Using Virtual Bids to Manipulate the Value of Financial Transmission Rights

Shaun D. Ledgerwood is a Senior Consultant in The Brattle Group’s Washington, DC, office, with 23 years of experience as an attorney and economist. Dr. Ledgerwood is an expert in market competitiveness with an emphasis on the economic analysis of market manipulation claims in financial and commodities markets. He specializes in the analysis of competitive matters within and across physical and financial markets; issues pertinent to economic regulation, ratemaking, and resource planning; asset valuations; and analyses pursuant to matters in tort, contracts, or involving fraud. Dr. Ledgerwood holds a Ph.D., M.A., and B.A. in Economics from the University of Oklahoma and a J.D. from the University of Texas (Austin). Johannes P. Pfeifenberger, a Principal in The Brattle Group’s Cambridge, Massachusetts, office, is an economist with a background in electrical engineering and over 20 years of experience in the areas of regulatory economics and finance. Mr. Pfeifenberger specializes in energy and capacity market design, transmission and network access, ratemaking and incentive regulation, analysis and mitigation of market power, financial evaluation, and commercial litigation. He holds an M.A. in Economics and Finance from Brandeis University and an M.S. (Diplom Ingenieur) in Electrical Engineering and Energy Economics University of Technology from the University of Technology (Vienna).

November 2013, Vol. 26, Issue 9

Using Virtual Bids to Manipulate the Value of Financial Transmission Rights This article describes the economics of market manipulation using an applied electricity market construct: using virtual bids in conjunction with FTRs. The model was applied in the Federal Energy Regulatory Commission’s case against Constellation Energy Commodities Group and is applicable to a variety of cases alleging the use of uneconomic trading to trigger a manipulation, such as those alleged against JP Morgan, Deutsche Bank and Barclays. Shaun D. Ledgerwood and Johannes P. Pfeifenberger

I. Introduction The Energy Policy Act of 2005 (EPAct)1 provided the Federal Energy Regulatory Commission (FERC) with a fraud-based antimanipulation statute tied to the case precedent that underlies the SEC’s Rule 10b-5.2 Market manipulation ‘‘Rule 1c’’ was adopted by the Commission on Jan. 19, 20063 and gave FERC the ability to prohibit the use of ‘‘any

device, scheme, or artifice to defraud’’ the wholesale natural gas and electricity markets and to prosecute market participants who ‘‘engage in any act, practice, or course of business that operates or would operate as a fraud or deceit upon any [entity].’’4 Through its Office of Enforcement—which consists of the Divisions of Analytics and Surveillance, Audits, Energy Market Oversight, and

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

9

Investigations5—the Commission has the authority under EPAct to order the disgorgement of profits6 and to assess civil penalties up to $1 million per incident, per day.7 Given the risk involved, market participants rightfully seek guidance as to the types of behavior that may be considered manipulative under these rules. However, to ensure that robust trading survives in the future environment of anti-manipulation enforcement, it is perhaps even more important to identify the types of behavior that are clearly not manipulative. n the past year, FERC brought several enforcement cases against participants in the regional transmission organizations (RTOs)8 under its anti-manipulation authority.9 A settlement between FERC and Constellation Energy Commodities Group (CCG) demonstrated the magnitude of potential liability in such cases.10 The Commission levied $135 million in civil penalties and $110 million in disgorgement based on findings that CCG entered ‘‘into virtual transactions and [day-ahead] physical schedules without regard for their profitability, but with the intent of impacting [day-ahead] prices in the NYISO and ISO-NE to the benefit of certain significant [contract for differences] positions held by CCG.’’11 Key to this finding was a determination that the virtual trades placed by CCG were uneconomic, meaning that they caused divergence of the day-ahead and real-time prices and thus consistently lost

I

10

money on a stand-alone basis.12 The trades were allegedly intended to move day-ahead prices, which increased the value of CCG’s financial transmission rights (FTRs) and other swaps benefitted by the manipulation.13 Because the size of its FTR and swaps positions14 exceeded those of its unprofitable virtual and other trades, the resulting financial leverage allowed CCG to earn greater profits from its swaps than

Given the risk involved, market participants rightfully seek guidance as to the types of behavior that may be considered manipulative under these rules. it lost in its virtual and physical trades, netting CCG a profit from its alleged scheme. odeling the mechanics of the CCG case could provide valuable insight into FERC’s approach in determining whether specific trading behavior is or is not manipulative. To this end, we present an economic model that explains the incentives to manipulate the value of FTRs using virtual bids. This model shows the incentives that drive a trader’s stand-alone decision to place virtual bids at a node, the change in those incentives that results if the trader also holds a FTR position that sinks at that node, and the

M

profit-maximizing decision the trader may make to lose money on the virtual bids to increase the value of its FTR position and overall portfolio. We show that intentional uneconomic trading of virtual bids to trigger a manipulation diverges day-ahead and real-time nodal prices and thus creates market distortions and inefficiencies. The model therefore distinguishes between legitimate market participation that increases overall market efficiency and manipulative behavior that distorts markets and reduces efficiency, in a manner that is consistent with a general manipulation framework presented in another paper (Ledgerwood and Carpenter, 2012). These results are also consistent with the allegations made in several other market manipulation cases brought recently by the FERC, and so may provide guidance to compliance personnel on how to avoid potentially manipulative behavior. The remainder of this article consists of six sections. Section II explains the function of virtual bids within ‘‘Day 2’’ RTOs. Section III models the stand-alone incentives that underlie a trader’s decision to place virtual bids. Section IV explains the role and function of FTRs in Day 2 markets and discusses their potential manipulation through virtual bids. Section V then builds on the model of Section III to demonstrate a trader’s incentive to place an excessive quantity of virtual bids into the market to manipulate the value of its FTR position.

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

The Electricity Journal

Section VI identifies the market conditions that allow for successful manipulations to provide a basis for distinguishing manipulative behavior from legitimate trading. Section VII concludes.

