Economic Inefficiencies of Cost-based Electricity Market Designs

Transcription

1 Economic Inefficiencies of Cost-based Eectricity Market Designs Francisco D. Munoz,* Sonja Wogrin,** Shmue S. Oren,*** and Benjamin F. Hobbs**** ABSTRACT Some restructured power systems rey on audited cost information instead of competitive bids for the dispatch and pricing of eectricity in rea time, particuary in hydro systems in Latin America. Audited costs are aso substituted for bids in U.S. markets when oca market power is demonstrated to be present. Reguators that favor a cost-based design argue that this is more appropriate for systems with a sma number of generation firms because it eiminates the possibiities for generators to behave strategicay in the spot market, which is a main concern in bid-based markets. We discuss existing resuts on market power issues in costand bid-based designs and present a counterintuitive exampe, in which forcing spot prices to be equa to margina costs in a concentrated market can actuay yied ower socia wefare than under a bid-based market design due to perverse investment incentives. Additionay, we discuss the difficuty of auditing the true opportunity costs of generators in cost-based markets and how this can ead to distorted dispatch schedues and prices, utimatey affecting the ong-term economic efficiency of a system. An important exampe is opportunity costs that diverge from direct fue costs due to energy or start imits, or other generator constraints. Most of these arise because of physica and financia infexibiities that become more reevant with increasing shares of variabe and unpredictabe generation from renewabes. Keywords: Eectricity market design, market power, equiibrium modeing, opportunity costs 1. INTRODUCTION Dereguated eectricity markets have been in pace for neary three decades. In the 1980s Chie and the U.K. were the first countries to divide the od verticay-integrated monopoies into private generation, transmission, and distribution companies. To date more than 30 countries and states in the U.S. rey on merchant generation firms to ensure adequate investment in new generation capacity to suppy eectricity at minimum cost to fina consumers, with varying degrees of success (Griffin & Puer, 2009). One feature of dereguation that is not uniform among a countries and states that have impemented competitive markets for generation is how they dispatch and price eectricity in rea * Corresponding autor. Facutad de Ingeniería y Ciencias, Universidad Adofo Ibáñez, Santiago, Chie. E-mai: fdmunoz@uai.c. ** Instituto de Investigación Tecnoógica, Escuea Técnica Superior de Ingeniería, Universidad Pontificia Comias, Madrid, Spain. *** Department of Industria Engineering and Operations Research, University of Caifornia, Berkeey, Berkeey CA. **** Department of Environmenta Heath & Engineering, The Johns Hopkins University, Batimore MD. The Energy Journa, Vo. 39, No. 3. Copyright 2018 by the IAEE. A rights reserved. 51

2 52 / The Energy Journa time. A arge group of dereguated markets aow generators to submit their suppy curves (which may refect opportunity cost) as bids to a centraized auction that is coordinated and ceared by a System Operator (SO). Some exampes of such bid-based markets incude the ones in PJM, Caifornia, Texas, MISO, New Engand, New Zeaand, and Coombia. In contrast, there is another group of restructured markets where generators are not aowed to bid their suppy curves. Some of these incude Chie, Boivia, Peru, Brazi, and countries in Centra America (Hammons et a., 2002). In these markets, the SO or system reguator performs audits of a generation parameters (i.e., heat rates, minimum oads, ramping imits, etc.) and fue costs to estimate each generator s direct margina operating costs. Based on this information, the SO dispatches generators hour by hour to meet demand at minimum direct cost. The optimization of the dispatch schedues for a generators to meet demand aso yieds shadow prices for power on every hour and bus in the network, which, depending on the market design, might be used to sette energy purchases and saes. Hereinafter we refer to this pricing mechanism as a cost-based market design. One issue that has received itte attention in the academic iterature is how cost-based markets fare when compared to bid-based ones in terms of their overa economic efficiency. In this artice, we discuss severa types of economic inefficiencies that can resut from reying on audited information instead of on bids that typicay refect both direct and opportunity costs, to determine the optima dispatch and prices of eectricity in a power system. We make two basic arguments. First, it is incorrect to argue that forcing generators to bid their margina fue costs eiminates a possibiities for the exercise of market power and thereby increases the economic efficiency of the system. As we wi discuss in the iterature review and demonstrate with a simpe exampe, by design cost-based markets do indeed, effectivey, prevent the exercise of strategic behavior in the short-run exacty the type of market power that reguators and fina consumers are most sensitive to. However, those markets aso can provide incentives for generation firms to strategicay seect capacities and technoogies that ead to a ong-run equiibrium that is distant from a perfecty competitive one, and that the resuting market inefficiency is difficut to correct through market rues if investments are dereguated. Our second argument is that, even in the absence of strategic behavior, identifying and auditing the tota margina costs of a generators in rea time is chaenging and ikey to ead to incorrect estimates and inefficient dispatch. Margina costs have two components: 1) direct costs that are directy attributabe expenditures on fue, operation and maintenance (O&M), and any other variabe inputs and 2) opportunity costs. The inefficient dispatch coud, in theory, be avoided if generators were aowed to bid both direct and opportunity costs instead of ony directy attributabe costs. Furthermore, the information required to compute the opportunity costs of a generators in rea time goes far beyond the responsibiities of the SO. This requires access to information concerning intertempora generator constraints as we as on parae markets, such as natura gas, emissions permits, and renewabe energy certificates. We organize the rest of this paper as foows. In Section 2 we discuss market efficiency and the roe of information in the price formation process in competitive eectricity markets. In Section 3 we discuss how market power can arise in bid- and cost-based market designs and how difficut it is to detect it. In this section we aso provide a counterintuitive exampe where forcing spot prices to equa margina costs in a concentrated market can resut in ower investment and socia wefare than under a bid-based design where generators can behave strategicay in the short term. In Section 4 we enumerate and discuss specific chaenges of estimating opportunity costs and how inaccurate estimates can distort market efficiency. Some exampes incude the opportunity costs of conventiona generators with infexibe fue contracts, opportunity costs of generators in markets A rights reserved. Copyright 2018 by the IAEE.

