 2,9 + 2, ,757 deh dsp - 1,263-1,43-2,36 de - 1,598-4,296-5, 894 Table 2. Emsson reductons by component and perod, As shown n the lower, rght-hand cell, 22 SO 2 emssons from electrc utlty generatng unts had fallen by 5.9 mllon tons from ther level n 1985, and they wll fall another mllon tons n order to meet the Phase 2 cap as the Phase I bank of allowances s drawn down. The decomposton of ths change shows that emssons would have ncreased by 4.7 mllon tons over ths perod as a result of ncreasng generaton from fossl-fuel-fred generatng unts 7, but ths effect s more than offset by the combned effect of reductons n emsson rates at exstng unts and the general dsplacement of generaton to lower emttng unts. Of these two emsson-reducng effects, by far the 7 Ths effect would be larger f t were calculated from some unchangng base year emsson rate, such as n 1985, nstead of from each succeedng year, whch reduces the effect of ncreasng demand n each year by the emsson rate reducton and dsplacement effects n pror years.

14 SOURCES OF SO 2 EMISSION REDUCTIONS 13 greater s the effect of emsson rate reductons. Ths effect s also notably larger after 1995 than before. Moreover, not all of the emsson reductons observed over the pre-ttle IV perod can be attrbuted to ar emsson regulatons. Ellerman and Montero (1998) estmate that the effect of ralroad deregulaton n makng low-sulfur western coals economcally compettve at Mdwestern generatng unts burnng local, hgh-sulfur coals reduced SO 2 emssons by about two mllon tons between 1985 and Ths reducton occurred by swtchng unts burnng hgh sulfur md-western coal partally or entrely to lower sulfur western coal and by the greater utlzaton of these unts. Applyng ther estmate to ths analyss suggests that about half of the 3.6 mllon ton reducton n SO 2 emssons resultng from emsson rate reductons and dsplacement from 1985 through 1994 was due to reasons other than ar emsson regulaton. 8 Accordngly, the contrast n the magntude of the emsson reductons assocated wth conventonal prescrptve regulaton and the cap-and-trade requrements nsttuted by Ttle IV s even greater than s suggested by the cumulatve amounts n Table 2. The dsplacement component n emsson reductons observed snce 1985 can be further decomposed to reflect the emsson effects of shfts n the relatve shares of coal and ol/gas and of greater or less use of lower emttng unts wthn each of these fuel types, as shown below by year n Fgure 3 and Table A2 of the appendx and cumulatvely n Table 3. 8 Keohane (23) shows that the reductons n the delvered prce of low-sulfur western coal n the Mdwest came to an end n the early 199s so that the one-year dfference n termnal years between the Ellerman-Montero analyss and the analyss n ths paper s not lkely to be great.

15 SOURCES OF SO 2 EMISSION REDUCTIONS 14 3, 2, 1, tons SO2 (1,) (2,) (3,) (4,) (5,) deh bet deh w/_coal deh w/_og Fgure 3. Decomposton of the dsplacement effect by year, ( tons SO 2 ) deh bet deh w/_coal - 1, ,7 deh w/_og deh dsp - 1,263-1,43-2,36 Table 3. Cumulatve decomposton of the dsplacement effect By far, the largest component of the 2.3 mllon ton reducton due to dsplacement of generaton among fossl-fuel-fred generatng unts over the perod s that due to dsplacement among coal-fred unts. Ths s not surprsng snce the potental for reducton s large gven the range of sulfur content among coals, from as low as.5 lbs. SO 2 /mmbtu to more than 5 lbs./mmbtu. Most of ths reducton occurred n the years before Ttle IV became effectve and t s largely due to the shft to low-sulfur western coal dentfed by Ellerman and Montero (1998). 9 Once Ttle IV became effectve, the three components of the dsplacement effect are more balanced and the largest dsplacement component s a shft to more ol/gas fred generaton. Ths shft s consstent wth the abnormally low ol prces experenced n 1998 and the nstallaton of over 15 9 Snce unts are dspatched on the bass of varable costs, whch are largely fuel costs, unts swtchng to lower cost, lower sulfur western coal would tend to be dspatched more.