II. The Function of Virtual Bids and Offers within Day 2 RTOs ‘‘Day 2’’ wholesale power markets use a market design in which power is competitively traded in an hourly ‘‘day-ahead’’ market designed to optimize supply to meet forecasted load and a ‘‘real-time’’ market wherein electricity is traded to balance the system for variances from the planned day-ahead dispatch. Power is traded at numerous ‘‘nodes’’ throughout each RTO system, which may be aggregated into trading hubs or zones. The price of power at each node (referred to as the ‘‘locational marginal price’’ or ‘‘LMP’’) varies depending on the system energy price, the cost of transmission losses incurred in serving the node, and the ‘‘congestion’’ cost created by transmission constraints encountered in supplying power at that node. Given the innate complexities associated with managing the physical operation and economic optimization of a transmission grid, significant and unexpected variability can emerge between the day-ahead and real-time LMPs at particular nodes as well as across nodes. Each RTO offers instruments designed to hedge November 2013, Vol. 26, Issue 9

against or profit from the risk associated with such price differentials. While some products such as FTRs are price-taking, others (including virtual bids) can influence electricity prices and thus be used for manipulative purposes (Celebi et al., 2010). raders participating in Day 2 wholesale electricity markets use virtual demand bids (also known as ‘‘virtual load,’’ ‘‘decremental bids,’’ or ‘‘DECs’’) and

T

Despite the trader’s perspective, characterization of the effect of virtual trades on the day-ahead and real-time markets as purely ‘financial’ is erroneous. virtual supply offers (also known as ‘‘virtual supply,’’ ‘‘incremental bids,’’ or ‘‘INCs’’) to profit from expected differences between day-ahead and real-time prices.15 Virtual trades simultaneously clear with bilateral and other physical trades and therefore will tend to impact day-ahead and (at times) real-time LMPs. The distinguishing characteristic of a virtual trade is that the quantity of megawatts (MW) bought or sold by the trader in the day-ahead is exactly offset by a sale or purchase of an identical quantity in the real-time, such that the net effect on the physical market quantity traded is zero. For

example, if in a given hour a trader expects the day-ahead LMP at a point to exceed the real-time LMP at that point, it would offer virtual supply (INCs) at that point for that hour. If the INCs clear, the trader receives the day-ahead price on all MW ‘‘sold’’ and pays the real-time price for the same number of MW ‘‘purchased,’’ netting a profit or loss equal to the difference of the day-ahead and real-time prices multiplied by the quantity of MW cleared. In essence, the trader uses the dayahead market to sell to itself in the future real-time market such that the net effect is that of a purely ‘‘financial’’ transaction. A. The physical and price impacts of virtual trading Despite the trader’s perspective, characterization of the effect of virtual trades on the day-ahead and real-time markets as purely ‘‘financial’’ is erroneous because they influence LMPs and can affect physical power flows. For example, in each hour and for each pricing point, a RTO could choose to use the higher of the total demand forecasted by the load-serving entities (LSEs) in the market, the total LSE forecasted loads plus the net virtual demand in the market (equal to cleared DECs less cleared INCs), or the load forecasted by the RTO.16 Adding virtual trades to the day-ahead load will therefore affect the day-ahead price and could affect the real-time price if the bids affect the RTO’s physical commitment of generating

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

11

[(Figure_1)TD$IG]

Figure 1: The Price Effect of Virtual Load on the Day-Ahead Market

units. To demonstrate this, consider the representation of the day-ahead market shown in Figure 1. For each hour, net virtual trades (VT) add to the demand forecast for load (DL) if the quantity of DECs exceeds the quantity of INCs. This raises the price in the day-ahead market from P0DA to P1DA and increases the amount of generation resources procured by the RTO from MW0 to MW1. Since these resources will be available to the real-time market, the virtual demand bids may also affect real-time prices.17 Because the placement of virtual trades can affect the dispatch of real-time physical capacity, it is therefore inaccurate to characterize their impact as solely financial in nature. otwithstanding this observation, there is a clear price effect which virtual trades exert on the day-ahead market. As Figure 1 shows, DECs tend to raise prices in the day-ahead market while INCs tend to lower day-ahead prices. The concave nature of the offer curve accentuates these pricing effects in power markets. Thus, as load

increases, each MW of successive clearing DEC bids will have an increasingly powerful effect on price. The reverse is true for successive INCs, as the first INC negates the last (most powerful) DEC and subsequent INCs negate demand with less and less price impact. An opposite price effect may occur in the real-time market when the offsetting physical leg of the virtual transaction is executed. B. The role of virtual trading in price convergence The main benefit cited in support of the use of virtual transactions derives from their

[(Figure_2)TD$IG]

tendency to equilibrate the prices between the day-ahead and realtime markets. To demonstrate this, consider a numeric scenario wherein a virtual trader expects the day-ahead and real-time LMPs at a node to be $30/MW and $70/MW, respectively. This is represented in Figure 2. The trader would place DECs that node for that hour, paying the day-ahead price on all MW purchased and receiving the realtime price for an identical number of MW sold. However, the act of buying tends to push the LMP up in the day-ahead market from $30, while the act of reversing the transaction in the real-time market can push the price down below $70. As long as a differential exists, some trader in the market will have the incentive to continue to place DECs until the day-ahead and real-time LMPs are equal. For this reason, virtual trading is referred to as ‘‘convergence bidding’’ as a competitive virtual market should tend to cause the day-ahead and realtime prices to converge in each hour.18

N

12

Figure 2: The Convergence Principle of Virtual Bidding

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

The Electricity Journal

B

ecause the convergence of day-ahead and real-time prices mitigates market power and improves the efficiency of serving load, virtual trades not only affect market prices but also have a physical impact on the operations of the RTO and upon market participants that physically transact at the LMPs set in the day-ahead and real-time markets.19 The act of creating convergence acts against the self-interest of virtual traders, as the revenue from all virtual trades will be zero if the day-ahead and real-time prices are equal. In the presence of transactions costs, traders should therefore be averse to placing excessive virtual bids or offers for fear of losses. The resulting corollary is that if a trader places virtual trades beyond the point of convergence, they will lose money on the entire lot. In that case, the trader would be diverging the dayahead and real-time LMPs, thereby creating inefficiency in the market. The choice by a trader to continually place virtual bids or offers in a manner that tends to lose money therefore raises concerns that the trader may be attempting to trigger a manipulation by moving LMPs to accentuate the value of price-taking financial positions tied to the market prices.20

in Figure 2 to model a trader’s stand-alone incentive to place DECs in a single hour at a single node. For every MW of DECs that clear, the trader will pay the dayahead LMP ‘‘PDA’’ and receive the real-time LMP ‘‘PRT’’ at that node for that hour. It is rational for the trader to bid its first DEC MW into the market only if it initially expects that P0RT > P0DA . Thus,

III. A Model of the Economic Decision to Place Virtual Bids

T

The following presentation generalizes the principles shown November 2013, Vol. 26, Issue 9

using Figure 2 as an example, if the trader expects the initial day-ahead price to be $30/MW and the initial real-time price to be $70/MW, a 1 MW DEC would garner about P0RT  P0DA ¼ $70  $30 ¼ $40.21 If the trader then bids more DECs into the market, all cleared bids will make money if PRT > PDA, a condition that Figure 2 shows grows less likely with each additional MW of DEC cleared because the dayahead and real-time prices then tend to converge. o understand the trader’s decision-making as to the quantity of DECs it should bid into the market to maximize profits on a stand-alone basis, our model initially assumes a single

trader bids DECs in the quantity of ‘‘X’’ MW at a single pricing location without any competition from other traders. Consistent with Figure 2, the model must reflect that the continual placement of DECs ultimately causes the day-ahead and real-time C prices to converge at PC DA ¼ PRT , thus yielding no profit to the trader. The trader therefore has an incentive to bid an amount of DECs designed to cause partial convergence, allowing it to profit from the remaining spread on all MW cleared. Following this discussion of a simplified market, we discuss the outcome that results if multiple traders are able to place virtual bids and offers into a competitive market. The resulting movement toward convergence produces the efficiency gains associated with virtual bidding, as the coordination of day-ahead and real-time pricing through statistical arbitrage simultaneously minimizes a source of uplift and neutralizes pricing variances caused by the use of market power. A. Profit maximization for a single trader of virtual bids Assume that a single trader can choose the quantity of DECs (in the amount of ‘‘X’’ MW) that it will bid in a single hour at a single node. Based on the logic of Figure 2, the trader can develop a derived demand curve for its DEC bids given the expected price spread (PS) between the real-time and day-ahead prices assuming different quantities of DECs clear

