Preemption of TCCs and Deadweight Loss in Centrally Dispatched Electricity Systems with Competitive Generation. Shmuel S. Oren

Transcription

1 PWP-04 Preempton of TCCs and Deadweght Loss n Centrally Dspatched Electrcty Systems wth Compettve Generaton Shmuel S. Oren August 996 Ths paper s part of the workng papers seres of the Program on Workable Energy Regulaton (POWER). POWER s a program of the Unversty of Calforna Energy Insttute, a multcampus research unt of the Unversty of Calforna, located on the Berkeley campus. Unversty of Calforna Energy Insttute 539 Channng Way Berkeley, Calforna

2 Preempton of TCCs and Deadweght Loss n Centrally Dspatched Electrcty Systems wth Compettve Generaton.. Introducton Shmuel S. Oren Unversty of Calforna at Berkeley. August 5, 996 The electrc power ndustry around the world s undergong a revolutonary transton from vertcally ntegrated regulated or government run monopoles to a compettve ndustry. The transmsson network plays an essental role n ths transton by provdng the crtcal nterconnecton between geographcally dspersed markets. It s, therefore, generally agreed that open access va the transmsson network s the key to a compettve electrcty ndustry. The US electrcty system s characterzed by a hgh degree of nterconnecton and dverse ownershp of transmsson assets and generaton resources. Consequently, the mplementaton of open transmsson access n the US hnges on the establshment of property rghts and usage protocols for the transmsson network along wth mechansms for ownershp compensaton and usage charges. The Federal Energy Regulatory Commsson (FERC) (Aprl 4, 996) has recently proposed a set of gudelnes for transmsson capacty reservaton and use that would mplement the basc requrements for open transmsson access to support a compettve electrcty market. The FERC proposal defnes transmsson servce n terms of frm transmsson capacty from specfc ponts of delvery to specfc ponts of recept on the transmsson grd. Capacty reservatons can be traded among market partcpants and reconfgured as long as the reconfguraton does not nfrnge on other capacty reservatons. Unutlzed capacty reservatons or ncremental use of capacty when the transmsson system encounters constrants, may be compensated or charged for, based on opportunty costs. In a recent report, Harvey, Hogan and Pope (996) advocate the mplementaton of FERC's Transmsson Capacty Reservaton (TCR) by means of a spot market wth Transmsson Congeston Contracts (TCCs). The authors renterpret the orgnal Network Contracts dea ntroduced by Hogan (99) n terms of the language and prncples artculated n the FERC proposal. Specfcally, the authors propose a market structure n whch generaton as well as demand are bd compettvely at each node on the grd nto a central pool and are dspatched centrally by an Independent System Operator (ISO), so as to meet some optmalty crtera and satsfy transmsson constrants. The resultng market

