IS INVESTMENT IN PEAK GENERATION ASSETS EFFICIENT IN A DEREGULATED ELECTRICITY SECTOR?

Transcription

1 IS INVESTMENT IN PEAK GENERATION ASSETS EFFICIENT IN A DEREGULATED ELECTRICITY SECTOR? Benoît Peluchon MATISSE Unversty Pars 1, Boulevard de l hosptal 75013, Pars, France Abstract Electrcty markets have known turbulent tmes the last two years, from Calforna crss to nordpool s present wnter. Very hgh prces, wth no evdent lnk to costs of producton, have been recorded n exceptonal demand peaks. Each tme a lack of generaton capacty has been put n evdence. Durng December 2001, ths has even lead to falures n Span. Ths paper seeks to explan ths by focusng on peak capacty nvestments. Peak generators have a crucal role durng peak demand tmes, snce they prevent any falure n the delvery of electrcty. Though, ther fnancal vablty s very random snce they produce electrcty only when demand s exceptonally hgh. Under nvestment n peak generators may thus be underlyng the peaks n prces. Thus a model derved from electrcty markets models (Green and Newberry 92, Borensten and Bushnell 99) s buld, ntroducng randomness and the nvestment decson n the pcture. It allows to understand why a frm may not be wllng to nvest as much as a publc monopoly and why market power may be more dffcult to measure than by computng mark up rates. 1. INTRODUCTION From Calforna s hot 2000 summer to Europe s hard 2003 summer, prce spkes are now a common occurrence on electrcty markets. Though hgh prces are a normal phenomenon durng demand peaks, those spkes have been regarded manly as an evdence of dysfunctonal markets. Frstly, they seemed to bear no lnk wth the costs of (peak) producton (1000 /MWh n France on the 11 th of august). But more mportant, the possblty of falures n the supply of electrcty rose sharply, as reserve margns were very low. Each tme a prce spke has been recorded, a very hgh demand was coupled wth low remanng capactes, even negatve ones when shortcuts had to be done (Calforna 2000, Span 2001, Italy 2003, France 200?). Stll, that there was too much producton capacty n Europe was the general opnon three years ago. The non storablty of electrcty s crucal here. To meet a gven demand, there must be at least the same level of supply. But, as electrcty consumpton s cyclcal through the year, the hgher levels of demand are suppled by generators wth low fxed costs and hgh proportonnal costs, peak unts. Under nvestment n peak generaton unts may thus be an explanaton for the recent crses, snce peak generators are used to balance supply and demand. They work as an nsurance aganst shortcuts for a gven electrc system. The need for peak generaton depends on the per-

2 cepton producers have about the dstrbuton of probablty of demand. It s subject to changes snce ths dstrbuton changes too wth consumpton habts (the development of ar-cooled devces for example). Because they produce only when demand s very hgh, ther proftablty s very random and may be problematc. Moreover, as many economsts have already observed, electrcty markets tend to functon badly when demand s exceptonally hgh (Joskow 2001). One of the reasons s that competton for peak demand s less sharp, as many producers are already usng all of ther capacty. Thus, producers that have some remanng capactes have more market power and are able to get hgher prces. Under nvestng may then reward hgher profts. Ths paper s an attempt to characterze and explan ths under nvestment n peak generaton capacty. Investment and randomness are ntroduced n a model very smlar to electrcty market models, such as those developed by Green and Newberry (1992), Green (1996) or Borensten and Bushnell (1999). These models have been used to descrbe the prce mechansm on wholesale electrcty markets and to nvestgate the possblty of market power on such markets. Ther conclusons are that wholesale prces are hgher than margnal costs and they have ponted stuatons where frms had market power (the Pool n 1992, Calforna n 1999). But market power s consdered exceptonal (Wolfram 1999) or can be attenuated by adequate regulaton (Bushnell, Borensten, Knttel 1999). Includng the nvestment process seems to pont a deeper and more structural flaw n electrcty markets. 