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

13

[(Figure_3)TD$IG]

Figure 3: The Derived Demand for Decremental Bids and Associated Revenues by a Single Trader (No Competition from Other Traders)

(PS = PRT  PDA). Given the inverse relationship between the spread paid to the DECs and the total quantity of DECs that clear, the demand for DECs is downward sloping, with a vertical intercept at a price equal to P0S ¼ P0RT  P0DA and a horizontal intercept at the quantity of DECs ‘‘Xmax’’ that creates price converC gence (PC DA ¼ PRT , thus C C PC S ¼ PRT  PDA ¼ 0). For simplicity, this demand function is represented as linear and is shown in Figure 3.22 he clearing of successive DECs (and resulting movement down the demand curve) has two countervailing effects on the trader’s revenues. Each additional MW of DEC successfully bid into the market brings added marginal revenue to the trader. However, the additional DEC sold simultaneously causes greater convergence between the day-ahead and real-time prices, thus reducing the revenues obtained across the entire lot of

T

14

DECs bid by the trader. The ‘‘Marginal Revenues of Virtual Trades’’ curve in Figure 3 illustrates this combined effect. Marginal revenues are positive up to X*, the point where the gain from the last DEC MW bid exactly offsets the loss caused by price convergence across the lot of all DECs bid previously. Beyond X*, the losses from convergence exceed the incremental value of each 1 MW sale, such that the trader’s marginal revenues are increasingly negative. Assuming transactions are costless, the total profit made by the trader is then the sum of the marginal revenues earned across all MW of DECs that clear, shown in Figure 3 by the curve titled ‘‘Total Revenues of Virtual Trades.’’ As the trader bids successive DECs up to X*, each additional MW that clears adds positive revenues to the trader’s book. Beyond X*, Total Revenues will decline at an increasing rate due to increasing losses caused by convergence.

Because the single trader does not face competition from other traders and is solely interested in maximizing the total revenues derived from its virtual trades, it will therefore bid DECs in the amount X* and earn TR* as a profit. Table 1 combines the information presented in Figures 2 and 3 in an algebraic construct. The trader is assumed to expect the pre-DEC spread between the realtime and day-ahead prices to be $40 and expects that convergence will occur if 40 MW of DEC clear. If the derived demand is linear, the trader’s demand for DECs is PS = 40  X, where ‘‘PS’’ is the difference in the day-ahead and real-time prices given the quantity of DECs that clear, ‘‘X’’ is the quantity of DEC MW bid by the trader that clear, and 40 is the initial spread corresponding to the vertical intercept P0RT  P0DA .23 Assuming the trader then bids DECs (which, by assumption, clear) in 5 MW increments, the calculations shown in Table 1 result. The trader maximizes total revenues by bidding 20 MW of DECs into the market and driving the day-ahead/real-time spread to $20 per MW, corresponding with the profit-maximizing quantity of bids X* and resulting price spread PS shown in Figure 3. B. Convergence in the presence of multiple traders of virtual bids The previous discussion assumes that only one trader bids DECs into the market, yielding a

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

The Electricity Journal

Table 1: Profit Maximization of Virtual Bids Placed by a Single Trader (No Competition from Other Traders). Total Revenue Location on Horizontal Axis Origin (0)

Maximized revenues (X*)

Convergence (Xmax)

Real-Time vs.

Quantity of

from DEC

Day-Ahead Price Spread (PS = 40  X)

DEC MW Cleared (X )

MW Cleared (TR = PS  X)

40

0

0

35

5

175

30 25

10 15

300 375

20 15

20 25

400 375

10

30

300

5 0

35 40

175 0

Bold values correspond to the trader’s profit maximizing quantity of DEC bids.

bid of X* MW of DECs that will not fully converge the day-ahead and real-time prices upon clearing (i.e., PS > 0). However, assuming sequential bidding,24 there is an incentive for another trader to enter the virtual market and bid DECs to take advantage of the remaining price spread associated with PS . In the context of Table 1, the second trader will observe that there remains a $20 spread in the market after the first trader’s bids clear. The second trader’s residual demand for DECs is then PS = 20  X, prompting it to maximize revenues by bidding 10 MW of DECs. If these clear, the total number of DECs in the market will then be 30 and the resulting spread will fall to $10. Continuing in this sequential bidding paradigm, other traders would then opportunistically bid for the remaining surplus, moving the market toward a total quantity of DEC trades equal to Xmax and bringing convergence of the day-ahead and November 2013, Vol. 26, Issue 9

real-time prices toward the transaction costs of virtual trading. This process typifies arguments of why virtual trades assist market efficiency. Because physical capital is not required to participate in the virtual market, a multitude of traders can participate and use virtual transactions to eliminate the market inefficiency that would exist if only physical entities or single virtual bidders were able to trade.

IV. Financial Transmission Rights and the Incentive to Influence Day-Ahead Prices FTRs allow customers to protect against the risk of congestiondriven price increases in the day-ahead market for transmission service in the RTOs. Congestion costs occur as the demand for scheduled power over a transmission path exceeds that path’s flow capabilities. For

example, if the transmission capacity going from Point A (the ‘‘source’’) to Point B (the ‘‘sink’’) is 500 MW, but the RTO seeks to send 600 MW of power from Point A to Point B when calling on the least-cost generators to serve load, congestion occurs on the path. This causes the congestion price at the source (CPSource DA ) to decline and the congestion price at the sink (CPSink DA ) to increase, raising the cost of serving Point B from Point A to a level at which the RTO can economically limit power flows to the available 500 MW of transmission capacity. By obtaining a FTR (of ‘‘F’’ MW in size) over the path from Point A to Point B, the FTR holder receives the difference of the congestion prices between the sink and Source source ðCPSink DA  CPDA Þ  F, allowing it to hedge against the congestion charges it would otherwise incur in the day-ahead market. FTRs were originally developed to give traditional utilities and other LSEs price certainty similar to that available to traditional vertically integrated utilities prior to the introduction of market-based congestion management. This practice continues, as FTRs (or auction revenue rights that are convertible into FTRs) are allocated to LSEs and to other entities that fund the construction of specific new transmission facilities in the Day 2 RTOs. lthough FTRs are used by LSEs as a hedge, they can also be purchased by any creditworthy entity seeking their financial attributes as a risk