3 clearng prces at each node (adjusted for losses) become the nodal spot prces pad to generators and charged to consumers at the respectve nodes. In ths framework the rght (or oblgaton) to nject power at one node and take out power at another node takes the form of a TCC whch s a fnancal contract that enttles (or oblgates) ts holder to receve (or pay) the nodal spot prce dfference between the respectve nodes. In ths desgn transmsson rghts ownershp plays a passve role n the market whle the actve tradng takes place n generaton. The market value of TCCs (whch s determned ex-post) s establshed mplctly by the compettve generaton n conjuncton wth optmal dspatch by the ISO. In ther paper, Harvey, Hogan and Pope (996) argue that tradng, reconfguraton and opportunty cost compensaton of TCRs can be accomplshed wthn a pool based system by turnng these TCRs over to the ISO n return for TCCs. The optmal dspatch by the ISO can be vewed as an optmal reconfguraton whch maxmzes the value of the TCRs and the opportunty costs (or value) resultng from such reconfguraton accrues to the TCC holders. The argument that transmsson tradng s mplct n economc dspatch fals, however, to recognze the strategc mplcaton of replacng actve tradng n transmsson capacty wth passve ownershp compensated ex-post based on the energy tradng outcomes. The analyss supportng the above argument and ts concluson hnge on the premses that the energy market does not react to the way n whch transmsson property rghts are beng exercsed and that n the absence of locatonal market power n generaton bd prces wll be drven to margnal costs. These premses are based on economc theory that has not dealt wth explctly wth the mplcaton of congested dstrbuton channels and has not been substantated by any emprcal evdence. On the contrary, lmted expermental results suggest that n the absence of actve market partcpaton by transmsson rghts owners bd prces for generaton may devate from margnal costs whch wll defeat the TCC based approach. The purpose of ths paper s to hghlght the mportance of actve tradng n transmsson capacty and the possble mplcatons on competton n generaton of passve transmsson ownershp (.e., no ex-ante charges or actve market partcpaton through bddng or tradng of rghts). By usng a Cournot model of competton n a congested transmsson network we llustrate that even n the absence of local market share concentraton n generaton, passve transmsson can result n mplct colluson among generators who wll capture the congeston rents and preempt the TCCs. The prce dstortons wll result n short term neffcency due to suboptmal dspatch and msplaced ncentves for long-term market decsons.

4 The remander of ths paper s organzed as follows: In Secton two we dscuss the use of Cournot models of competton n the context of market power analyss and analyze the strategc mplcatons of a passve congested transmsson lnk separatng the supply sources from the demand. In secton 3 we extend our analyss to Cournot competton n a three node system wth two supply nodes and one demand node. Secton 4 examnes the prce dstortons and short term effcency mplcatons of the Cournot equlbrum. Secton 5 contans a bref perspecve on alternatve approaches and concludng remarks.. Cournot Competton over a Congested Transmsson Lnk A common aspect of pool-based market organzaton paradgms for compettve electrcty s the relance on compettve generaton (n conjuncton wth central control of the transmsson system) to acheve an economcally effcent outcome. The above desgn s based on the ratonale that competton n generaton wll drve prces at each locaton on the network to margnal cost whle centralzed control of the transmsson system could (wth proper nsttutonal desgn) acheve optmal dspatch. The UK system, the generc POOLCO descrbed by Ruff (994) and the Calforna Power Exchange/ ISO desgn advocated by Joskow (996) are all based on the above ratonale. From a theoretcal perspectve, the assumpton that prces wll be drven to margnal cost would be supported by a Bertrand competton model. However, emprcal evdence from the UK system ndcates that electrcty spot prces systematcally exceed margnal costs (see e.g. Newbery (996)) suggestng that a Bertrand competton model s unrealstc n ths context. Green and Newbery (99) attrbute the hgh prce markup n the UK system to the fact that there are only two generatng companes that can exercse market power. They analyze the Duopoly market power and the potental effect of more compettors usng the theoretcal framework of supply functon equlbra ntroduced by Klemperer, and Meyer (989). Ther analyss does not address, however, effects of locatonal concentraton or transmsson constrants. The possble strategc explotaton of transmsson constrants by generators n order to gan market power has been explored by Hogan (995) who consdered possble nteracton between a local monopoly generator facng competton across a congested network. The effect of transmsson constrants on market power has also been examned by Borensten, Bushnell, Kahn and Stoft (995) and (996) n the framework of Cournot equlbrum among generators across a congested transmsson system. An mplct assumpton n all the models attemptng to analyze market power (wth or wthout transmsson effects) s that departure from margnal cost prcng s due to some 3