3 Economic Inefficiencies of Cost-based Eectricity Market Designs / 53 with emissions and/or renewabe energy poicies, and intertempora imits on starts, operating hours, and energy. Finay, in Section 5 we provide concusions and some poicy recommendations. 2. MARKET EFFICIENCY, PRICES, AND INFORMATION There are two independent concepts that are frequenty used to assess the overa economic efficiency of a market: aocative or Pareto efficiency and production efficiency. In simpe terms, aocative efficiency is achieved in eectricity markets where prices are equa to the margina cost of suppying an additiona unit of energy (Green, 2000). This margina cost captures not ony the incrementa fue cost, if the additiona power is produced using fue-powered pant, but aso any opportunity cost incurred by a generator that suppies power to the market or, if there is scarcity, the margina benefit of consumption. Aocative inefficiency means that too much or too itte eectricity is produced and consumed. On the other hand, an eectricity market is said to achieve production efficiency if power is produced in a way that minimizes the tota cost of producing that amount, given the generation technoogies that can be chosen from and the transmission infrastructure. 1 Production inefficiency means that more cost (incuding opportunity costs) is incurred in production than is necessary. In theory, a centra panner (i.e., a verticay-integrated pubic utiity that serves a customers in a geographica region) coud achieve both aocative and production efficiency under the assumption of perfect and compete information concerning investment aternatives, future technoogy and fue costs, and the true opportunity costs of using any resource to produce eectricity at a given time and ocation, among many other factors. If this assumption is true, a centra panner can achieve production efficiency by deveoping the portfoio of generation technoogies that minimizes the present worth of tota system costs, which can be identified by soving an optimization mode such as the one described in Hobbs (1995). In addition, aocative efficiency can be achieved by ater setting eectricity prices equa to the Lagrange mutipiers of the demand-baance constraints in the mode (Wenders, 1976; Sherai et a., 1982), so as to meet any demand whose margina benefit exceeds margina cost. 2 Unfortunatey, accessing a the needed information is impossibe for a centraized authority, which can ead the system to outcomes that are neither aocative nor production efficient. This is a major reason why most generation markets around the word are structured around private, but reguated, utiities or based on dereguated markets for generation. Private firms that seek to maximize profits can have more incentive to obtain information than a centra panner. 3 Private utiities that operate as reguated monopoies have one feature in common with power systems that are centray panned by a pubic utiity since a singe agent, the private utiity, makes a investment and operationa decisions. However, they differ in what Woak (2003) refers to as the individua rationaity constraint: the objective function of a private utiity is to maximize its profits, not to minimize tota system costs as pubic utiity. This feature, combined with information asymmetries between the private utiity and the reguator resuts in that It is virtuay impossibe to design a reguatory mechanism that causes a privatey owned profit-maximizing firm to produce in a east-cost manner (Woak, 2003, p. 14). Athough aocative efficiency coud be partiay achieved 1. In the rest of the artice we ignore transmission as a piece of infrastructure that coud affect the production efficiency of an eectricity market and focus soey on generation capacities and technoogies. 2. This is based on the theory of peak-oad pricing deveoped by Boiteux (1960) and ater extended by a series of different authors (Crew et a., 1995). 3. For some, an even stronger argument against centra panning is the independence of the eectricity market from costy poitica agendas (Joskow, 2008). Copyright 2018 by the IAEE. A rights reserved.

4 54 / The Energy Journa in a system served by a reguated monopoy under the unikey scenario that the reguator coud access a cost information of the eectric utiity firms facing cost-of-service reguation have itte incentive to attain production efficiency if a costs coud be passed on to fina consumers. The main objective of the introduction of competition in generation has been to increase the overa economic efficiency of the system with respect to what it coud be achieved through centra panning by pubicay-owned utiities or reguated monopoies. According to microeconomic theory, if there are no barriers to entry, there is access to perfect information, and scae economies and generation firms are sma reative to the size of the market among severa other assumptions a dereguated market shoud yied a perfecty competitive outcome that achieves both aocative and production efficiency (Green, 2000). In other words, the best strategy for profit-maximizing firms is to seect investments that minimize the average system cost and to have a suppy curve that refects the short-run margina cost of generation, incuding a opportunity costs. Depending on how they are coordinated, dereguated markets are cassified as sef-committed (biatera) or centray-committed (Pooco). In sef-committed systems, equiibrium prices emerge as a resut of biatera negotiations between generation firms and retaiers or consumers, and the responsibiities of the SO are imited to rea-time management of imbaances and ensuring the reiabe operation of the grid. In contrast, in a Pooco system, a SO cears the market through a competitive auction where generators can submit bids that refect the minimum price they are wiing to receive for producing power at a given ocation and time (incuding non-convex features of their cost structure such as start-up and no-oad cost), i.e., their direct and opportunity costs. These bids are an input to the SO s optimization probem that seects dispatch schedues for a generation units and which are used to determine prices that refect the incrementa cost of suppying an additiona unit of eectricity at a specific time and, in systems with ocationa margina pricing, in every node of the network. If information is perfect and markets are perfecty competitive, both biatera and Pooco structures shoud reach the same outcome (Wison, 2002). Both types of dereguated markets rey on the vountary participation of agents in the sense that generation firms are free to set the offers that best refect their direct and opportunity costs of generation, as recommended in the idea mode for dereguation described by Joskow (2008). If generation firms beieve they cannot affect the equiibrium price and they behave rationay as price takers, bidding their true direct and opportunity costs is a dominant strategy (i.e., no other bids can increase their profit), which then eads to prices that are aocative efficient in the short run and to investments that resut in production efficiency in the ong run (Green, 2000). 4 However, if markets are concentrated or there exist barriers to entry, generation firms can make strategic offers that wi affect equiibrium prices to their benefit. This means that submitting bids above their margina costs or withhoding capacity in order to increase prices above margina cost arises as a new dominant strategy (Cramton, 2004). The resut of such exercise of market power is ikey to be a decrease in the aocative and production efficiency of the market. Aocation efficiency is harmed because some demand whose margina benefit is ess than price but greater than the margina cost of serving it wi go unsatisfied. Short-run production efficiency decreases because higher prices wi encourage smaer firms to produce more power at a margina cost near to price, partiay repacing the ower cost power that arger firms have withdrawn from the market. Long-run production efficiency decreases if arge efficient firms decine to expand capacity in order to maintain high prices, whie costier smaer companies partiay fi the gap with more expensive investment. 4. This discussion disregards some of the compexities arising from mutipart bids of nonconvex costs and the use of bid-cost recovery systems by SOs. A rights reserved. Copyright 2018 by the IAEE.