A

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

15

management tool or a speculative investment. In this regard, FTRs are simply financial instruments executed as contracts for differences between the day-ahead prices at two locations. However, FTRs differ from other types of financial ‘‘swaps’’ because the physical constraints of the transmission grid limit the available quantity of FTRs. RTOs limit the quantity of available FTRs based on a simultaneous feasibility test performed across all potential transmission constraints, which limits the size of the net positions that can be held by market participants.25 By comparison, because financial swaps have no physical dimension, traders can accumulate swaps in quantities bounded only by the willingness of counterparties to take opposing financial positions. Another factor that can distinguish FTRs from other types of swaps is that FTR payments are based only upon the congestion component of dayahead LMPs, whereas swaps can be based upon the total LMP or any components thereof.26 ecause FTRs and other types of instruments designed to allow market participants to hedge congestion are price-taking in the day-ahead market, there is legitimate concern that market participants with FTR positions of sufficient size may have incentives to manipulate day-ahead market prices in an effort to benefit their FTR or other swaps positions (Celebi et al., at 20–23). The model in Section III demonstrates that virtual transactions can provide a tool for

B

16

executing such manipulations due to their price-making role in the day-ahead market. This is especially true if, as was alleged in the CCG case, there is evidence that the trader consistently executed virtual trades in a manner that was intentionally designed to be uneconomic– i.e., intended to lose money on a stand-alone basis. To understand the

incentives that drive such behavior, we next add a FTR position to the virtual trader’s portfolio to extend the analyses of Section III and examine the trader’s joint decision-making process.

V. The Effect of Placing DECs at the Sink of a Prevailing Flow FTR A prevailing flow FTR pays its holder the difference of the dayahead congestion charges at the FTR’s sink and source, thus hedging the holder against the cost of increased congestion at the sink (or, less likely but conceivably, a glut of supply at the source). Placing DECs at (or near) the

FTR’s sink tends to worsen congestion in the day-ahead market at that node, causing the dayahead congestion price differential to increase and the resulting value of the FTR to increase (Celebi et al., at 22–23). The effect of this phenomenon on a trader’s incentives can be captured by adapting the model presented in Section III to include a FTR position in the trader’s portfolio. We first consider the case of a ‘‘financially unleveraged’’ FTR, where the size of the FTR position is not sufficient to fully hedge the trader’s losses. This shows that the addition of an unleveraged FTR to the trader’s portfolio incents the trader to bid DECs in quantities that reduce its virtual trading profits on a stand-alone basis, but which simultaneously improve market efficiency while maximizing the value of the trader’s overall portfolio. We next evaluate the effect of increasing the trader’s FTR position to a size that is greater than is necessary to hedge the trader’s risk from its virtual trades. In this case, the trader is incented to bid more DECs into the market than is economically efficient, causing the day-ahead and real-time prices to diverge. We conclude the section by extending the model to consider a multi-trader environment. A. Bidding DECs at the sink of an unleveraged FTR position For simplicity, the relationship between the quantity of DECs cleared and the resulting effect on

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

The Electricity Journal

[(Figure_4)TD$IG]

Figure 4: The Total Revenues Derived from Placing DECs at the Sink of an Unleveraged FTR Position

the day-ahead price is assumed to be linear, as shown in Figure 4 by the line labeled ‘‘Enhanced Value of FTRs.’’27 he combined revenues from the FTR and the virtual trades are shown by the curve labeled ‘‘Total Revenues: Virtual Trades + FTR Profits.’’ This curve achieves a maximum at a level of DECs (X^) and a level of revenues (TR^) above the DECs (X*) and revenues (TR*) of a trader bidding only DECs without the benefit of a related FTR position. The ability of the trader to enhance the value of its FTR gives it the incentive to bid more DECs at the sink than it would absent the FTR. Strictly speaking, this behavior is consistent with the mechanism of a manipulation triggered by uneconomic trading because the trader intentionally forgoes revenue in the virtual market to accrue a greater benefit from the enhanced value of its FTR. However, there are three interrelated reasons why the behavior illustrated in this

T

November 2013, Vol. 26, Issue 9

example should not be seen as manipulative. irst, as Figure 4 demonstrates, the trader’s behavior yields a desirable result for the market. While X^ lies to the right of X*, it is below Xmax, the level of DECs needed to converge the day-ahead and real-time prices. As the goal of convergence bidding is for the market to supply an amount of DECs equal to Xmax, the trader can legitimately argue that any quantity of DECs it places such that X^  Xmax benefits the market, thus serving a legitimate purpose that should be above scrutiny for market manipulation. Second, while it is certainly true that an optimal quantity of DECs (X*) exists, there is no way for the trader to know with precision what that quantity is ex ante or even ex post the clearing of the day-ahead and real-time markets. Because its opportunity costs are unknown and unprovable, the trader can always assert that its actual DEC bid X^ represented its

F

best effort to find the optimal quantity of DECs X*. This depends heavily on a third aspect of the outcome shown in Figure 4: that the trader made money on its virtual transactions on a stand-alone basis. As we will discuss, the expected value of a virtual trade is zero in the presence of other traders operating in a competitive market. Thus, because the only available benchmark against which to gauge the trader’s opportunity costs is zero, any profitable virtual transaction must be considered legitimate if viewed on a stand-alone basis.28 B. Bidding DECs at the sink of a leveraged FTR position As the quantity of FTR megawatts exceeds and increases relative to the size of the DECs bid at the FTR sink, the trader can ‘‘leverage’’ its FTR gains against its virtual trading losses. This may incent the placement of more DECs than necessary to create market convergence. If the quantity of FTR MW (F) exceeds the quantity of DEC MW associated with Xmax, the trader may have the incentive to set X^ > Xmax such that gains on its FTR will more than offset the resulting losses incurred on its virtual trades. Specifically, the trader must own a larger FTR position than it places in DECs (of ‘‘X’’ MW) to profitably execute the manipulation such that the gains in its FTR position exceeds its losses on the virtual transactions (Ledgerwood and Carpenter at

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

17

[(Figure_5)TD$IG]

Figure 5: The Total Revenues Derived from Placing DECs at the Sink of a Leveraged FTR Position

39–42; Ledgerwood and Harris at 31–32). 29 he potential manipulation of a leveraged FTR position is shown above in Figure 5. If more than Xmax MW of DECs clear, the trader receives an increasingly negative price (PS ) such that the loss on the last DEC MW traded (X^) equals the marginal gain on the enhanced value of the FTR, thus maximizing the total revenues garnered from the combined position (TR^). The shaded region shows the losses the trader incurs on its virtual transactions that enhance the value of its FTR to FTR(X^). By definition, the trader’s losses on all cleared DECs are due to divergence of the dayahead and real-time prices such that the resulting day-ahead price is now higher than that in the real-time. Table 2 captures the uneconomic trading that can trigger the manipulation by showing the results for DECs bid and cleared in 10 MW increments. In addition to the assumptions used for Table 1, we assume that the trader