5 form of general or locatonal market share concentraton. It s mpled by the above assumpton that n a congested network, elmnaton of locatonal market power through dversfcaton of generaton ownershp at each locaton would drve prces to margnal cost and acheve economc effcency. The above mplcaton s reflected n the analyss of horzontal market power for the proposed Calforna electrcty market by Joskow, Frame and Heonymous (996) whch s based on evaluatng a market concentraton ndex (HHI) wthn each congeston zone. Ths perspectve mght be rooted n the fact that a smple Cournot olgopoly model produces margnal cost prces when the number of compettors s very large. As we shall demonstrate, however, ths concluson may not be vald when the compettors are separated from the demand by a congested transmsson network. Recent expermental results reported by Backerman, Rassent and Smth (996), suggest that factors other than locatonal market power may cause prces n congested electrcty network to devate from margnal costs. The paper descrbes a unform prce double aucton experment conducted at the Economc Scence Laboratory at the Unversty of Arzona, wth multple sellers and buyers tradng across a constraned passve lnk (.e., there are no ex-ante transmsson charges or actve tradng by transmsson rghts holders). The authors observed that the bdders on both sdes wll ndependently adjust ther prces so capture the congeston rents. Fgure, llustrates conceptually how the bd supply curve and bd demand curve (n sold lnes) can dvate from the correspondng cost and wllngness-topay functons (n dashed lne) so as to capture all the congeston rent. Ths observaton suggests that n a congested electrcty network, even n the absence of locatonal market power, prces may devate from margnal costs due to mplct colluson among generators (and possbly large consumers) that s nduced by ratonal expectaton of congeston. To model ths phenomenon, we employ a smple two node system consstng of a supply node and a demand lnked by a transmsson lne wth a maxmum lne flow capacty Q. The demand node s represented by a down slopng nverse demand functon P(q) wth prce elastcty ε( p), whle at the supply node there are n symmetrc supplers wth dentcal margnal costs c. Fgure llustrates the model and the demand functon P*(q) as seen at the supply node. Because the lne capacty s publc knowledge the demand functon s seen as becomng nelastc at transmsson capacty Q. ε( p) = p / qp ( q) 4

6 MC S q < Q D WTP q q Prce Wllngness-to-pay Bd Supply curve Margnal Cost curve Congeston rent Bd Demand curve Q Quantty (q) Fgure : Buyers and seller capture congeston rent n suppler s MC=c S q < Q D P(q) Demand q P*(q) Demand funct on as seen at supply node P(q) Congeste d equlbr um P(Q) c Congest. rent Q q Fgure : Cournot Competton Over a Congested Supply Channel 5

7 In Cournot competton each suppler takes the supply quanttes qj of each of ts compettors as fxed and sets the market prce p by selectng ts own quantty q. Hence the optmzaton problem defnng the best response of suppler s: Max π = q [ P( q + q ) c] q j j Subject to : 0 q Q q j j Invokng symmetry n the frst order necessary condtons for the above problem yelds the Cournot equlbrum condtons: dπ q = Q and = p( + c dq nε( p) ) 0 Or dπ q < Q and = p( + c = dq nε( p) ) 0 Note that the above condtons do not determne a unque prce when the equlbrum s on the vertcal segment of the perceved demand functon snce the above condtons can be satsfed by any par (p,q) such that q=q and c p P( Q) (assumng c P( Q)). However, the proft maxmzaton by each compettor mples that the equlbrum prce correspondng to q=q must be at ts hghest possble level p=p(q) regardless of n. It thus follows that when the equlbrum s on the downslopng porton of the demand curve, the prce decreases as the number of compettors ncreases untl t ether drop to margnal cost or to P(Q), whchever s larger. From there on the equlbrum prce does not change regardless of n, Thus for suffcently large n, a congested Cournot equlbrum s at the knk of the effectve demand curve or n general: p = Max[, c P( Q)], q = Mn[ P (), c Q] and q = q / n for all The key mplcaton of the above example s that although none of the generators have market power by any measure of concentraton, ratonal expectaton of congeston leads to mplct colluson whch leads to prcng at the level that the market wll bear for the constraned supply quantty. Consequently, the congeston rents are extracted by the actve traders (n ths case the generators.) Passve owners of transmsson rghts n the form of TCCs that are compensated ex-post based on nodal prce dfferences wll be dsapponted to dscover that no rents have been generated. Ths concluson s consstent wth Coase Theorem (960) whch supports the argument that n the absence of transacton costs and publc knowledge of transmsson capacty, barganng among buyers and sellers wll capture all the congeston rents. Implct colluson based on ratonal expectatons s a 6