2. INVESTMENT IN GENERATION CAPACITY 2.1 The problem of peak nvestment Investment n the generaton ndustry s a very complex phenomenon. Ths ndustry has been progressvely deregulated n Europe n the 90 s, but has largely thrved on the genraton capacty bult by the publc natonal monopoles that prevaled before. At the excepton of England, where snce the begnnng of the 90 s many gas fred unts have been nstalled, the ntroducton of market mechansms has been done n a context free of nvestment (at the excepton of renewable energy unts). Stll, the growth of consumpton coupled wth the dvestures of many generators n Europe s changng the pcture. The need for new nvestments s growng. We do not know how the frame of nsttuton put n place to regulate the new markets are able to deal wth ths new context. Investment seems to be an nterestng subject to nvestgate and has not been treated as such by the academc lterature. Peak capacty nvestment, especally, seems qute problematc. An nvestment n a base generaton plant s a decson that requres forecastng the base future prces. An nvestment n a peak generaton plant requres much more nformaton as the peak prces depend from the base prces as well as from the future nvestments n every knd of generaton capacty. The revenue generated by a peak plant s much more hazardous than a base plant, snce t produces electrcty only when every other plant produces to full capacty or cannot produce. In the same way that an opton sad to be out the money, peak plant has a value that may change drastcally wth any change n the way demand evolves. Take for example a generator whose proportonal cost puts at the hgher end of the load curve, and suppose that such a demand happens wth a mean of one n ten years. Ths means that, on mean, nvestng n such a unt s proftable, but wll not produce nne years out of ten. Hence our nterest n how frms may deal wth ths problem. 2.2 Whch level of nvestment s optmal? As already told, peak capacty has a role of nsurance for the whole electrc system. It means that there must be enough nstalled peak capacty to ensure a low shortcut probablty. But what The Hague - September 2003 AER/CPB/ECN

3 s the prce of that nsurance? And what would be an optmal level of capacty from the pont of vew of the regulator? A possble answer s to defne a publc cost of falure, that s the publc cost of not supplyng one kwh of energy. Tradtonal French economc calculus used ths cost to compute an optmal set of generators by mnmzng the total cost of supplyng demand. The probablty of a shortcut, whch depends on the capacty nstalled, has then an economc groundng. It s the probablty that gves an expected cost of shortcut equal to the cost of buldng one more unt of capacty. We may defne thus an optmal level of peak capacty. In the model, we make the comparson between the nvestment polcy of two regmes: a publc monopoly, whose objectve s to mnmze ts costs, and a deregulated ndustry, where frms maxmse ther profts. It allows us to get some deas about the strateges frms may mplement and how to judge these strateges. Ths paper shows that generaton frms tend to nvest less than a publc monopoly, n order to rase ther proft. 3. THE MODEL 3.1 General features In order to test ths ntuton about peak nvestment, a model of the market for peak energy has been bult. Followng Schenkman and Kreps (1983), we suppose the nvestment decson s taken before the demand s known, then the prce s fxed by the equalty between supply and demand. The model lmts tself to peak demand, that s the resdual demand of energy once all the pre peak generators produce wth all of ther avalable capacty. Ths mples that every change n pre peak capacty must be ncluded as a change n resdual demand, snce t lowers by the same level the length of peak demand. We suppose the tme to be reduced to a sngle pont, we thus set asde the possblty that power may be lackng. But ncludng that possblty n a model suppose to modelze the whole load curve. We focus nstead on the possblty of energy defct. Peak demand s assumed to follow a normal dstrbuton. We can thus wrte peak demand at tme t as the sum of a mean and a random shock of mean zero and varance 2 σ : D = + ε, t m t t There s no prce n ths functon. We don t assume that consumpton s unelastc to prce, even f elastcty s