T

18

holds a 60 MW FTR position that sinks at the node where the DECs clear, that the change in the dayahead price at the sink is attributable solely to the change in congestion prices, and that the clearing of DECs affects neither real-time prices nor the dayahead congestion price at the FTR’s source.30 Beyond X*, every dollar the day-ahead price increases will increase the profits on the FTR position but will simultaneously reduce the profitability of the virtual trades. The trader now maximizes its revenues by bidding 50 MW of DECs into the market, causing the spread to diverge such that the day-ahead price exceeds the realtime price by $10/MW. he net profitability of the trader’s behavior is theoretically measurable from either an opportunity cost or accounting cost basis. The opportunity cost measure would compare the trader’s post-manipulation bid of 50 MW (X^) against the 20 MW of DECs it would bid if it wished to maximize the revenues from its

T

virtual trades on a stand-alone basis (X*). This would show that the trader loses $400  ($500) = $900 on its virtual trades to gain $3,000  $1,200 = $1,800 on its FTR, a net gain of $900 overall. However, because the true value of X* is realistically unobservable, this calculation is not possible. By comparison, measurement of the net profitability of the trader’s behavior based on its accounting costs is easily accomplished by comparing its post-manipulation bid of 50 MW (X^) against the 40 MW of DECs needed for market convergence (Xmax). Relative to the 40 MW of DECs at convergence, the trader’s potential manipulation loses $500 on its virtual trades to trigger a $3,600  $3,000 = $600 gain on its FTR, a net gain of $100. It is noteworthy that a manipulation triggered by uneconomic trading requires the trader to bid DECs into the market in excess of the amount required for convergence. While this can be accomplished using fixed-price bids placed above the expected dayahead market clearing price, the trader then risks the possibility that either the bids will not clear or that a regulator might inquire as to the trader’s motivation for seeking to buy at high prices. It is less complicated for the trader to place the DEC bids as a price taker, such that the volume is guaranteed to clear irrespective of the ultimate market price. The resulting increase in market demand will then tend only to increase day-ahead congestion,

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

The Electricity Journal

Table 2: Profit Maximization Using Uneconomic DECs to Enhance FTR Values. Real-Time vs.

Enhancement of

Day-Ahead Location on Horizontal Axis

Price Spread (PS = 40  X)

Quantity of DECs Cleared (X)

Total Revenue

FTR Value from

Total Profit

of Virtuals (TR = PS  X)

Cleared DEC MWs (FTR = 60  (40  PS))

from Scheme [TR + FTR]

Origin (0)

40 10

0 100

0 300

0 600

0 900

Max profit virtuals (X*)

20 10

20 30

400 300

1,200 1,800

1,600 2,100

0

40

0

2,400

2,400

10 20

50 60

500 1,200

3,000 3,600

2,500 2,400

30 40

70 80

2,100 3,200

4,200 4,800

2,100 1,600

50

90

4,500

5,400

900

60

100

6,000

6,000

0

Convergence (Xmax) Maximum revenues (X^)

TR intercept (Xmax + XFTR)

Bold values correspond to the trader’s profit maximizing quantity of DEC bids.

thus increasing the value of the trader’s FTR that sinks at that node. That a price-taking virtual trade can be executed for a pricemaking purpose to benefit a different price-taking position is a somewhat confounding mechanism to grasp, as the trader can then potentially bias the dayahead market price. While the presence of other virtual traders should theoretically mitigate this potential, this theory would not explain the behavior alleged in the CCG and other similar market manipulation cases before the FERC. C. FTR manipulation using virtual transactions in a multitrader environment As discussed above in Section III.B, the presence of multiple traders in an efficient market that includes virtual bidding should consistently tend to drive the November 2013, Vol. 26, Issue 9

day-ahead and real-time prices to convergence. This ostensibly should mitigate any attempts at market manipulation. For example, if other traders observe the price divergence caused by the manipulation shown in Figure 5, they have the incentive to offer virtual supply bids (INCs) into the market to profit from the relatively high real-time price. Assuming that sufficient liquidity exists, refined virtual bids and offers will continue until the expected difference between the day-ahead and real-time prices is zero. Thus, any manipulation attempt is futile since virtual trading by others will ultimately cause convergence at Xmax and will drive the value of the trader’s FTR position shown in Figure 5 to FTR (Xmax). owever, there are (at least) three factors that can prevent convergence from occurring given intentional uneconomic behavior. First, the complex

H

interaction of the physical, regulatory and economic dimensions of an electrical system provides asymmetric market opportunities for market participants to act strategically to create or exploit anomalous system constraints or seams. Such behavior then impinges the effective formation and use of liquidity to capitalize on the divergences created by the uneconomic trades. Second, asymmetric risk may accompany the transactions needed to counter the uneconomic trades. This is true of INCs offered to counter uneconomic DEC bids, for INCs may carry different transaction costs31 and face relatively unlimited exposure to variability in real-time prices.32 Third, the manipulator’s expected losses on its virtual trades are more than subsidized by the expected profits earned from its leveraged FTR position. This affords the manipulator a hedge against the risk of

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

19

errors that other would-be traders seeking to capitalize on the divergence do not enjoy.33 bsent sufficient mitigation to restore convergence, the intentional placement of uneconomic virtual trades causes inefficiency in the RTO’s markets. By purposely causing the day-ahead and real-time prices to diverge, the manipulator causes sub-optimal generation commitment and dispatch at the manipulated node(s) and across other nodes that are affected. The resulting costs must ultimately pass to the RTO’s wholesale power customers, who pay uplift charges for resources scheduled in the dayahead market but not used in the real-time or higher energy and ancillary services charges for the real-time dispatch of resources not scheduled in the day-ahead market. The manipulation also redistributes wealth to the detriment of the physical and financial market participants that are on the same side of the day-ahead and real-time markets as the manipulator’s virtual trades. For example, in the example shown in Figure 5, buyers in the day-ahead market will pay an uneconomically high price for power while sellers in the real-time market may receive an artificially low price due to the over-commitment of generation in the day-ahead market. Given these issues, RTO market monitors screen for uneconomic virtual transactions as part of their oversight function.34 However, because these screens are applied on an ad hoc, ex post basis, there is

A

20

concern that such analyses may result in ‘‘false positives’’ wherein legitimate trades that just happened to lose money are erroneously identified as manipulative.35 The risk of false positives can be minimized by identifying the market conditions that favor the successful use of uneconomic trading as a manipulation trigger and by better

distinguishing the characteristics that separate legitimate and manipulative trading. The model presented in Figure 5 is useful to assist these purposes.