8 realstc phenomenon n an envronment were bds are submtted day after day every day of the year. As demonstrated by the evoluton of the supply curves n the UK (see Newbery (995)) the repettve nature of power auctons allows learnng-by-dong and refnement of the bddng strateges so as to explot market power or other rent extracton opportuntes. The above analyss hghlght the role of actve tradng by transmsson rghts owners vs. ex-post collecton of resdual rents from the compettve generaton market. Actve tradng of transmsson rghts can take place n parallel wth energy tradng n a decentralzed market framework as descrbed by Chao and Peck 996), or through any ex-ante settlement such as a set fee or prenegotated prce. It s the separaton of the transmsson ownershp from the generaton ownershp n conjuncton wth the ex-post settlement aspect of TCCs that allows ther preempton by the generaton owners. In the two node example descrbed n ths secton, there are no short term effcency consequences from the overprcng of generaton. The only short term dstorton s a redstrbuton of rents among the stakeholders. In the long run, however, the prce dstorton wll msplace generaton nvestment ncentves by encouragng more generaton at the supply node. Furthermore, the preempton of TCC value by generators, creates a free rder effect wth respect to transmsson nvestment rewarded wth TCCs (as proposed by Bushnell and Stoft (995).). Whle both generators and consumers could beneft from expandng the transmsson lne by ncreasng, sales or reducng prces, the free rder effect dlutes the ncentves for such nvestments. A stakeholder makng such an nvestment unlaterally can only capture a proportonal share of the beneft through use of the transmsson lne whle addtonal property rghts n the form of TCCs whose remuneraton are determned ex-post by the ISO are vrtually worthless. In a more complcated system passve transmsson ownershp can also lead to short term neffcency as wll be demonstrated n the subsequent sectons. 3. Cournot Equlbrum n a Congested Three-Node Network. In ths secton we consder a three node case as llustrated n Fgure 3. One consumpton node represented by an nverse demand functon P(q), and two supply nodes: Node (CHEAPGEN) where there are n symmetrc generators wth margnal cost c and Node (DEARGEN) where there are n symmetrc generators wth margnal cost c (we assume c<c). For smplcty we assume a lossless DC network wth equal mpedance on all lnes and a lneflow lmt of Q on the lnk connectng the two generaton nodes. Because of the 7

9 cost symmetry at each supply node we may assume that n equlbrum there wll be a sngle bd prce at each node, p and p respectvely where p p. n generators MC = c, Prce = p q q = n = q Y Y q Q P(q) 3 Y q n generators MC = c, Prce = p q = n q = Fgure 3: Competton n a Three Node Network Gven the prces bd at each node and the publcly known demand functon, the ISO determnes the optmal dspatch.e,. the total njecton at each of the supply nodes so as to maxmze total welfare subject to the transmsson constrants. Hence, n equlbrum the ISO's optmzaton problem for ths example s: Max P( q) dq pq pq (Socal Surplus) q, q q + q 0 Subj. to : q q 3Q (Krchoff Law) q q 0 (Equlbrum condton) The optmal dspatch soluton has two possble cases dependng on whether power s or s not generated at DEARGEN (ths wll depend on the relatonshps between the varous problem parameters). Case I (no power from DEARGEN): q = 0 or q = P ( p ) < 3Q and p = p 3 q = 3Q and p = P( 3Q) p 3 3 8