5 Economic Inefficiencies of Cost-based Eectricity Market Designs / MARKET POWER IN BID- AND COST-BASED MARKET DESIGNS You can ead a horse to water, but you can t make it drink. Anonymous 3.1 Theoretica and empirica evidence In a bid-based market, generation firms have incentives to exercise market power anytime they face a ess than perfecty-eastic residua demand curve. They can do so by withhoding a fraction of their instaed capacity in their offers or by raising their bids, particuary at times when they are pivota suppiers (Stoft, 2002), 5 i.e., they can induce scarcity by uniateray withhoding production. Part of this behavior can be captured in a Cournot mode of oigopoistic competition. The Cournot mode is frequenty used in eectricity markets to depict the effect of strategic behavior, even when the bidding mechanism is not accuratey represented by the Cournot mode that assumes quantity setting. There is empirica evidence that actua prices in bid-based markets in the U.S. (i.e., Caifornia, PJM, and New Engand) are cose to the ones predicted using a static Cournot mode, particuary at times when demand is high and generation capacity is scarce (Bushne et a., 2008). Puer (2007) states that pricing in the Caifornia market was approximatey Cournot for the major firms during much of the period. The exercise of market power has been an important concern in bid-based eectricity markets because it can resut in potentiay arge weath transfers from consumers to producers and deadweight osses. The Caifornia eectricity crisis is the best exampe of a restructured eectricity market that was gamed at the expense of consumers because market rues were not set correcty. Puer (2007) as we as Lo Prete & Hobbs (2015) anayze the behavior of the 5 argest therma generators in the Caifornia eectricity crisis and contrast it with Nash equiibrium eves as we as tacit cousion, and find statisticay significant evidence of the exercise of market power. Borenstein et a. (2002) found that neary 60% of the $6.94 biion increase in eectricity expenditures from the summer of 1999 to the summer of 2000 were a direct consequence of non-competitive behavior of generation firms. Since the Caifornia crisis, the SOs and the Federa Energy Reguatory Commission (FERC) in the U.S. have impemented a series of measures to mitigate market power in restructured markets with generay satisfactory resuts, by essentiay outawing any form of price manipuation even if it is technicay achieved through actions that are not expicity prohibited by the tariff. In other words, any price manipuation that woud have not been rationa for a price taker is deemed iega. 6 In contrast, spot markets coordinated through cost-based dispatch and pricing mechanisms do, by design, eiminate most of the possibiities that exist in bid-based designs for generators to exercise uniatera market power. In cost-based markets, fue costs and technica parameters are audited by the SO and generators have a must offer obigation, which diminishes (if not eiminates) the abiity of generation firms to withhod capacity or to raise prices above margina costs. 7 However, contrary to what many reguators beieve, forcing the spot market to operate based on short-run 5. A suppier is said to be pivota if its capacity is needed to meet demand at a specific ocation and time. 6. Recent actions by J.P. Morgan, abusing the Bid Cost Recovery rues in Caifornia (FERC, 2013), and by Morgan Staney and KeySpan in New York, gaming the instaed capacity market (DOJ, 2011), are exampes of such behavior. 7. This is ony true if the SO or reguator performs periodic revisions to cost and technica parameters of a generators in the system. However, if audits are carried out ony sporadicay, generation firms might have incentives to report higher fue costs, higher technica imits, or ower eves of avaiabe capacity than the actua ones. Furthermore, some generators may have incentives to report higher than actua forced or other outage rates, which are hardy verifiabe by the reguator, eading to the sick day probem (Harvey et a., 2004). Copyright 2018 by the IAEE. A rights reserved.