16 SOURCES OF SO 2 EMISSION REDUCTIONS 15 GWe of new natural gas fred generatng capacty ncludng nearly 1 GWe of new combned cycle capacty to whch we now turn. THE EFFECT OF COMBINED CYCLES ON SO 2 EMISSIONS The crtcal ssue n estmatng the reducton n SO 2 emsson due to the new combned cycle capacty s determnng what generaton s dsplaced. Ths queston cannot be answered satsfactorly wthout a dsaggregaton to at least the regonal level because of the sgnfcant dfferences n the regonal dstrbuton of the new combned cycle capacty, dfferng patterns of dsplacement by regon, and dfferent regonal emsson rates for coal and ol/gas fred generaton. The regonal dstrbuton of the new combned cycle capacty s gven n Table 4 and addtonal data used n calculatng the effect of the new combned cycle capacty on SO 2 emssons s provded n Table A3 of the appendx. Census Regon CC Capacty 22 (MW e ) Regonal Share of CC Capacty 22 Regonal Share of US Ol/Gas Generaton, 1997 Regonal Share of Total Fossl Generaton, 1997 New England 6,19 11% 1% 2% Md-Atlantc 3,248 6% 13% 8% East North 3,827 7% 2% 2% West South 918 2% 1% 1% South Atlantc 6,489 11% 19% 2% East South 5,537 1% 2% 11% West South 22,448 39% 4% 16% Mountan 4,318 8% 2% 1% Pacfc 4,395 8% 11% 2% USA (lower 48) 57,289 1% 1% 1% Table 4. New combned cycle capacty and regonal shares of generaton The regonal dstrbuton of combned cycle capacty follows the pre-exstng dstrbuton of ol and gas generaton far more closely than t does the pre-exstng generaton of fossl-fuel fred generaton. Fve regons consttutng 93% of ol and gas generaton n 1997 account for 75% of the combned cycle capacty but only 48% of total fossl

17 SOURCES OF SO 2 EMISSION REDUCTIONS 16 generaton. The largest share by far of new combned-cycle capacty s n the West South Central census regon, encompassng Texas, Oklahoma, Arkansas, and Lousana, whch s also the regon wth the largest share (and absolute amount) of ol and gas generaton. Conversely, regons n whch there was lttle pre-exstng ol and gas generaton receved a smaller share of the new combned cycle capacty. A sold economc reason explans ths pattern. When new, more effcent unts compete wth exstng unts usng the same fuel, they can be assured of beng dspatched frst f all other factors are equal. However, when the competng unts use a dfferent fuel, dsplacement depends upon the prce dfference between natural gas and the other fuel. If the prce of the fuel frng the more effcent generaton s greater percentage-wse than percent savngs n heat nput, dsplacement wll not occur. Ths has been the case for the new combned cycle unts when they compete aganst exstng coal-fred unts n the U.S., especally snce late 22 when natural gas prces rose to levels that are two to three tmes the level of coal prces. There are, of course, other factors concernng locaton and network dynamcs that nfluence dspatch, but buldng combned cycle unts where relance on less effcent natural gas generaton s already hgh provdes greater assurance of demand for generaton from the new capacty, but also less reducton of emssons. Two dstnct patterns of dsplacement occur, as llustrated by the two charts n Fgure 4.

18 SOURCES OF SO 2 EMISSION REDUCTIONS 17 West South Central (TX, OK, AR, LA) East South Central (KY, TN, AL, MS) 5% 12% 45% Ol/Gas Share 4% 1% Share of total heat nput 35% 3% 25% 2% 15% 1% Combned Cycle Share Share of total heat np ut 8% 6% 4% 2% Ol/Gas Share Combned Cycle Share 5% % % Fgure 4. Combned cycle and ol/gas shares n two census regons The uppermost lne on each chart represents the share of heat nput nto ol/gas generatng unts n that regon, whle the bottom lne shows the share of heat nput gong nto combned cycle unts. In all regons, the share of combned cycle capacty rses from nearly zero n 1998 to some notceable postve share by 22. In cases such as the West South Central census regon, the share of combned cycle heat nput rose from 1% n 1999 to 19% n 22. Over the same perod, the total ol and gas share of heat nput remaned relatvely constant at 42%-44%. Obvously, the new combned cycle capacty n ths regon has been dsplacng exstng ol and gas capacty, not coal capacty. The East South Central regon presents a dfferent pcture. The 22 shares of ol/gas and combned cycle heat nput are much smaller than n the West South Central regon, but the ncrease n the combned cycle share from zero percent n 1999 to 8% n 22 causes the ol/gas share of heat nput to ncrease by fve percentage ponts, from 5% n 1999 to 1% n 22. Accordngly, t can be sad that fve percentage ponts of the 8% ncrease n combned cycle generaton dsplaced coal generaton and the remanng three percentage ponts dsplaced exstng ol/gas generaton, whch s now 2% nstead of 5%.