10 Case II (Both nodes generate) q > 0 or p = p = p, q + q = P ( p ), q q 3Q 3 3 p = ( p + p )/, q = [ P ( p ) + 3Q]/, q = [ P ( p ) 3Q]/ Note that for ths example, under optmal dspatch, both nodes wll supply power smultaneously only f the bd prces are equal or f the network s congested (t s possble however, to have congeston wth only the cheaper node generatng). Furthermore, when both supply nodes generate, the sellng prce at node 3 (under optmal dspatch) s the average of the bd prces at DEARGEN and CHEAPGEN. The ISO's behavor s publc knowledge and can be taken nto consderaton by the compettors at each node. Thus, we can derve an effectve demand functon at each of the supply nodes condtoned on the total njecton at the other supply node as llustrated n Fgure 4. Demand functon at CHEAPGEN Demand functon at DEARGEN P (q ;q ) P (q ;q ) On vertcal segment f q >0 then On vertcal segment p =P(q -3Q)-p P(3Q+q ) P(q +q ) p =P(3Q+q )-p P(q -3Q) P(q +q ) o 3Q+q q q -3Q q q Fgure 4: Condtonal Demand Functons at Supply Nodes The condtonng of each demand functon on the njecton at the other supply node s consstent wth the conjectural varaton of Cournot competton. Accordngly each compettor selects ts supply quantty and sets market prces so as to maxmze ts proft takng as gven the quanttes of all other supplers (at ts own node) as well as the other node. The ISO's response s accounted for n the nodal demand functon shown above. On the down slopng segments of the demand curves, selecton of quanttes by a compettor sets the unform market prce at all nodes. On the vertcal segments of the demand functon 9

11 (congeston case) prces at the two supply node may dverge but the average prce of the two supply nodes sets the prce at the demand node (through the optmal dspatch by ISO). In dervng an equlbrum for ths three node system we need agan to dfferentate between the case where only CHEAPGEN generates power and the case were both nodes generate. Ths depends on the relatonshp between the demand functon and the margnal prces at the two supply nodes. When no power s generated at DEARGEN the demand functon seen at CHEAPGEN s the same as n the two node case but wth total flow capacty of 3Q. Hence, the correspondng symmetrc equlbrum can be derved as n the two node case whch, for a suffcently large n yelds: p = Max[ c, P( 3Q)], q = Mn[ P ( c ), 3Q], q = q / n for all Next we consder the case where the two supply nodes generate. In theory, t s possble to have an uncongested equlbrum on the down slopng segments of the two demand functons wth equal bd prces at both supply nodes. Such an equlbrum prce p would have to satsfy the two smultaneous equatons: p( c and p c n ( p) ) ( + 0 n ( p) ) = + = 0 ε ε For suffcently large number of compettors at both nodes, as the prces approach margnal cost, these equatons become nconsstent and such an equlbrum can be ruled out. It thus remans to consder only congested equlbra whch correspond to the vertcal segments of the effectve demand. We argue that for a suffcently large number of generators at both supply nodes the only possble Nash equlbrum s at the knks of both demand curves, where q q = 3Q. For CHEAPGEN we may use the best response argument made for the two node case to prove that for a suffcently large number of supplers the local Cournot equlbrum (condtonal on q ) s at the knk. Compettors at DEARGEN would prefer a hgher prce on the vertcal segment of ther demand functon (above the knk) whch would correspond to a lower prce at CHEAPGEN. However, such prces would not support a Nash Equlbrum snce any generator at CHEAPGEN can unlateraly (and proftably) set the knk prce wthout affectng any of the other supply quanttes by slghtly "backng off" from the vertcal segment of the demand functon. It follows that the only possble congested Cournot-Nash equlbrum wth generaton at both supply nodes s at the knks. Furthermore, snce under Cournot competton, each compettor acts as a unlateral prce setter and no sde payments are allowed the prces at each supply are bounded below by the correspondng margnal costs. Consequently the congested equlbrum must satsfy: 0