6 56 / The Energy Journa fue costs can affect the production efficiency of a market, even if prices are equa to the margina cost of generation for reasons we now expain. The undesirabe effect upon production efficiency of forcing prices to be equa to the theoreticay competitive eves were first noticed in the era of reguated monopoies. Averch & Johnson (1962) demonstrated that monopoies facing a rate-of-return reguation have incentives to overinvest in capita with respect to the wefare maximizing eves in order to increase their profits, if the aowed rate of return exceeds the market s cost of capita. Currenty, many reguated utiities operate under price-cap incentives, which eiminates the Averch-Johnson effect by imiting revenues rather than the return that firms get on invested capita (Braeutigam & Panzar, 1993). Joskow (1974) notes that there can aso be a reverse Averch-Johnson effect if rates of return are hed beow market returns to capita whie fue or other operating expenses are subject to pass-through provisions, so that utiities overspend on operating costs and underinvest in capita. Simiary, in cost-based eectricity markets, firms do not necessariy have incentives to invest in the portfoio of generation technoogies that minimizes cost. This issue was first demonstrated in a more genera context by Kreps & Scheinkman (1983). The authors show that when firms first seect production capacities and then ater engage in price competition when determining shortrun operating eves (Nash-Bertrand strategy), the equiibrium of that cosed-oop game yieds the same resuts of an open-oop Cournot mode. In an open-oop Cournot mode, firms seect investment eves and production quantities simutaneousy (rather than in sequence as in the cosed-oop case), anaogous to a system where a production is sod in ong-term contracts and where there is no spot market (Murphy & Smeers, 2005). Areano & Serra (2007) extend Kreps & Scheinkman s resut to cost-based market designs using a simpe cosed-oop mode with ony two generation technoogies, peaking and baseoad. The authors assume that a generation firms first seect investment eves simutaneousy; the dispatch and pricing of eectricity is done in a ater period based on audited short-run fue costs by an independent SO. They find that under this form of market design, generation firms have incentives to increase the share of the peaking technoogy beyond the eve that maximizes socia wefare. By doing so, firms can increase the fraction of time the peaking technoogy sets the price, which increases the economic profits they receive from generating eectricity using their baseoad capacity. That soution is not, in genera, the same as the open-oop Cournot mode. Consequenty, the theory shows that if reguators with their best intention try to mitigate market power by forcing prices to be equa to short-run margina cost, producers wi have incentives to invest in inefficient generation portfoios, which woud utimatey reduce socia wefare just ike the exercise of short-run market power in bid-based markets. However, those resuts do not necessariy impy that it is more desirabe to aow the exercise of market power in the short-run by removing restrictions on bidding. The obvious questions therefore remain: What market design is more efficient in the ong-run then? A bid-based market, where generation firms can bid strategicay, or a cost-based design, where firms might have strong incentives to deviate from the wefare maximizing generation portfoio? Unfortunatey, there is no unique answer to these questions. Fershtman & Kamien (1987) show that the equiibrium of a cosed-oop game, where firms commit to capacities first and are free to choose productions eves ater anaogous to a bid-based design is coser to the competitive eve than the one for the standard open-oop Cournot game. Murphy & Smeers (2005) reach a simiar concusion for investment and spot eectricity markets with imperfect competition. The resuting prices and investment eves of the cosed-oop equiibrium eves ie between the perfecty competitive eves and the ones observed in the open-oop Cournot equiibrium. Wogrin et a. (2013) compare cosed-oop investment A rights reserved. Copyright 2018 by the IAEE.

7 Economic Inefficiencies of Cost-based Eectricity Market Designs / 57 modes with a generaized production stage using conjectura variations. They find that the rank ordering of cosed-oop equiibria in terms of market efficiency is ambiguous and parameter dependent. Thus, the resuts of Fershtman & Kamien (1987), Murphy & Smeers (2005), and Wogrin et a. (2013) suggest that forcing eectricity prices to be equa to short-run margina fue costs can be more harmfu to the ong-run economic efficiency of the system than etting generators exercise market power in bid-based market designs. Cramton (2004) even goes one step further and states that the exercise of market power is a desirabe and necessary condition to incentivize new entry. Whie we do not go as far as Cramton (2004), with the next simpe iustrative exampe we show that forcing prices to be equa to short-run margina cost can ead to ower market efficiency than a bid-based market. However, these resuts are not genera and, depending on the parameters, the opposite can occur. 3.2 Market power in cost-based markets: A simpe numerica exampe Consider a simpe system with two oad periods L { peak, base} = that represent operation in a representative year. The parameter T denotes the duration of each period, which we define as T peak=3,760 [hours/year] and T base =5,000 [hours/year]. Consumers choose their demand eves d [MW] depending on the price of eectricity p [$/MWh] according to a inear demand curve d( p) = IC DSp (the inverse demand curve being p( d) = IC / DS d / DS), where the demand intercept is IC peak= 2,000 [MW] and IC base=1,400 [MW] and the demand sope is DS peak= IC peak/250 and DS base= IC base/200, resuting in price intercepts of 250 and 200 [$/MWh], respectivey. There are two competing generation firms i I = { 1, 2} in the market that seect generation capacities x i [MW] and generation dispatch eves q i [MW] in order to maximize their profits. For simpicity we assume that there is ony one type of technoogy avaiabe for investment and operation for both firms. Its annuaized capita cost is K =46,000 [$/MW-year] and its margina cost is C =11.8 [$/MWh]. Consider first the perfecty competitive assumption, where both firms are price takers. As we discussed earier, the equiibrium under such idea conditions is equivaent to the investments and dispatch eves that a centra panner with perfect information woud seect trying to maximize socia wefare (Samueson, 1952; Mas-Coe et a., 1995). We find the perfecty competitive equiibrium soving the foowing optimization probem: max IC T d 1 d C q K x (1) 2 i i d, qi, xi L DS 2DS i I i I s.t. qi = d L (2) q i i x L i d, qi, xi 0 i I, L (4) In (1) we maximize socia wefare subject to suppy-demand baance (2), maximum generation imits (3), and nonnegativity (4). Next, we consider two variations of the cosed-oop equiibrium modes proposed in Wogrin et a. (2013): a cost- and a bid-based market. In both modes generation firms first (upper eve) seect generation investments simutaneousy and ater participate in a spot market (ower eve) where the dispatch and pricing of eectricity is based on either audited costs (cost-based market) or on bids (bid-based market). We emuate a bid-based market using a (3) Copyright 2018 by the IAEE. A rights reserved.