VI. Market Conditions that Promote and Characteristics that Define Manipulative Behavior As Figures 4 and 5 suggest, the decision to intentionally trigger a manipulation through uneconomic trading is based on a cost-benefit analysis, where the expected cost of losses in the virtual trades is compared to the expected benefit derived from

gains on the FTR. This is not a static exercise because the costs and benefits of the manipulation are inextricably linked subject to diminishing returns. For analytical purposes, it is therefore helpful to think of a manipulation as having three interrelated components: a trigger, a nexus, and a target (Ledgerwood and Carpenter; Ledgerwood and Harris). The trigger begins the manipulation with an act designed to bias a market outcome (such as a price) to cause the manipulation to occur. This biased outcome is the nexus that links the manipulation’s cause and effect. This effect alters the worth of the manipulation’s target, which is valued using the nexus and thus influenced by the trigger. In the context of the potential manipulation shown in Figure 5, the uneconomic virtual trades (the trigger) increase the congestion component of the nodal LMP (the nexus) to benefit the value of the trader’s FTR (the target). This framework is useful to evaluate each of the three components of the manipulation while conceptually maintaining the causal nexus between marginal losses in the trigger and marginal gains in the target. A. Market conditions that improve the likelihood of a successful manipulation Figure 5 shows that the trader’s goal is to maximize the value of its combined virtual/FTR portfolio. It at first seems obvious that factors which increase the trader’s benefits or reduce its costs in

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

The Electricity Journal

aggregate must tend to potentiate the manipulation, but this assertion requires careful consideration. Consider for example a change in the costs of the trigger. A reduction in the transactions costs of placing and clearing virtual trades would make their use more appealing for manipulation, but would also aid their legitimate use by other traders to arbitrage attempted price manipulative movements. Likewise, a reduction in the market-based loss used to trigger the manipulation (i.e., the divergence of the real-time and day-ahead prices) reduces its cost but could dampen the price movement needed to make the manipulation possible.36 Another factor that can reduce the cost of the trigger is if other market participants unwittingly assist the directional price movement by panicking in reaction to the manipulated price.37 This is a key reason why the price movement caused by uneconomic trading should be thought of as a type of fraud (Ledgerwood and Carpenter). ncreased leverage in the manipulation’s target will unambiguously tend to increase the likelihood of the manipulation’s success. In the context of Figures 4 and 5, this is because an increase in the size of the FTR target improves the benefit derived by the trader from movement in the nexus as each successive DEC MW clears. From a broader perspective, it follows that markets that facilitate traders in the accumulation of large, price-taking positions provide

I

November 2013, Vol. 26, Issue 9

greater potential for manipulation. Day 2 electricity markets provide an excellent example of this. As alleged by the FERC, CCG used combinations of financial derivatives, FTRs, and physical electricity trades priced at index to accumulate the financial leverage needed to manipulate power prices across four RTOs in the eastern interconnect, using a

combination of losing physical and virtual trades to trigger its scheme. The scale of the manipulation was made possible only because profits from the leveraged positions CCG had in place system-wide continually financed the losing transactions that perpetuated its alleged scheme. he strength of nexus is a direct function of the inelasticity of market supply and demand. The logic of this is clear, for more vertically sloped curves will generate a greater price reaction for any given change in the market quantity traded. This suggests why energy markets are particularly susceptible to market manipulation given frequent or episodic periods of inelasticity of

T

demand and supply arising during price formation. This is especially true of electricity markets, wherein lack of storage mandates that supply must be prepared in advance to meet an expected (but highly variable) demand and the quantity of the good produced must be constantly balanced with load given a litany of economic and physical constraints. These market conditions favor acts designed to trigger directional price movements through concentrated uneconomic trading designed to distort market prices. Given access to cheap triggers through virtual trades and an array of price-taking instruments that can be leveraged to accrue gains in excess of losses, Day 2 and other indexed energy and commodities markets will continue to provide opportunities for gaming unless meaningful and adaptive standards are developed to distinguish intentional uneconomic trading from the ‘‘noise’’ of losing transactions that are an otherwise normal outcome for roughly half of all competitive, profit-seeking trades (Hellwig at 493). B. The characteristics of legitimate versus manipulative virtual trading Based upon a casual interpretation of our presentation in Section V, the reader might erroneously assume that we are proposing to define manipulative virtual trading by the existence of a loss in any given hour. This is not so. Absent an extraordinarily

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

21

anomalous event,38 evidence that a trader lost money on one bid placed at one node in one hour provides little meaningful evidence of manipulation, irrespective of how that trade affected the trader’s FTRs or other price-taking positions. Indeed, if the virtual market is competitive, the trader should lose money on about half of its virtual trades placed over time because the mean value of those trades should trend to zero given market convergence, the law of large numbers, and the central limit theorem. To properly define the manipulative placement of virtual transactions, it is therefore necessary to prove that a trader frequently and consistently incurred losses on these trades over the course of time such that a presumption of legitimacy of the trades is brought into question.39 Even if such anomalous losses are demonstrated, the proof of a manipulation requires further showing that the trades in question were used to benefit financially leveraged positions through a nexus sufficiently causal to link the manipulative trigger to the targeted positions. evertheless, virtual trades provide a ‘‘bright line’’ standard against which to evaluate legitimate versus manipulative trades because the expected value of a virtual trade is typically zero. Thus, while a fixed-size DEC placed at one location in any single hour may earn a significant profit or loss, placement of that DEC over time should result in offsetting gains and losses such that the trader’s cumulative profit

N

22

would often net to zero under competitive market conditions. Certainly, if a trader is discovered to consistently lose money on the virtuals it places over time, this opens the possibility that it is strategically placing the virtual trades in excess of the level needed for convergence to intentionally bias day-ahead prices to benefit its related positions. While

it is conceivable that there may be a stand-alone legitimate business purpose for incurring such losses over time, the burden shifts to the trader to demonstrate why this is so. y way of contradiction, the characteristics of manipulative trading are equally useful in identifying the characteristics of legitimate trading. For example, if the trader can demonstrate that positive or zero profits were made on its virtual trades over time, concerns as to whether the behavior was manipulative are irrelevant because the outcome enhanced the efficiency of the market by converging the dayahead and real-time prices. Even if suspicious losses are shown, the

B

trader can refute a manipulation claim by demonstrating that its net price-taking positions exposed to the manipulated price were held in financially unleveraged quantities, for the manipulation cannot be profitable if the alleged target is sufficient only to hedge positional risk. Finally, the presence or absence of the nexus between different pricing relationships could be used to establish a defense. For example, a trader could refute an assertion that it held a financially leveraged FTR position by showing that the congestion component of the dayahead LMP moved less than the total LMP, thus making the size of the FTR position too small to support a profitable manipulation. Likewise, a suspected financially leveraged FTR position held at one location could be proven to be only a hedge if the impact of FTR or other price-taking positions held at other related locations is considered.