19 SOURCES OF SO 2 EMISSION REDUCTIONS 18 When dsplacement s calculated n ths manner for all nne regons for each year, the amount of dsplacement depends not only on the amount of heat nput dsplaced by the new combned cycle unts, but also on the emsson rate of the Btu s beng dsplaced from coal or other ol and gas-fred unts. Fgure 5 and Table 5 below provde the yearby-year results for the nne census regons and the naton as a whole. NEW MAT ENC WNC SAT ESC WSC MON PAC (2,) (4,) (6,) tons SO2 (8,) (1,) (12,) (14,) (16,) Census Regon Fgure 5. SO 2 emsson reductons due to new combned cycle capacty, by regon and year tons SO Cumulatve New England Md Atlantc East North Central West North Central South Atlantc East South Central West South Central Mountan Pacfc Lower 48 States Table 5. SO 2 emsson reductons due to new combned cycle unts, by year and regon

20 SOURCES OF SO 2 EMISSION REDUCTIONS 19 The two regons wth the largest cumulatve reducton (the sum of the annual amounts) are the East South Central and South Atlantc census regons. Even though they consttute only a quarter of natonal combned cycle generaton, they account for 53% of the natonal SO 2 reducton attrbutable to the new combned cycle capacty. The reason s that the new combned cycle capacty n these regons dsplaced more coal generaton and the emsson rate assocated wth the dsplaced coal generaton s relatvely hgh. In contrast, the much larger dsplacement of exstng generaton n the West South Central regon reduced SO 2 emssons by consderably less because no coal generaton was dsplaced. A fnal observaton about the effect of the new combned cycle capacty on SO 2 emssons concerns the nteracton between these new unts and the Ttle IV cap. Whle the new combned cycle capacty clearly dsplaced generaton that had hgher SO 2 emssons, aggregate SO 2 emssons are no lower than they would otherwse be snce the SO 2 emssons cap s fxed. 1 The effect of the new capacty s then to reduce the amount of abatement requred from the other, mostly coal-fred unts. Consequently, the effect of the new combned cycle capacty s not to reduce actual SO 2 emssons but the emsson reducton requred of other generators of electrcty and therefore the cost of achevng the SO 2 cap. The extent to whch the cost of Ttle IV has been reduced can be estmated. As shown n Table 5, the cumulatve reducton n SO 2 emssons attrbutable to the combned cycle unts as of the end of 22 s approxmately 8, tons. The method for calculatng the smple counterfactual for 22 (cf. Ellerman et al., 2) yelds counterfactual emssons that are 6.9 mllon tons greater than observed emssons of 1.2 mllon tons; however, ths method does not take account of the assumed 35% effcency gan and greater dsplacement per unt of heat nput assocated wth the combned cycle unts. When ths correcton s made, counterfactual emssons are 7.1 mllon tons hgher than observed emssons. After subtractng the 8, ton emsson reducton due to the 1 Ths effect does not apply for any uncapped emssons, such as NO x emssons n many states and CO 2 emssons.