very weak. But we make the assumpton that the value of elastcty s unknown and that t manfests tself by an exogenous prce determnaton n the case of the duopoly, whle there s no elastcty n the case of the publc monopoly on behalf of tarfcaton. The random part s an oversmplfcaton of realty, but allows us to ntroduce randomness and stll derve general results. There are three dfferent costs: the cost of falure (not supplyng one unt of electrcty), the cost of fuel that we admt to be completely proportonal and the cost of nvestment (cost of buldng one unt of capacty). The set of varables s: Varable Cost of falure Cost of fuel Notaton C D C p Cost of nvestment I Capacty nstalled Q Under-capacty prce P

4 3.2 The publc monopoly The objectve of the monopoly s to mnmze ts expected total cost whch s composed of three parts : nvestment cost, cost of producton, cost of demand that has not been suppled. Its varaable of decson s the level of capacty t wshes to buld n order to meet demand. The frm s objectve can be wrtten as: mn E cost Q [ ] [ ] = mn E IQ + CP D1{ } + ( C PQ + C D ( D Q ) 1 D< Q { D Q } Q where 1 { D<Q } s a functon that yelds 1 f the condton{ D < Q } s true ; and 0 otherwse It s not possble to get an explct expresson of the optmal level of nvestment, but the mplct equaton that defnes t permts the comparson between the two regmes. Ths equaton s : ( < Q M ) = P( D < Q ) 1 P ε 1 = I ( ) C D C P It traces the lnk between the dfferent costs and the probablty of under-capacty. Second order condton s verfed. The normal dstrbuton of demand allows us to compute the mpact of a postve varaton of exogenous varables. A rse n the cost of falure mples a postve varaton of optmal capacty. A rse n the cost of fuel or the cost of nvestment have the opposte effect. A rse of the demand trend (M) mples a postve varaton of capacty so that the probablty of under-capacty remans constant. 3.3 The prvate duopoly We represent the regme of a deregulated ndustry by a symmetrc duopoly. Each frm n the duopoly maxmzes ts proft. The prce mechansm n the duopoly s very smple. If demand s lower than the nstalled capacty, then competton between frms yelds a Bertrand outcome, and prce s equal to proportonnal cost. If demand s greater than the nstalled capacty, then each frm holds a knd a market power. If one of the frm produces wth all of ts capacty, then the other frm s n the stuaton of monopoly on the remanng demand. We suppose then that a knd of elastcty takes place, so that all excess demand s cleaned by an exogenous prce greater than the propotonnal cost that we wll note as under-capacty prce. Ths last prce s of course lower than the cost of falure. Ths mechansm s very near from Kreps and Schenkman (1983) model where capacty constrants n a two perod game wth prce competton yelds a Cournot equlbrum. But ther model does not exhbt pure strategy equlbrums n case of over-capacty, t can not be appled to the electrcty ndustry. We suppose frms are rsk neutral. Ths s not a realstc assumpton, but supposng they are rsk averse would only strengthen the results. The frm s expectaton of proft can then be wrtten as: max E Q [ ] [ π ] = max E ( CPQ C PQ IQ )1{ } + ( PQ C PQ IQ ) 1 D< Q + Q { D Q + Q } Q j j Equlbrum s defned by the ntersecton of the functon of reacton of each frm. The followng equaton gves the functon of reacton of frm : The Hague - September 2003 AER/CPB/ECN

5 A 1 Q A = Q I ( P ) C P where A s the probablty of over-capacty ( P ( D + ) < Q Q j ) Because equlbrum s symmetrc, the equaton also defnes the optmal level of nvestment of each frm. Equlbrum s unque and second order condton s verfed on a reasonable range of values : total (peak) capacty must be nferor to a value defned by the parameters of demand. Ths value s very hgh and smply stresses the fact that f capacty s too great then fxed costs can not be recovered. The optmal capacty of a frm bears the same relatonshps wth exogenous factors than the optmal capacty of the duopoly, the exogenous prce assumng the role of the cost of falure. 