12 q q = 3Q p = p = p = P( q 3Q) = P( q + 3Q) c 3 Suppose, however that P( q 3Q) > c than for large n, p c whch mples q < q 3Q, thus contradctng the congeston hypothess. Consequently, a lmtng congested equlbrum (for a suffcently compettve generaton) must satsfy the necessary condtons: p = p = p = c, q q = 3Q, q + q = P ( c ) 3 The proof that these condtons ndeed defne a Cournot-Nash equlbrum follows from the fact that no compettor can ncrease ts profts unlaterly gven that everyone else hold ther quanttes fxed. ) P( 3Q) c q = 0, q = P ( c ), p = p = c 3 ) c P( 3Q) c q = 0, q = 3Q, p = p = P( 3Q) 3 ) c P( 3Q) q = [ Pc ( ) 3Q]/, q = [ Pc ( ) + 3Q]/, p = p = p = c 3 As n the two node case none of these equlbra produce resdual congeston rents to be pad to TCC holders. All such rents n ths case are captured by the CHEAPGEN generators through mplct colluson. In case ) there s no congeston, There s no generaton at DEARGEN and competton drves bd prces at CHEAPGEN to margnal cost. In case ) there s congeston but DEARGEN does not generate and the bd prce at CHEAPGEN equals the market prce at the demand node whch s between the margnal costs at the two supply nodes. In ths case the congeston rent whch totals 3Q [ P( 3Q) c] s captured as profts by the CHEAPGEN supplers. In Case ) There s congeston and both supply nodes are generatng. The equlbrum prce at all nodes equals the margnal cost at DEARGEN and the congeston rent whch s captured as proft by the CHEAPGEN supplers equals ( c c ) [ P( c ) + Q] Short Term Effcency Losses. Unlke n the two node case, the Cournot equlbrum derved above for three nodes may correspond to a socally neffcent dspatch. The socally effcent dspatch s gven by the optmal dspatch correspondng to bd prces that equal margnal cost. Thus the dfferent cases for the socally effcent dspatch can be obtaned from the earler soluton to the ISO problem by substtutng p = c and p = c as follows:

13 ) P( 3Q) c p = c, q = 0, q = P ( c ), 3 c P Q c c + ) ( 3 ) p3 = P( 3Q), q = 0, q = 3Q c c c c ) P( 3Q) p, q [ P( c c = = ) 3Q]/, q = [ P( c c ) + 3Q ]/. Fgure 5 compares the Cournot equlbrum wth the socally effcent dspatch for the dfferent cases arsng from the relatonshp between the demand functon and the margnal costs at the two supply nodes. The black areas ndcate the deadweght losses due to prce dstorton n the equlbrum soluton and the gray area marks the rents captured by the CHEAPGEN supplers. P(3Q) P(3Q)>c Cournot eq. Effcency loss c >P(3Q)> Cournot eq. c +c Effcency loss c c +c c profts at Effcent dsp. c P(3Q) c +c c profts at Effcent dsp. 3Q q +q 3Q q +q c +c >P(3Q)>c c >P(3Q) c c +c P(3Q) c Cournot eq.= Effcent dsp. c c +c c P(3Q) Cournot eq.= Effcent dsp. 3Q q +q 3Q q +q Fgure 5: Deadweght Loss and Supplers Profts at Cournot Equlbrum In the two bottom cases there s no effcency loss but the congeston rents are captured by the CHEAPGEN supplers. In the two upper cases, however, the Cournot equlbrum results n a deadweght loss whch cannot be avoded through unlateral best response at