VII. Conclusion Virtual bidding provides a mechanism to improve market efficiency by adding liquidity to the market and thus incenting convergence between the dayahead and real-time markets. Virtual trades can also provide a cheap and effective vehicle to trigger a manipulation if executed by financially leveraged market participants, particularly at locations and during periods when constraints are binding and market liquidity is limited. The model

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

The Electricity Journal

we discussed herein provides the logical economic foundation for such behavior and thus defines a way to consistently and logically evaluate uneconomic trading generally and virtual trading specifically. Consistency and transparency of enforcement will clarify the behavior that is and is not considered manipulative, thereby identifying trading ‘‘safe harbors’’ to simplify the compliance requirements expected of traders and reduce the uncertainty and costs of market participation. This will improve market liquidity, reduce regulatory risk and mitigate the likelihood of successful manipulations. oncerns that explaining the workings of a loss-based manipulation as we have herein will lead to the deployment of overly-aggressive screening methodologies that produce false positives. The concepts suggested herein do not focus on isolated incidents of unlucky or riskaverse trading, but on uneconomic behavior that persists with such repetition or magnitude that a rational trader would avoid the loss but for the existence of some benefiting physical or financial target. Clarification of the behavior that is and, perhaps more importantly, that is not manipulative will help improve screens for market manipulation, reduce the likelihood of false positives and false negatives and, importantly, reduce the cost of future compliance and enforcement efforts through a better understanding of the behavior prohibited.&

Endnotes: 1. Energy Policy Act of 2005, Pub. L. No. 109–58, §§ 315, 1283, 119 Stat. 594, 691, 979–80. 2. The authority is based on 15 U.S.C. § 78j (2006). 3. Order No. 670, Prohibition of Energy Market Manipulation, FERC STATS. & REGS. ô 31,202, 71 Fed. Reg. 4,244 (2006) (codified at 18 C.F.R. pt. 1c). 4. Id. at p. 1.

11. Id., at p. 12. 12. Id., at pp. 2, 8, 9, 12, and 17. See also FERC Chairman Wellinghoff’s advice concerning the message sent by the stipulation: ‘‘do not trade uneconomically on one position in order to benefit the value of another.’’ Chairman Wellinghoff’s statement on the Constellation Energy Commodities Group Investigation, at http:// www.ferc.gov/media/statementsspeeches/wellinghoff/2012/03-15-12wellinghoff.asp (March 15, 2012). 13. Constellation, at pp. 5 and 15. 14. Id., at p. 6.

C

November 2013, Vol. 26, Issue 9

10. Constellation Energy Commodities Group, Inc., 138 FERC ô 61,168 (2012) (‘‘Constellation’’).

5. Office of Enforcement, FERC, at http:// www.ferc.gov/about/offices/oe/orgoe.asp. 6. Enforcement of Statutes, Order, Rules, and Regulations, 132 FERC ô 61,216 at p. 216 (2010). 7. 16 U.S.C. § 825o-1(b) (2006). 8. We use the term RTO to represent all types of entities tasked with operating regional transmission systems in furtherance of wholesale competition, including independent system operators (ISOs). 9. See Rumford Paper Company, 140 FERC ô 61,030 (2012); Lincoln Paper and Tissue, LLC, 140 FERC ô 61,031 (2012); Competitive Energy Services, LLC, 140 FERC ô 61,032 (2012); Richard Silkman, 140 FERC ô 61,033 (2012); Deutsche Bank Energy Trading, LLC, 141 FERC ô 61,084 (2012); Barclays Bank PLC, Daniel Brin, Scott Connelly, Karen Levine, and Ryan Smith, 141 FERC ô 61,084 (2012); and Gila River Power, LLC, 141 FERC ô 61,084 (2012).

15. Virtual transactions are allowed in all U.S. RTOs with a ‘‘Day 2’’ market design. Market participants can also place ‘‘synthetic’’ virtual trades using other types of transactions, such as scheduling to sell power into a region in the day-ahead market then cutting that schedule in the real-time. These ‘‘no-flow’’ transactions are functionally identical to virtual bids and are executable between RTOs and in markets that do not have a Day 2 market design. 16. Note that use of the maximum of these three calculations mandates that net short virtual positions (when the number of INCs cleared exceeds the number of cleared DECs) will not be considered in the day-ahead market. This is consistent with procuring a sufficient supply needed to preserve reliability in the market. 17. One could argue that because realtime dispatch is independent of all load forecasts, virtual trades should not affect real-time prices. However, because the RTO prepares physical generation for dispatch to meet virtual load, the failure of that load to materialize could logically serve only to decrease the real-time price unless the real-time market is operating in a perfectly elastic region of the offer curve. Whether this price effect manifests itself through the real-time LMP or through a different pricing mechanism (such as ‘‘uplift’’ charges) is irrelevant to this presentation, assuming such charges transparently

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

23

tie back to their associated virtual trades (Celebi et al. at 21). 18. See Order [G]ranting [sic] Motion for Extension of Time And Addressing Convergence Bidding Design Policy Filing (130 FERC ô 61,122 [Feb. 18, 2010]). Figure 2 shows that real-time prices fall as a result of the cleared DEC, which may or may not be the case given the RTOs’ method for dispatching its system. However, it is ultimately irrelevant to the analysis herein as to whether virtuals actually affect real-time prices because the placement of DECs will continue as long as a spread between the day-ahead and real-time prices exists; if real-time prices are unaffected by the virtuals placed in the example, the $40 spread will be reduced to zero at a price of $70 instead of a price between $30 and $70 as shown (Celebi et al. at 18–19).

intermittently due to portions of the day-ahead and real-time offer curves where blocks of generation are offered at a constant price. Subsequent analysis in Section V will consider the negative region of this demand curve beyond the point of convergence, wherein the trader’s decision to place DECs is based on a joint portfolio of virtual bids and a FTR. 23. Since Demand is given by PS = 40  X, Total Revenues (TR) = PS  X = (40  X) 

19. For further discussion of the merits of convergence bidding, see the California ISO’s filing Convergence Bidding Design Policy (FERC Docket No. ER10-300-000 [Nov. 20, 2009]). 20. A trader may have a legitimate purpose in its willingness to persistently take losses on virtual trades. For example, the owner of generation might be willing to consistently lose money on a DEC placed at a generation source to hedge the possibility of outages during peak periods. However, if the virtual market is competitive, other traders should see the profit opportunity afforded by the divergence and execute INCs to restore convergence to the market. 21. Note that this is the trader’s expected profit based upon its own private information set, which may or may not prove to be correct once the day-ahead and real-time markets actually clear. Indeed, it is possible that the trader could lose money on this transaction if its information is incorrect, making the trade appear ‘‘uneconomic’’ if viewed on a singular basis. The model we provide herein is designed to assist efforts to distinguish losses incurred through legitimate trading from those anomalously incurred to trigger a manipulation. 22. More realistically, the derived demand has a non-positive slope with horizontal segments occurring 24