3.4 Comparson We are now able to compare the optmal level of nvestment n each regme. The two equlbrums are defned by : A I 1 Q A =, for the duopoly Q ( P ) C P I 1 A =, for the monopoly ( C D C P ) The two expressons are very smlar, the only dfference beng the dervatve of the probablty. It s possble to show that optmal level of capacty defned by the monopoly equaton s greater than the total capacty of the duopoly (two tmes the optmal capacty of a frm). Ths s true even f the over-capacty prce equals the cost of falure, but on the range of values on whch the duopoly equlbrum s unque. 4. RESULTS 4.1 The mechansm of under-nvestment The model tells us that the level of nvestment of the duopoly wll always be lower than the level of the monopoly. Even f the prce n case of demand n excess s equal to the cost of falure, the duopoly nvests less. The fundamental dfference between the duopoly and the publc monopoly s that what remans a cost for the monopoly, energy not suppled, rases prces and profts for the duopoly. If a frm bulds less capacty t lowers the quantty of electrcty t may sell, but t rases the prce of ths same electrcty. Of course competton between frms lmts ths effect, but not enough to compensate for t. Would the result be dfferent f there was more frms? It does not seem : f we suppose more frms compete for peak demand, the under-capacty prce may be lower than n the case of the duopoly f total peak capacty s unchanged. But, as proftablty would be lower, frms wll nvest less, so that the probablty of under-capacty rses. What s crucal here s that peak generators have to sell power to a prce hgher than ther margnal cost n order to fnance ther fxed costs. As Borensten (2000) shows, a prce equal to the margnal cost of the system can be suffcent to contrbute to fxed costs of base and half base

6 generators. Peak generators must beneft from market power, or they have to get some fxed revenues to balance fxed costs. Borensten ndcates that ths s the case n Calforna where stand by payments (e to be pad for gettng ready to produce n case of demand peak) exst. But those last revenues are generated by ntervenng on reserve markets, where unexpected varaton n demand are suppled n real tme. Some peak generators may not get much from those markets snce they remunerate speed, not capacty n tself. And no revenues are generated for peak generators when demand s correctly expected on a day to day bass. Furthermore, the total cost of nvestment (buldng the generators and stand by payments ) n peak capacty can be hgher than smply buldng the capacty. Reserve markets may not be enough to tackle the problem of peak nvestment. Another mechansm s explctly desgned to prevent stuatons of under-capacty : a (hgh) crss prce s mposed on the balancng market n order to ncte producers. It may not be effcent snce, as the model shows, even f the prce s equal to the cost of falure (a very theoretcal stuaton snce consumers are then ndfferent between consumng or not) the nstalled capacty by the duopoly s lower than that nstalled by a publc monopoly. All seems to work as f there was a knd of n-bult mechansm n electrcty market that nduce producers to lmt ther nstalled peak capacty n order to rase prces. Every dsposton desgned to lower market power dmnshes at the same tme the nctaton to nvest. 4.2 Market power and ts measures Those results also tell us that market power may be more dffcult to measure than we thnk. In our model we have two dstnct stuatons : one of under-capacty and market power, the other of over-capacty wth a prce equal to margnal cost. Over a long perod of tme, the two stuatons appear. A measure of market power by estmatng the mark up rate over a year (as done by Wolfram (1999) n order to compare dfferent years) can then be based. If t s measured over a hot year, demand s lower than a normal year, and over-capacty prevents the exstence of market power. On the opposte, a cold year wll yeld more tenson on prces because of a lack of capacty durng demand peaks, thus gvng hgher rates. Mark up rates are relevant f measured n normal demand condtons (from a statstcal pont of vew). Of course, ths s a much more dffcult operaton, whch requres modelzng the whole load curve and ts probablty dstrbuton. The conclusons of nvestgatons smlar to Wolfram (1999) and especally comparsons between rates measured on dfferent years, are then less nstructng. Other dffcultes arse when we consder that many unexpected events alter the