8 58 / The Energy Journa Cournot mode, since there is empirica evidence that this setting can, approximatey, capture strategic firm behavior in bid-based markets (Bushne et a., 2008; Puer, 2007). In the ower eve (spot market), given in (5), generation capacities are fixed, thus, each firm maximizes its profits subject to the (now fixed) generation capacities x i seected in the upper-eve probem: ( ) max T p C qi p, q i i I st.. qi xi ( λi ) L qi 0 L where in the competitive mode, p is viewed as a fixed parameter by firm i, who ony optimizes over q i, whie in the Cournot (unconstrained bidding) mode, p = p ( qi, q i ) = p qj, and the j I firm optimizes over q i and p. The variabe λ i represents the Lagrange mutipier of the capacity constraint. In the cost-based market, the equiibrium of the ower-eve probem is equivaent to a perfecty competitive one, which means that generation firms are price takers and the conjectured p ( qi, q i ) price response is = 0, i I, L. The Karush-Kuhn-Tucker (KKT) conditions of qi the ower-eve cost-based market equiibrium probem in compementarity form are the foowing: 0 qi T p T C λi 0, i I, L (6) 0 λi qi xi 0, i I, L (7) In contrast, in a bid-based market generation firms are aware that they can affect the equiibrium price since they recognize that they face a residua demand curve, i.e., price is infuenced by p ( qi, q i ) 1 their output so that = qi DS, i I, L. In a Cournot version of a bid-based mode, each firm i views its rivas production q i as fixed. The KKT conditions of the ower-eve bidbased market equiibrium probem in compementarity form are the foowing ones: (5) 0 qi 1 T p T T C λi 0, i I, L (8) DS 0 λi qi xi 0, i I, L (9) Consumers choose their consumption eves d trying to maximize their surpus: IC 1 max d d p d d DS DS s.t. d 0 L 2 2 (10) (11) The KKT conditions for consumers in compementarity form are as foows: 0 d IC DS 1 d p 0, L (12) DS A rights reserved. Copyright 2018 by the IAEE.

9 Economic Inefficiencies of Cost-based Eectricity Market Designs / 59 Tabe 1: Summary of resuts from three different market modes. Cost-based market Bid-based market Centra panner Investments per firm [MW] p peak [$/MWh] p base [$/MWh] Consumer surpus [Biion $] Tota profits [Biion $] Tota wefare [Biion $] The producer and consumer optimization probems in each spot market are inked together by the foowing market cearing condition: p free) d qi = 0, L (13) ( i I We refer to the ower-eve equiibrium (spot market) of the cost-based market mode as the market equiibrium under perfect competition (MEPC) probem. The soution of the MEPC is defined by the KKT conditions (6), (7), (12), and market cearing condition (13). Simiary, we refer to the ower-eve equiibrium of the bid-based market mode as the market equiibrium under Cournot-type competition (MECO) probem. The soution of the MECO is defined by the KKT conditions (8), (9), (12), and market cearing (13). We now compete the bieve cosed-oop mode by adding the upper-eve decision of capacity investment. In the upper eve, each firm maximizes tota profit (net of capita cost), subject to the short-run market equiibrium conditions (either MEPC or MECO). We formuate the bieve equiibrium modes as foows: ( ) max T p C qi K xi p, d, qi, xi i I s. t. MEPC / MECO. The probems for a i must be soved simutaneousy because the market cearing constraints (pus p ( qi, q i ) in the bid-based case) coupe the firms probems. Tabe 1 shows investments, equiibrium prices, consumer surpus, tota profits, and tota wefare for the three different equiibrium modes: a) centra panning, equivaent to a perfecty competitive market both in operations (MEPC) and investment, b) a duopoy cost-based spot market (MEPC) with two firms who make investments strategicay, and c) a duopoy bid-based spot market (MECO) invoving two firms behaving strategicay with respect to both investment and operations. By definition, if investments are centray panned and bid at short-run margina cost, the market attains the highest possibe eve of tota wefare, which is a captured by consumers. 8 The centra panner prices eectricity based on its margina cost, which yieds the exact amount of revenues needed to cover both the operation and capita costs, thereby making zero profits. 9 Scarcity prices that occur when generation is operating at capacity serve as a mechanism to ration demand and provide a return to capita. This equiibrium attains both aocative and production efficiency and it is equivaent to the outcome of a perfecty competitive market in both investments and oper- (14) 8. Note that we haven t considered the possibiity of Averch-Johnson-type production inefficiency effects. 9. The zero-profit resut occurs because of the inear capita and operating cost assumptions. Copyright 2018 by the IAEE. A rights reserved.