14 ether supply node. At CHEAPGEN, whle generators could beneft from ncreasng ther generaton, they are prevented from dong so by the congeston constrants whle at DEARGEN, a unlateral ncrease n generaton would reduce the prce below the margnal cost at that node. 5. Concluson Externaltes resultng from the physcal laws governng the flow of electrcty make transmsson capacty an essental nput factor or a by product of generaton n the provson of electrc power n an nterconnected grd. The Federal Energy Regulatory Commsson has recognzed the crucal role of transmsson servce n the provson of compettve electrcty and establshed gudelnes for defnng and tradng that nput factor. Some current proposals clam, however, that an explct transmsson servce market s unnecessary and effcent transmsson tradng can be acheved ndrectly n a pool framework through a compettve generaton market and optmal dspatch by an ISO. Ths asserton s based on the premse that compettve generaton can drve generaton prces to margnal costs and any departure from margnal cost bddng s attrbutable to market power. In ths paper we demonstrated the mportance of actve partcpaton of transmsson rghts owners n the market n order to acheve the desred effcent outcome of competton. Through smple stylzed examples t shown that even n the absence of market power, expectaton of congeston and passve transmsson can lead to mplct colluson among generators and departure from margnal cost prcng. Ths nvaldates the key premse underlyng the ndrect mplementaton of transmsson rghts tradng through optmal dspatch by the ISO. We show that passve transmsson rghts (n the form of TCCs) by the ISO wll be preempted by the actve traders who wll adjust ther prces so as to capture the congeston rents. Prce dstortons due to congeston and passve transmsson ownershp can result n short and long term neffcency. Short term effcency losses resultng from prce dstorton, may be mtgated by market mechansms that would allow sdepayments among generators. For nstance, supplers at CHEAPGEN could offset the losses ncurred at DEARGEN due to counterflow-producng generaton that releves the congeston constrant at CHEAPGEN. The Chao and Peck (996a, (996b) approach provdes such a mechansm through parallel pre-dspatch markets for transmsson rghts and energy forwards admnstered through a power 3

15 exchange. A key feature s that through actve market tradng, the owners of rghts on congested lnks can capture rents on these lnks, and a set of effcent prces wll emerge to gude nvestment n future transmsson capacty. Furthermore, a generator creatng counterflow on a congested lnk s enttled to the rghts for the relved congeston whch can be sold n the transmsson rghts market. Ths creates the mechansm for sde payments that are needed to move the Cournot equlbrum to the socally effcent dspatch. In our partcular three node example, generators at DEARGEN can subsdzed ther sales of power below margnal cost by sellng ther counterflow rghts on the congested lnk to generators at CHEAPGEN. Wth suffcently compettve generaton at each node such tradng wll result n an effcent market equlbrum wth nodal generaton prces equal to the margnal costs. In ths framework TCCs can serve as a valuable hedgng nstruments whose prce wll are determned (correctly) by the actve transmsson rghts market. It should be noted that short term effcency losses resultng from passve transmsson rghts can be, alternatvely, mtgated through drect negotaton among the generators (possbly facltated by brokers) that wll result n ncremental "congeston neutral" proftable transactons. Such a scheme has been descrbed by Wu and Varaya (995) who advocate a "mnmal" ISO that manages congeston by enforcng feasblty of the proposed generaton schedules and who provdes coordnatng nformaton on feasble ncremental multlateral transactons. However, whle such an approach could potentally elmnate the deadweght loss, t does not resolve the generaton prce dstorton problem caused by the absence of actve transmsson rghts tradng. In summary, the analyss presented n ths paper offers an explanaton to the recent expermental results wth tradng over passve congested lnks. The stylzed examples are ntended to hghlght possble ptfalls n some recent proposals for mplementaton of FERC polcy rulngs. Even wth a three node system there are many ways of organzng supply and demand and many questons that were not addressed by our analyss. For example, local demand at the cheap generaton node may ntensfy local competton and drve prces down at that node. We have not examned the effect of asymmetry n generaton costs at a partcular node and the effect of demand sde bddng. In should be noted, however, that the experments at the Unversty of Arzona ESL whch revealed the capture of congeston rents by the tradng partes, ncluded supply and demand sde bddng of multple blocks wth dfferent margnal costs and values (but wth no loop flow effects.) After the market concludes, the system operator performs the actual dspatch to mplement these contracts and to ensure the securty of the system. The dfferences between the actual dspatch and the forward contracts wll be reconcled subsequently n a settlement procedure. 4