way demand evolves. For example, the development of ar-cooled devces n France have made summer electrcty consumpton a lot more temperature dependant. Ths s one of the feature that has provoked the august 2003 prce spke n Europe. Such an unexpected development s only perceved when temperature effectvely provokes a rse n consumpton, that s at the worst moment. Ths explans why, though the general opnon two years ago was that too much capactes were nstalled n Europe, prce spkes happened. If we consder that peak producers market power s relatve to potental demand, then we may not be able to record t or only too late. Ths s a strong source of concern for regulators. 5. CONCLUSION Peak nvestment s a crucal problem to ensure a good supply of energy n a country. It also seems to be a growng problem, as recent crss on wholesale markets n Europe have shown. The model presented here s an attempt to show that deregulated systems may have dffcultes to cope wth t, snce prvate frms have a strong nterest n lmtng peak capactes. And ths problem s all the more sharp that beng able to observe market power on peak supply market The Hague - September 2003 AER/CPB/ECN

7 requres a very good knowledge of the way consumpton evolves. Computng mark up rates wthout that nformaton may lead to wrong conclusons. Furthermore, mechansms whose object s to deal wth peak problems (reserve markets, crss prces on balancng markets) does not seem adapted to deal wth peak nvestment. The model also stresses the lmts of tradtonal electrcty markets models. By not takng nto account capacty constrants and randomness they gnore very mportant characterstcs of the generaton ndustry. Comparng alternatve regulatons on the unque bass of those models s lkely to nduce mstakes. Of course, the present model s very lmted tself, especally concernng the prcng mechansm. It does not nclude dynamcs, nor does t examne the effects of the threat of entry. And though t shows that a publc monopoly better deals wth peak nvestment than a prvate duopoly, t s not enough to conclude to the superorty of the later over deregulaton. It s way too smple to ntegrate all the parameters we have to take nto account n order to compare the two regmes. But, as an extenson of tradtonal models t helps to put a lght on the bases of theory n order to correct them. We have a strong need to enhance our knowledge of how a deregulated electrcty ndustry tackle the problem of nvestment. As recent examples show, t s not lkely that t always does well. It would be a strong contrbuton to the publc debate about ndustry regulaton and energy ndustry. If we do not, and f regulators do not handle the problem, the future s prone to brng new crses. REFERENCES Borensten S., Bushnell J. (1999) An Emprcal Analyss of the Potental for Market Power n Calforna s Electrcty Industry. The journal of Industral Economcs, vol xlv, No 3 sep, pp USA. Borensten S. (2000). Understandng Compettve Prcng and Market Power n Wholesale Electrcty Markets. The Electrcty Journal, vol 13, n 9, july, pp USA. Borensten S., Bushnell J., Knttel C. R. (1999). Market Power n Electrcty Markets: Beyond Concentraton Measures. The Energy Journal, 20(4): USA. von der Fehr N-H. M., Harbord D. (1993) Long Term Contracts and Imperfectly Compettve Spot Markets: a Study of the UK Electrcty Industry. The Economc Journal, 103, pp UK. Green R., Newberry D. M. (1992) Competton n the Brtsh Electrcty Spot Market. Journal of Poltcal Economy, 100, pp UK. Green R. (1996) Increasng Competton n the Brtsh Electrcty Spot Market. Journal of Industral Economcs, 44, pp UK. Joskow P. L. (2001) Calforna s Electrcty Market Meltdown. Economes et Socétés, sére EN n 8, jan-fev. USA. Klemperer P. D., Meyer M. A. (1989) Supply Functon Equlbra n Olgopoly under Uncertanty. Econometrca, 57, pp Kreps D., Schenkman J. A. (1983) Quantty Precommtment and Bertrand Competton Yeld Cournot Outcomes. RAND Journal of Economcs, 14, pp USA. Staropol C. (2001) Concepton de marchés effcaces pour les secteurs déréglementés. Le cas des marchés de gros d électrcté. Thèse pour obtenr le grade de Docteur en Scences Economques de l Unversté Pars 1. France. Wolfram C. (1999) Measurng Duopoly Power n the Brtsh Electrcty Market. Amercan Economc Revew, 89 (4 sep): USA.