21 SOURCES OF SO 2 EMISSION REDUCTIONS 2 new combned cycle unts, the remanng unts reduced SO 2 emssons by only 6.3 mllon tons or about 11% less than what would have been requred to meet the same electrcty demand wthout the new combned cycle unts.. Assumng a lnear relaton between quantty and prce for ncremental abatement at the current margn, the margnal cost of abatement and the prce of allowances s 11% less than t would be absent the ntroducton of the new combned cycle capacty. 11 The average prce of allowances n 22 was about $15, whch would mply margnal costs that would have been $16-$17 hgher. Addtonal combned cycle capacty came on lne n 23, approxmately equal n capacty to that added n 22, so that the ultmate effect mght be larger, but ths would depend upon the amount of dsplacement by ths new capacty and the data reported so far for 23 ndcates decreasng total ol/gas generaton over the past year, probably because of the hgh natural gas prces that have been observed snce the end of 22. If a round number were to be used for the total effect of the new combned cycle capacty n reducng the margnal cost of abatement, say $2 per ton, the mpled annual savngs n electrcty cost s $18 mllon when multpled by the Phase 2 cap of 9 mllon tons of SO 2 emssons per year. CONCLUSION The major source of SO 2 emsson reductons n the Unted States snce 1985 has been the reducton of emsson rates at exstng unts. Dsplacement of hgher emttng unts by lower emttng ones, whether newly constructed or exstng unts, has also contrbuted an mportant share of the total reducton; however, ths factor alone has not been suffcent to offset the ncrease n emssons that would have occurred as a result of contnung growth n aggregate demand. Our analyss also ndcates that Ttle IV has been more effectve n reducng emssons durng the eght years t has been n effect than the conventonal, source-specfc, prescrptve regulaton had been n reducng emssons n the ten years precedng Ths s not to say that allowance prces have fallen as the new combned cycle capacty came on lne snce ts effect of the allowance market would have been antcpated.

22 SOURCES OF SO 2 EMISSION REDUCTIONS 21 The effect of the new combned cycle capacty s not what mght be expected at frst sght. These unts have clearly dsplaced more hghly emttng generatng unts, although most often not coal-fred unts, but the effect has been to reduce the cost of abatement, not total SO 2 emssons. When emssons are capped, exogenous factors such as the ntroducton of more effcent combned cycle generaton results n less requred abatement by other affected unts, n ths nstance, mostly coal-fred unts. From the standpont of the competton among contendng fuels, ths effect s ronc but t s small and the ultmate benefcary s the consumer who thereby pays slghtly less for electrcty wthout any change n ths attrbute of envronmental qualty.

23 SOURCES OF SO 2 EMISSION REDUCTIONS 22 REFERENCES Ellerman, A. Denny, Paul L. Joskow, Rchard Schmalensee, Juan-Pablo Montero, and Elzabeth Baley (Ellerman et al., 2). Markets for Clean Ar: The U.S. Acd Ran Program. Cambrdge Unversty Press, 2. Ellerman, A. Denny and Juan-Pablo Montero The Declnng Trend n Sulfur Doxde Emssons: Implcatons for Allowance Prces, Journal of Envronmental Economcs and Management 36: (Ths artcle s substantvely reproduced as chapter 4 of Ellerman et al. (2). Energy Ventures Analyss, Inc. 23. Trackng the Boom of New Power Plants n the U.S. (Propretary quarterly report dated September 23). Arlngton, VA. Keohane, Nathanel O. and Meghan Busse, 23. Polluton control and nput markets: The creaton and capture of rents from sulfur doxde regulaton. Workng paper dated July 28, 23, avalable at U. S. Envronmental Protecton Agency, 23. Average annual emssons, all crtera pollutants, years ncludng , Washngton, D.C. Avalable at

24 SOURCES OF SO 2 EMISSION REDUCTIONS 23 APPENDIX I: DECOMPOSITION METHODOLOGY Decomposton of changes n emssons at the unt level The SO 2 emssons produced by a generatng unt n the tth year can be descrbed as: e = r * h t t t where h t s the heat nput (.e. the energy contaned n the fuel burnt durng year t) and r t s the emsson rate (.e. the amount of SO 2 emtted per unt of heat nput). The change n SO 2 emsson between year and t can be descrbed as a functon of four observed values, h, h t, r, and r t, such that de ( r + dr, t )( h + dh, t ) r, t = rt ht r h = h de, t = r dh, t + hdr, t + dr, tdh, t where the d,t s denote the observed change n e, r or h between year and t. The change n emssons, de,t, can also be represented n a (h,r) dagram: r t r dr,t h dr,t dh,t dh,t r h h t Fgure 6. Representaton of the heat nput, emsson rate and emssons of a generatng unt n a (h,r) dagram In ths dagram the surface of the h x r rectangle s equal to the emssons e, and the surface of the h t x r t rectangle s equal to the emssons e t. The dfference e t e s represented by the strped areas. The dagram clearly shows that de,t can be separated nto three components: 1 = (r t - r )h whch s created by a change of the unt s emsson rate