10 60 / The Energy Journa ations one with many price-taking generation firms, perfect information, and no barriers to entry (Samueson, 1952; Mas-Coe et a., 1995). Perhaps counterintuitivey, if we force the spot market to operate based on audited margina costs (cost-based market), this does not ead generation firms to seect the investment eves that woud be chosen in a perfecty competitive market or by a centra panner. Since generation firms are aware that the dispatch and pricing of eectricity is done based on their true margina costs in the spot market, they determine that it is optima to strategicay size their power pants beow the sociay-optima eves in order to incur scarcity prices more often. The net effect is a reduction of socia wefare with respect to the centray-panned soution. In this simpe exampe we find that a cost-based market design eads to ower market efficiency reative to the bid-based market. Aowing for the exertion of market power in the bids aows companies to incur higher profits, which in this particuar case aso eads to higher capacity investments (603 [MW] per firm in the bid-based market). When forced to bid margina cost, as in the cost-based market regime, capacity investments are not as profitabe for private companies and hence they invest ess (504 [MW] per firm in the cost-based market). Consequenty, it is not necessariy true that a cost-based market design yieds a better ong-run economic efficiency than a bid-based market, even if in the atter generators can behave strategicay in the short term. However, we want to highight that this resut is parameter dependent. In the onine Appendix A.1 we carry out a sensitivity anaysis for this numerica exampe in order to identify when one type of market design outperforms the other. 4. THE CHALLENGE OF AUDITING OPPORTUNITY COSTS IN COST-BASED MARKET DESIGNS Cost-based markets present an additiona chaenge that can create a series of production and aocation inefficiencies. In a cost-based market, the reguator or SO is responsibe for estimating or verifying both the direct and opportunity costs of a generators at a hours based on audited information in order to, ideay, set dispatch schedues and prices that resut in productive and aocative efficiency. However, if for some reason a generator is forced to se its power at a price equa to its directy attributabe margina cost, but that disregards a opportunity costs then a market faiure wi occur because consumers wi buy too much or too itte of the good in question (Mas-Coe et a., 1995). Production inefficiencies can occur if a generator s actua opportunity cost is understated (overstated) by its audited cost, so it wi be overused (underused) reative to other pants. We now discuss a series of settings where the opportunity costs of generators are non-zero, incuding dispatch with intertempora generation constraints, infexibe fue contracts, and tradabe emissions and renewabe permits. 4.1 Intertempora imits on starts, operating hours, and energy The operation of eectric generators is subject to intertempora constraints that give rise to opportunity costs that can greaty differ from audited fue-based variabe costs. An obvious exampe is hydroeectric power pants; even though water is free, discharging water from a reservoir today in order to generate eectricity in many cases means that ess water is avaiabe tomorrow to generate power then, so that the vaue of water is equa to the revenue that woud be earned if energy was A rights reserved. Copyright 2018 by the IAEE.

11 Economic Inefficiencies of Cost-based Eectricity Market Designs / 61 generated at the optima time ater on. This opportunity cost is widey recognized in power markets and is, for exampe, the basis of power pricing in the buk market in Brazi (Maceira et a., 2008). Another exampe whose importance has been recenty recognized by SOs is the foowing. Many peaking generators have restrictions on the number of starts per time interva; for instance, some generators can ony be started and shut down once per day because of therma stresses or crew considerations, whie for others, maintenance contracts often specify a maximum number of starts per season before a generator must be taken down for maintenance (Kumar et a., 2012). For instance, severa generators in the Caifornia market have such imitations (CAISO, 2016). As a resut, an opportunity cost arises as foows: if a generator is started today, it has one fewer start avaiabe over some reevant time horizon. Thus, the opportunity cost is the revenue that the generator woud earn if that incrementa start woud instead be used ater. For instance, consider a 30 [MW] peaking pant with a variabe cost of 70 [$/MWh]. A start might invove a fue cost of ony $1000, but if starting the generator now precudes starting it at a ater time or date when the eectricity prices are 200 [$/MWh] for a four-hour peak, the generator s true start-up cost is the foregone revenue 200 [$/MWh]*30 [MW]*4 [hours], or $24,000. Assuming that this price represents the margina vaue of power to the system, the SO shoud compare that foregone revenue with the revenue that woud be earned today in order to make a rationa decision about when to start the unit. It woud be economicay undesirabe for the system to burn a start today under, say, prices of 100 [$/MWh], if that prevents the generator from being used during that 200 [$/MWh] period. Recognizing this, the Caifornia SO now cacuates opportunity costs associated with monthy or seasona imitations on numbers of starts, number of operating hours, and tota energy production, and wi aow generators to submit those costs as the start-up cost instead of fue costs needed to start or run the unit (CAISO, 2016). 4.2 Natura gas contracts A arge number of countries rey on imports of iquefied natura gas to serve the domestic demand for the fue using take-or-pay contracts (Creti & Vieneuve, 2004). Unike an option contract that gives the hoder the option but no obigation to buy a commodity at a pre-arranged price, a take-or-pay agreement specifies both a price and a quantity at the time of deivery, pus a cause that forces the buyer to pay a fixed penaty to the seer if the quantity of fue taken is ess than specified in the contract (Masten & Crocker, 1985). Resuts simiar to those we are about to describe can aso occur in rea-time eectricity markets when short-term gas imbaance penaties make it costy to deviate too far from day-ahead gas deivery quantities; this can cause generator margina costs to be much greater or ess than day-ahead costs (or bids) (CAISO, 2014). These contracts create a series of difficuties for natura gas-fueed generators in power systems with arge shares of variabe and unpredictabe generation from renewabe energy resources. Since renewabe resources have, in genera, ower opportunity costs than natura gas, an increase in the avaiabiity of hydro, wind, or soar resources within a time window in a power system eads to a decrease in the dispatch of therma generators. Under such a scenario, the owner of a natura gas-fueed power pant that secured its fue through a take-or-pay agreement coud find himsef in a situation where the amount of procured fue is in excess of his actua needs. In that case, the appropriate margina cost of gas-fired generation is not the contracted price, but rather something much ess, depending on the size and form of the penaty. Copyright 2018 by the IAEE. A rights reserved.