X = 40X  X2 and Marginal Revenues (MR) = @TR/@X = 40  2X. 24. For simplicity, this presentation assumes that traders place bids sequentially. This is clearly not the case since the day-ahead and real-time markets follow an auction format. Auction theory would nevertheless predict an outcome equal to this result if competitive market conditions prevail, as is the pretext for allowing virtual bidding in the first place. 25. Note that participants in FTR auctions can buy ‘‘counter flow’’ capacity which offsets ‘‘prevailing flow’’ FTRs, thus allowing the value at risk on a given path to exceed the physical limits of the line. However, such bids are ultimately physically constrained, as the net position held on the path should always conform to the simultaneous feasibility test. 26. At a given node for a given hour, the total day-ahead LMP (PDA) consists of an energy component (equal to the marginal cost of the last generation unit used to serve load), a

loss component (equal to the marginal transmission losses incurred in serving the node), and the congestion price (CPDA). While FTRs tie exclusively to the latter component, swaps bear no such restrictions. See http://www.nodalexchange.com/ products_and_services/ overview.php. The distinction between the total LMP and its congestion component is relevant to the discussion of Sections V and VI concerning the manipulation of FTRs using virtual bids, because virtual bids are paid based on the total LMP whereas FTRs are paid based only on the congestion price. 27. Because the day-ahead offer curve is generally concave-up, every added DEC cleared will tend to increase the price at the sink, thus making this curve concave-up. While a linear curve is used in this example for simplicity, this presentation confirms that a concave-up curve would increase the financial leverage of the FTR position as more DECs are placed, thus increasing trader’s incentive to manipulate the market. Note also that the clearing of DECs may change the loss component of the total LMP in addition to increasing congestion. If so, the slope of the ‘‘Enhanced Value of FTRs’’ curve would need to be adjusted, but the analysis would otherwise change little as 0  ð@CPSink DA =@XÞ  ð@PDA =@XÞ. 28. Note that this distinguishes virtual transactions from most other trading instruments, for which a nonzero opportunity cost can be determined. The reader should be aware that trades that are profitable on an accounting basis but are proven to have intentionally failed to cover the trader’s opportunity cost can be and have been prosecuted for triggering manipulative schemes. See for example Energy Transfer Partners, L.P., et al., Order to Show Cause and Notice of Penalties, 120 FERC ô 61,086 (July 26, 2007); BP America Inc. et al., Order to Show Cause and Notice of Proposed Penalty, 144 FERC ô 61,100 (August 5, 2013). 29. If the marginal increase in the congestion price is smaller in magnitude than marginal decrease in the spread caused by the clearing of

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

The Electricity Journal

each successive MW of DEC, a larger FTR position is needed to give the manipulator sufficient financial leverage such that its gains from the enhanced FTR position exceed its losses accrued from the virtual trades. This would occur if either ð@CPSink DA =@XÞ < ð@PDA =@XÞ (e.g., if the DECs increased losses as well as congestion such that the change in the total LMP exceeds that of the congestion component) or (@PDA/ @X) < (@PS/@X) (i.e., the DECs lowered the real-time price such that the change in the day-ahead/real-time spread exceeds the change in the dayahead price). Special thanks to Matthew Hunter for his thoughts related to this issue. 30. If the real-time price is not affected by the placement of DECs, the closure of the spread is due entirely to an increase in the day-ahead price at the sink, which is attributable solely to increased congestion costs. This simplifies the calculation of the enhancement to the FTR’s value shown in Table 2, for the change in the day-ahead price must then equal the pre-DEC spread ($40) less the spread that results after the trader’s DECs clear (PS). 31. For example, cleared INCs can incur different uplift charges than cleared DECs do. For example, see California Independent System Operator Corporation: Order on Rehearing Requests, Compliance filing, Instituting Section 206 Proceeding and Establishing Refund Effective Date, 134 FERC ô 61,070 (2011), at pp. 44–57. 32. Specifically, whereas the minimum real-time price at which a cleared DEC bid sells power is typically constrained to zero, the maximum real-time price at which an INC offer must buy power is limited only by a capped value set by the RTO. 33. Of course, a purist could posit that the other traders should then acquire their own countervailing FTR positions to empower their abilities to combat the manipulator. This ignores that the capitalization requirements then needed to fund participation in the virtual market expand drastically such that market participation (and thus liquidity) dries up. Further, it

November 2013, Vol. 26, Issue 9

assumes that the other traders are willing to bear the potential cost of being accused of manipulation themselves. 34. For example, the Midwest ISO screens for uneconomic trades ‘‘generating losses greater than $50 per MWh’’ but rarely finds transactions of concern, noting ‘‘Virtual losses that warrant further investigation have been rare and only one pattern of trades warranted mitigation.’’ 2010 State of the Market Report for the MISO

36. To clarify this point, the nexus of the manipulation ties only to the dayahead price. Thus, the trader will benefit from a smaller price divergence only if it is caused by a constant realtime price such that the entirety of its losses on the virtual trades contributes to a change in the day-ahead price. 37. The 24-hour day-ahead auction format used in Day 2 RTOs would tend to prevent this type of behavior unless the manipulator can project a false pricing signal across days. However, in cases where the timing of trades is more fluid (such as when energy is traded over a settlement period to form an indexed price), this potential is much greater. For example, see Brian Hunter, 130 FERC ô 63,004 (2010), approved by the Commission in Brian Hunter, 135 FERC ô 61,054 (2011). 38. Specifically, a virtual bid or offer placed in size near the physical limits of the node and at a price sufficiently uneconomic so as to guarantee clearing could reasonably give rise to inquiries as to the trader’s motivation.

Electricity Markets, pp. 37–38. By comparison, PJM employs a ‘‘Forfeiture Rule’’ which requires automatic forfeiture of a FTR’s profits if virtual trades are placed in a proximity that could influence their value. See PJM Open Access Transmission Tariff, Attachment M – Appendix, Part VI at p. 1907 (Dec. 17, 2012). 35. Also concerning are ‘‘false negatives’’ where the screens fail to identify manipulative trades. However, because of the substantial litigation and reputational costs that may be incurred by market participants to defend against allegations of manipulation, it is preferable that some deference be given to the prevention of false positives in the interest of due process. The analysis of the next section seeks to simultaneously minimize both false positives and false negatives through better identification of the characteristics that define legitimate and manipulative trades.

39. For a loss-based manipulation to be effective, the trigger must impact the price-making mechanism such that the information conveyed is not filtered out as noise or snuffed out by the weak law of large numbers (Hellwig, at 493). To detect such behavior thus requires the ability to screen for anomalous losses against the backdrop of the noise of the market. Because such ‘‘anomalies’’ will also include legitimate transactions through which informed traders gather information, a presumption of legitimacy must attach to all open market trades as a null hypothesis. The burden then falls on the party alleging the manipulation to prove that the behavior in question rejected this hypothesis and thus fell outside of the level of market ‘‘noise’’ associated with legitimate trading. In the context of the uneconomic placement of virtual bids or offers, this burden requires proof that the accused trader anomalously placed excessive bids or offers to intentionally cause or attempt to cause directional price movements through a pattern of trading exhibited over time (Ledgerwood and Carpenter).

1040-6190/$–see front matter # 2013 Elsevier Inc. All rights reserved., http://dx.doi.org/10.1016/j.tej.2013.09.014

25