25 SOURCES OF SO 2 EMISSION REDUCTIONS 24 2 = (h t - h )r whch s created by a change of the unt s heat nput 3 = (r t - r )(h - h t ) whch s created by both changes We adopt the conventon of splttng the thrd component evenly and attrbutng each half to the other two components so that we can attrbute /2 to a change of emsson rate and /2 to a change of heat nput, whch gves us: - the change n emssons due to a change n heat nput dr de = r dh + h, t, t dh 2, t - the change n emssons due to a change n the emsson rate dr de = dr h + r, t, t dh 2, t When a unt s ether shut down or put onlne (.e. ether h o or h t s equal to zero), we set de r = and attrbute all the change n emssons to a change n heat nput. Accountng for the nteracton of ndvdual unts wth others n some aggregate The two components accountng for changes n emssons at the unt level, de r and de h, have dfferng characterstcs when the unt s consdered as part of some aggregate, such as an electrcty grd. 12 A change n emsson rate, de r, such as that resultng from the nstallaton of a scrubber, s a unt specfc acton that does not mply a change n the emsson rate at other unts n the aggregate. In contrast, a change n heat nput at an ndvdual unt, de h, wll always reflect some change n the aggregate that s shared by other unts or s the result of the nteracton among the consttuent unts. For nstance, a change n aggregate demand would be expected to affect all unts n some measure. Smlarly, changes n fuel prces or n condtons on the electrcty network, would be expected to change the contrbuton of consttuent unts to meetng aggregate demand. Whle the observed change n heat nput at any sngle unt results from changes n 12 We use regonal aggregates defned along the lnes of the U.S. census regons, but the methodology apples for any aggregate.

26 SOURCES OF SO 2 EMISSION REDUCTIONS 25 condtons affectng the aggregate, the contrbutng factors can be analytcally separated nto three components: the change n aggregate demand, any change n the contrbuton of coal and ol/gas unts vewed as sub-sets, and any change n the utlzaton of ndvdual unts wthn each fuel share. More formally, heat nput at the th unt can be decomposed nto three components. h t h = dh agg + dh bet +d h w/. where dh agg s the th unt s share of the change of heat nput for the aggregate: dh agg h = H t H ( H ) wth H, H t beng the aggregate heat nput for years, t dh bet reflects what would be the change n the th unt s heat nput due to a change n the share of the subset of unts consttutng fuel X n year t assumng no change n the shares of the consttuent unts n that fuel subset: dh bet h = H, fuelx t ( ) = H t, fuelx t H t, fuelx H, fuelx h * H H H, fuelx wth H,fuelX, H t,fuelx the aggregate heat nput for fuel X unts for years, t. H H dh w/ s the remanng part of h t h whch wll be equal after substtuton and cancellaton to: dh w / = ht h H H t, fuelx, fuelx dh w/ reflects the effect of any change n the role of the th unt wthn the fuel X subset after allowng for changes n aggregate demand and for any change n fuel share. Wth these defntons, the new decomposton of de,t n the (h,r) dagram for a unt experencng an ncrease n heat nput due to all three factors s: r t dr,t h r r h agg r h bet dr,t h agg dr,t h bet

27 SOURCES OF SO 2 EMISSION REDUCTIONS 26 Fgure 7. Representaton of the decomposton of the heat nput and assocated emssons of a generatng unt n a (h,r) dagram Returnng now to the formulae for changes n emssons, we can decompose de,t nto four components: de r = dr, dh t h + 2, t de h, agg = r dr + 2, t dh agg dr = r + 2, t h H ( H H ) t de h, bet = r dr, t dr, t + = h H dhbet r + Ht, fuel H, fuel 2 2 H, fuel H t dr, t dr, t Ht, fuel = = de h, w / r + dhw / r + h t h 2 2 H, fuel All the above equatons are unt-level equatons. Aggregate numbers can be obtaned by addng up the de x of all the unts of the database: de =. X de X database Specal consderatons for combned cycle unts Combned cycle unts present a problem n accountng for ther effect on emssons because they are markedly more effcent n generatng electrcty than conventonal coal and ol/gas generatng unts. The decomposton methodology presented above s based upon heat nput, not electrcty, whch s the fnal product. So long as the heat rate, the