16 Fnally, the man objectve of ths paper s to rase questons and challenge some premses underlyng mportant polcy recommendatons that are not fully supported by ether theory or emprcal evdence. Hopefully ths work wll also stmulate fundamental research on olgopolstc competton and compettve bddng n congeston prone dstrbuton networks. References: Backerman Steven R., Stephen J. Rassent and Vernon L. Smth, "Effcency and Income Shares n Hgh Demand Networks: Who Gets the Congeston Rents When Lnes are Congested?., ESL Report, Unversty of Arzona (996). Borensten, Severn and James Bushnell, Edward Kahn and Steven Stoft, "Market Power n Calforna Electrcty Markets," Report PWP-036, Unversty of Calforna Energy Insttute, (December 995) Borensten, Severn and James Bushnell, and Steven Stoft, "The Compettve Effects of Transmsson Capacty n a deregulated Electrcty Industry", Draft Report, Unversty of Calforna Energy Insttute (May, 996) Bushnell, James and Steven Stoft, "Transmsson and Generaton Investment n a Compettve Electrc Power Industry," Journal of Regulatory Economcs, Vol. 0 (July 996). Chao, Hung-po and Stephen C. Peck "A Market Mechansm for Electrc Power Transmsson", Journal of Regulatory Economcs, Vol. 0 (July 996a). Chao, Hung-po and Stephen C. Peck "An Insttutonal Desgn for an Electrcty Contract Market wth Central Dspatch" Workng Paper, Electrc Power Research Insttute, Palo Alto CA, (June 996b) Coase, Ronald, "The Problem of Socal Cost", Journal of Law and Economcs, Vol. 3 (960) pp Federal Energy Regulatory Commsson, "Capacty Reservaton Open Access Transmsson Tarffs," Notce of Proposed Rule-makng, RM , Washngton DC, (Aprl 4, 996) Green, Rchard J., and Davd M. Newbery, "Competton n the Brtsh Electrcty Spot Market," Journal of Poltcal Economy, Vol. 00, No. 5 (99) pp Harvey, Scott M., Wllam W. Hogan and Susan L. Pope, "Transmsson Capacty Reservatons Implemented Through a Spot Market wth Transmsson Congeston Contracts," Techncal Report, (June 6, 996) Hogan, Wllam W., "Contract Networks for Electrc Power Transmsson," Journal of Regulatory Economcs, Vol 4, (99) pp. -4. Hogan, W. Wllam., "Market Power Analyss and Transmsson Constrants: An Exploraton", Harvard Electrcty Polcy Group Specal Semnar, (December 7, 995). Joskow, Paul L. "Restructurng to Promote Competton n Electrcty: In General and Regardng the Poolco Vs. Blateral Contracts Debate", Workng Paper, Presented at the AEA Meetng, San Francsco Calforna (January 996). 5

17 Joskow Paul L., Rodney Frame and Wllam Heronymous, "Report on Horzontal Market Power Issues; Chapter One: Basc Analytcal Framework", Docket No. ER , Flng by PG&E, SCE and SDG&E befor the Federal Energy Regulatory Commsson, (May 9, 996) Klemperer, Paul D. and Meyer, Margaret A. "Supply Functon Equlbra n Olgopoly under Uncertanty," Econometrca, Vol 57, No. 6, (989) pp Newbery, Davd M., "Power Markets and Market Power," Energy Journal, Vol. 6, N0. 3 (995) Newbery, Davd M., "Electrcty Power Sector Restructurng: England and Wales", POWER Conference on Electrcty Industry Restructurng, Berkeley, Calforna, (March 5, 996). Ruff, L., "Stop Wheelng and Start Dealng: Resolvng the Transmsson Dlemma,", Electrcty Journal, vol. 7, No. 5 (994) pp Wu, Felx and Pravn Varaya, "Coordnated Multlateral Trades for Electrc Power Networks: Theory and Implementaton", Report PWP-03, Unversty of Calforna Energy Insttute, (June 995) 6