12 62 / The Energy Journa Consider the optima dispatch of the foowing system for one hour with three avaiabe generation technoogies: 100 [MW] of hydropower (Hydro), 100 [MW] of coa (Coa), and 100 [MW] of combined-cyce gas turbines (CCGT). We assume that the opportunity cost of the water in the hydroeectric power pant is 5 [$/MWh] and the margina fue costs for the Coa generator is 30 [$/MWh]. The CCGT signed a take-or-pay contract for a fixed amount of natura gas, which resuts in a (onger run) margina fue cost of 50 [$/MWh]. We assume that a generators act as price takers and consider two scenarios of demand: high (150 [MW]) and ow (90 [MW]). In the scenario where there is no excess of natura gas, both cost-based and bid-based markets yied the same dispatch schedue and eectricity prices. In the high-demand scenario (150 [MW]), the eectricity price equas the margina fue cost of Coa, 30 [$/MWh], and the Hydro generator makes a positive profit. In the ow-demand scenario (90 [MW]) the eectricity price is equa to the opportunity cost of Hydro, 5 [$/MWh]. The schedues and prices derived in the cost-based market are Pareto efficient because the ony reevant costs for the CCGT are its fue expenditures based on the take-or-pay agreement, 50 [$/MWh], an amount that is easiy auditabe by the SO. 10 Let s consider now two scenarios where the CCGT faces an excess of natura gas committed in the take-or-pay agreement in a bid-based market. First, if the fue coud be re-sod by the buyer in a secondary market for the equivaent of 27 [$/MWh], this amount woud define the owest price the CCGT woud be wiing to receive for participating in this eectricity market rather than se the gas in the secondary one, i.e., its opportunity cost. In the high-demand scenario (150 [MW]) the CCGT woud dispace the generation from Coa and set the system s price, since the atter has now a higher margina cost than the opportunity cost of natura gas. This scenario woud not change the dispatch schedue or price in the ow-demand case (90 MW) compared to the situation in which there is no excess of natura gas. In the second scenario, assume now that there is no secondary market that woud take the excess of natura gas committed in the take-or-pay agreement. In this case, et s assume that the CCGT faces the equivaent of a $20 penaty for every [MWh] that is not deivered using natura gas. 11 This means that the CCGT woud be wiing to pay customers up to 20 [$/MWh] for taking its energy from natura gas. In a bid-based market, an offer for 20 [$/MWh] woud dispace Hydro and give the CCGT the first pace in the suppy curve. In the high-demand scenario Hydro woud set the eectricity price, since the CCGT woud be dispatched at its maximum capacity, 100 [MW]. However, in the ow-demand scenario the CCGT woud be the margina unit and the efficient eectricity price woud not equa its fue cost, but its opportunity cost equa to 20 [$/MWh]. These dispatch schedues and prices are Pareto-optima since they refect a reevant fue and opportunity costs. However, in a cost-based market the CCGT generator woud not be aowed to submit bids for its true opportunity cost that we showed in the two situations above, which woud ead too itte consumption of natura gas and too much of coa if the margina cost of the CCGT cannot be changed to a vaue other than its ong-term fue cost, 50 [$/MWh] If the generator was fuy using its take-or-pay contract and did not have the option of buying more than the contracted quantity at 50 [$/MWh] for the excess, then using the gas now woud mean that it coud not be used ater. This woud resut in a hydro-ike situation, were there woud be an opportunity cost equa to the revenue that the gas woud have earned in other periods. Our exampe here is much simper, since we assume that the power pant has a ong-term contract and aso an option of buying more gas than the contracted amount at a price equa to 50 [$/MWh]. 11. Note that venting natura gas off to the atmosphere is iega. Thus, if the procured fue is not needed, the buyer has no other aternative than to face the penaty stipuated in the take-or-pay agreement. 12. In Chie, for exampe, CCGTs facing an excess of natura gas can request the SO to give them a higher dispatch priority to avoid facing penaties for not taking the fue committed in ong-term contracts. In practice, the SO sets their margina costs in the dispatch optimization program to an administrative vaue equa to zero. Unfortunatey, this vaue does not A rights reserved. Copyright 2018 by the IAEE.