28 SOURCES OF SO 2 EMISSION REDUCTIONS 27 number of Btu s used to produce a klowatt-hour of electrcty remans relatvely constant from year to year and among unts that would substtute for one another on the electrcty grd, no great dstorton results from usng heat nput as the proxy for electrcty output. However, wth the recent ntroducton of a sgnfcant amount of combned cycle capacty, ths assumpton no longer holds and some allowance must be made to recognze that the emsson reducton resultng from the dsplacement of conventonal generaton by new combned cycle unts s greater than would be ndcated by a smlar dsplacement among conventonal unts. More formally, so long as combned cycles dd not play a large role n generaton, such as was the case untl 1999, t was reasonable to assume that deh agg deh elec where the left-hand-sde of the equaton s defned as the change n emssons due to the observed change n heat nput and the rght-hand-sde, as the change due to the assumed change n demand for electrcty from fossl-fuel-fred generatng unts. Wth the ntroducton of a sgnfcant amount of combned cycle generaton, a new term s requred, deh cc, defned as the change n emssons due to the unobserved heat nput savngs resultng from the zero-fuel (thus zero-emsson) electrcty generaton by the heat recovery unt of combned cycle faclty. Conceptually, ths new term can be defned n the followng manner: deh cc deh agg deh elec If the share of combned cycle generaton n the aggregate s ncreasng, then deh elec > deh agg and deh cc wll be negatve, and vce versa f the share of combned cycle generaton s decreasng. Estmatng deh cc requred two analytcal tasks to be performed. Frst, combned cycle unts were dentfed wthn the subset of ol/gas unts. Second, the heat nput savngs assocated wth combned cycles was estmated. At frst appearance, all of the nformaton requred to perform both tasks appeared to be n the quarterly reports whereby emssons are reported to the U.S. Envronmental Protecton Agency, hereafter

29 SOURCES OF SO 2 EMISSION REDUCTIONS 28 called the CEMS (for Contnuous Emssons Montorng System) database, whch record not only emssons but also heat nput and electrcty generaton at the unt level, as well as dentfers for the fuel burned and the type of unt. In fact, cross checks wth other sources revealed that labels dentfed as combned cycle n the CEMS database were not always such and that some not so labeled were combned cycle unts. Comparson wth data reported to the Energy Informaton Agency and data obtaned by web search and drect calls enabled us to dentfy 276 out of the 948 ol/gas unts that could be consdered combned cycle unts n that these Btu-usng generatng unts had an assocated heat recovery unt. A more serous problem was that the generaton reported for combned cycle unts n the CEMS database was often only the generaton from the gas turbne and not the addtonal power from the assocated heat recovery unt. For nstance n 21, out of the 276 combned cycles, 52 unts had an average heat rate above 1, Btu/kWh. From dscussons wth the owners of some of these unts, we found that the data reported to the EPA on the CEMS forms sometmes contans only the generaton for the turbne (whch s the Btu-usng and emttng unt) and not the generaton from the (non-emttng) recovery unt. Consequently, there s no relable method wthn the CEMS data to determne whch combned cycle unts had complete generaton data and whch were ncomplete. To remedy ths problem, we used another database from the Energy Informaton Admnstraton, EIA Form 96, whch reports electrcty generaton and heat nput for all unts for the year 21. We took all the unts from the EIA database that were also present n the 276 combned cycles lst of the CEMS data and selected a group of 41 unts that had been n operaton for more than two quarters (thereby avodng heatrate-dmnshng start-up problems) and showed steady generaton and generally hgh utlzaton. Ths subset of fully operatonal combned cycles experenced an average heat rate of 7,4 Btu/kWh. The heat nput savngs from combned cycle unts was then easly calculated usng an assumed average heat rate of 1, Btu/kwh for conventonal generatng unts. Dvdng

30 SOURCES OF SO 2 EMISSION REDUCTIONS 29 1, by 7,4 provdes the assumed heat nput savngs of 35% that we use for estmatng the emsson reductons from the conventonal generaton dsplaced by the new combned cycle unts. More formally, t s possble to calculate the total heat nput dsplaced by combned cycles as the sum of the observed heat nput of CCs and an estmate of the heat nput saved by the recovery unt: h cc,dsp = h cc,obs + h cc,sav for any CC unt I Snce the combned cycle unts are present n the database we use, the change n emssons assocated wth changes n h cc,obs are already ncluded n deh bet and deh w/,og. deh cc s an adjustment, requred to account for the emssons savngs due to the greater effcency of combned cycle unts, that depends on h cc,sav, whch s related n turn to h cc,obs as follows: h cc,sav =.35*h cc,obs. Fnally, the savngs for the Yth regon can be summed across unts as: H Y cc, sav = h cc, sav cc, regony Furthermore, f we assume that the dsplacement due to the heat nput savngs, H t,cc,sav, s proportonal n all respects to the dsplacement occasoned by the observed heat nput at combned cycle unts, H t,cc,obs, then the heat nput savngs can be smlarly broken down nto a component dsplacng coal generaton and another dsplacng other ol/gas unts. Thus: H t,cc,sav = H coal t,cc,sav + H O/G t,cc,sav = H t,cc,sav * ( % coal + % O/G ) The calculaton of % coal and % O/G can be llustrated takng New England between 1999 and 21 as an example. Fgure 8 represents the heat nput shares of coal unts, conventonal ol/gas unts, and combned cycle unts.

31 SOURCES OF SO 2 EMISSION REDUCTIONS 3 1% 8% 6% 4% Coal O/G sngle cycle O/G combned cycle 2% % Fgure 8. Shares of Heat Input from coal, ol/gas sngle cycles and ol/gas combned cycle unts n New England between 1999 and 21. Three patterns of heat nput dsplacement by combned cycles are possble. ) If the share of coal heat nput ether ncreases or stays equal from one year to the next, then there s no coal dsplacement and combned cycles exclusvely dsplaced ol/gas unts. We then smply calculate the emssons savngs by usng the average ol/gas emsson rate. Accordngly, % coal = and % O/G = 1. ) If the share of heat nput from coal decreases more than the share of combned cycles ncreases, then we assume that the combned cycle unts have dsplaced coal unts only (and that conventonal ol/gas sngle cycles have dsplaced coal as well). Thus % coal = 1 and % O/G =. ) Fnally, f there s a decrease n the share of heat nput from coal smaller than the ncrease n the share of combned cycles, then we assume that the combned cycle unts have dsplaced both coal and ol/gas unts as follows: % coal = - ( H ) ( t, coal H, coal Ht, cc, obs H, cc, obs) = % coal - 1. and % O/G Once % coal and % O/G have been calculated for each regon and each year between 1999 and 22, the correspondng emssons savngs can then be calculated usng the average

32 SOURCES OF SO 2 EMISSION REDUCTIONS 31 regonal (or other aggregate) emsson rates of coal r coal and ol/gas r O/G, whch are observed. Thus, E t,cc,sav = E coal t,cc,sav + E O/G t,cc,sav = H t,cc,sav * ( % coal *r coal + % O/G *r O/G ) Table 6 shows the values obtaned for deh cc, as well as the SO 2 savngs due to the dsplacement of generaton by the entre combned cycle unt, whch s related to the savngs by a factor of 1.35/.35 or about 4) SO 2 savngs due to CCs heat recovery unt SO 2 savngs due to CCs as a whole (deh cc ) NEW -1,51-4,974-16,397-17,73-4,52-19,185-63,246-68,283 MAT -3, ,96-12,547-2,353-2,438-62,83 ENC , ,4-28,213 WNC , ,744-2,532-5,79 SAT -3,343-3,228-1,989-38,397-12,893-12,452-42, ,12 ESC ,385-21,74-28,342-3,287-13,56-81,286-19,319 WSC , , ,738-3,212-13,511 MON -42-2,42-4,436-1,343-1,552-9,264-17,11-5,179 PAC -2, , USA -8,992-28,473-58, ,252-34,685-19, ,596-44,685 USA Cumul. -37,466-95,694-29, ,51-369,16-89,792 Table 6. SO 2 emssons savngs due to the electrcty generaton dsplaced by combned cycles between 1999 and 22 APPENDIX II: DATA TABLES Table A1. Heat nput, SO 2 emssons, and average emsson rates by regon New England Heat nput SO 2 emssons Emsson rate Mddle Atlantc Heat nput 1,818 1,962 2,31 1,923 1,843 1,737 1,657 1,627 SO 2 emssons 1,645 1,694 1,696 1,67 1,623 1,557 1,465 1,429 Emsson rate East North Central Heat nput 3,456 3,61 3,662 3,741 3,88 3,713 3,96 4,69 SO 2 emssons 5,435 5,149 5,232 5,167 5,91 4,784 4,673 4,686