13 Economic Inefficiencies of Cost-based Eectricity Market Designs / Environmenta reguation and poicy incentives for renewabes More than 100 countries (REN21, 2015) and 49 states in the U.S. (DSIRE, 2015) have enacted some form of environmenta poicy to promote generation from renewabe energy resources or to reduce carbon emissions. Some of these incude Renewabe Portfoio Standards (RPSs), production tax credits (PTCs), feed-in tariffs, carbon taxes, and carbon cap-and-trade programs. RPSs, PTCs, and feed-in tariffs are poicies that aim at pricing into the market the positive externaities caused by the production of power from quaifying renewabe resources, whereas carbon taxes and cap-and-trade programs seek to price the negative externaities from the production of power from carbon-intensive generation. Since these poicies are equivaent to a combination of subsidies for renewabe energy and a tax on conventiona technoogies, they can create opportunity costs for both conventiona and renewabe generators participating in spot markets. In some extreme cases, renewabe generators might be wiing to pay customers to take their power. The whoesae market in Texas is one exampe where renewabe generators have submitted bids to suppy their power at negative prices for severa hours in a row, particuary at times when demand was ow, wind resources were abundant, and transmission capacity was scarce. According to Badick (2012), there were minute intervas where prices in west Texas, an area rich in wind resources, were beow zero in The observed negative bids paced by wind generators were simiar in magnitude to the production tax credit of approximatey 30 [$/MWh]. In this case wind generators behaved rationay, since an eectricity price of 29 [$/MWh] woud sti eave them with a 1 [$/MWh] profit, assuming negigibe variabe O&M costs. Since PTCs and RECs can ony be accrued if a renewabe resource is producing power, a competitive bid submitted by a price-taking wind generator shoud refect the foregone opportunity of coecting revenues from these reguatory instruments. In this case, negative prices are efficient and provide strong incentives for consumers to shift their consumption to those hours when renewabe resources are abundant. 13 Negative prices during certain hours aso disincentivize investments in more wind or soar capacity that wi be mosty avaiabe within that time window. Unfortunatey, in cost-based market designs, these opportunity costs are not considered as directy attributabe expenditures, such as fue or O&M costs, and must be internaized by generators, which resuts in dispatch schedues and prices that are economicay inefficient. In Chie, for instance, there is an RPS poicy with tradabe Renewabe Energy Certificates (RECs) and a carbon tax in pace. However, the opportunity costs that resut from the saes of RECs or the tax on carbon emissions are not considered in the dispatch and pricing of eectricity in rea time. The reguator and SO expects generators to, somehow, recover the additiona costs or profits that resut from these environmenta poicies through ong-term purchased power agreements. However, this mechanism creates a distortion between ong- and short-run eectricity prices that utimatey affects the production and aocative efficiency of the market. necessariy refect the opportunity cost of the CCGTs, which shoud account for the option of seing the gas in a secondary market or the penaty. Therefore, this rue eads to inefficient dispatch schedues and prices as we show in the exampe above. 13. Another view is that negative eectricity prices that resut from PTCs are harmfu to the economy, since they are subsidies from taxpayers to consumers of eectricity. However, this inefficiency is a direct consequence of the use of taxes and subsidies in a market economy and not a faiure of the eectricity market. In other words, if it is true that the production of one additiona MWh of power from a renewabe energy source wi resut in economic and environmenta benefits with a present worth of $30, then paying consumers up to 30 [$/MWh] to increase their consumption when renewabe resources are abundant is economicay efficient. However, in actua situations, the PTC can be very different from the margina socia benefit when prices are negative, and it can be more efficient to curtai renewabe production (Deng et a. 2015). Copyright 2018 by the IAEE. A rights reserved.

14 64 / The Energy Journa 5. CONCLUSIONS The major goa of a market designer (or reguator) shoud be to ensure that a market (or a reguated monopoy) functions in the most economicay efficient manner. However, this seemingy simpe goa might prove quite chaenging in practice since a market designs are imperfect due to information asymmetries and incentive compatibiity issues (Woak, 2003). In this paper, we discussed some of potentia inefficiencies of cost-based eectricity market designs, in which investments are dereguated but the dispatch and pricing of power is conducted based on audited cost information from private generation firms. Many countries in Latin America opted for these hybrid designs at the time of dereguation with the goa of preventing the exercise of market power in the spot market (Hammons et a., 2002) the type of strategic behavior that reguators and consumers seem most sensitive to. Esewhere, Ireand unti recenty had a simiar system, and severa U.S. markets aso require cost-based bidding when oca market power is shown to exist. However, up unti now, itte has been said with regards to the economic efficiency of such markets and how effective they are at reducing the incentives for generators to behave strategicay compared to bid-based markets, especiay concerning ong-term investment. Our main arguments are that cost-based spot market designs have two main features that make them inefficient. First, the exercise of market power is sti possibe in concentrated markets where there are barriers to entry, since firms have incentives to underinvest or to increase the share of the peaking technoogy, deviating from the sociay-optima generation portfoio. We use a simpe numerica exampe to show that the wefare oss due to the exercise of market power can be arger in a cost-based market than under a bid-based one. Thus, a bid-based design can be more efficient than a cost-based one even if firms can behave strategicay in the spot market. Nevertheess, the possibe exercise of market power in the short run in a bid-based system is a rea threat, particuary under stress conditions such as congestion, temporary ow market iquidity, and outages of generation resources and transmission ines. To guard against such market power, U.S. markets have created market monitoring departments for SO-based markets as we as active mitigation mechanism that aow the market monitor to substitute cost-based defaut energy bids (DEBs) for the actua submitted bids when certain market competitiveness tests are not met. These mitigation techniques fa into two genera categories: 1) Conduct and Impact Tests and 2) Structura Tests. The Conduct and Impact approach monitors if bids exhibit unusua deviations from some set norm and estimates the impact of such behavior on market prices, then triggering the DEBs if the impact exceeds a specified threshod. On the other hand, the Structura Test monitors the opportunity for the exercise of oca market power due to congestion and triggers the DEBs if, for instance, the number of pivota suppiers who can reieve congestion on a congested ine is three or ess. In such a case, the ine is deemed noncompetitive and a bids significanty impacting such a ine are repaced with DEBs. Market monitors aso survei ong-term bidding patterns and reegate to FERC cases against market participants whose bidding strategies suggest that they do not behave as rationa price-takers, as was the case with J.P. Morgan in Caifornia whose bidding behavior expoited the Make Whoe Payment mechanism and was fined $410M (FERC, 2013). Aggressive market monitoring and active mitigation have proven effective in suppressing market power abuse in the U.S. They are superior to enforcing cost-based bidding in a hours since, in genera, the market is competitive. Annua reports by the monitor for the Caifornia market indicate that bidbased market prices most of the time fa beow the simuated benchmark prices that are based on estimated cost pus 10%. Market competitiveness and the prevention of strategic withhoding in the day-ahead markets are aso reinforced by aowing Virtua Bidding which increases market iquidity A rights reserved. Copyright 2018 by the IAEE.