Black Oak Energy, LLC v. Federal Energy Regulatory Commission

United States Court of Appeals
FOR THE DISTRICT OF COLUMBIA CIRCUIT

Argued April 16, 2013 Decided August 6, 2013

No. 08-1386

BLACK OAK ENERGY, LLC, ET AL.,
PETITIONERS

v.

FEDERAL ENERGY REGULATORY COMMISSION,
RESPONDENT

CITY POWER MARKETING, LLC, ET AL.,
INTERVENORS

Consolidated with 11-1275, 12-1286

On Petitions for Review of Orders of
the Federal Energy Regulatory Commission

Catherine R. Connors argued the cause for petitioners.
With her on the briefs were Carol A. Smoots and Timothy R.
Schneider.

Samuel Soopper, Attorney, Federal Energy Regulatory
Commission, argued the cause for respondent. With him on
the brief were David L. Morenoff, Acting General Counsel, and
Robert H. Solomon, Solicitor. Beth G. Pacella, Attorney,
 2
Federal Energy Regulatory Commission, Robert V. Eckenrod,
Gary J. Newell, and Robert A. Weishaar Jr.

Before: GARLAND, Chief Judge, ROGERS and GRIFFITH,
Circuit Judges.

Opinion for the Court filed by Circuit Judge GRIFFITH.

GRIFFITH, Circuit Judge: By statute, the Federal Energy
Regulatory Commission (FERC) regulates trading in energy
markets. This case concerns the markets operated by PJM
Interconnection LLC, a Regional Transmission Organization
(RTO) 1 covering the East Coast, Appalachia, and parts of the
Midwest. Some PJM market participants are known as “virtual
marketers.” Unlike participants who actually traffic in
electricity, the virtual marketers never deliver or take delivery
of electricity; they trade in order to profit from price
fluctuations.

The petitioners and petitioner-intervenors in this case – all
virtual marketers – petition for review of two sets of FERC
orders. The first orders approved PJM’s method for disbursing
a monetary surplus that results from the way it operates its
markets. The virtual marketers do not receive any of this large
pool of money, but they believe they should. To that end, they
argue that FERC’s orders violate the Federal Power Act (FPA)
and the Administrative Procedure Act (APA). In Part I, we set
forth the facts relevant to this petition for review, and in Part II,
we set forth our reasons for denying it.

1
RTOs coordinate the transmission of electricity across a
geographic region. RTOs must be independent of any individual
market participant and must possess certain forms of control over
transmission of electricity in the region. See generally Regional
Transmission Organizations, 89 FERC ¶ 61,285 (1999).
 3
The second petition – also brought by a group of virtual
marketers, albeit a larger set of them – seeks review of FERC’s
orders requiring PJM to recoup money refunded to the virtual
marketers in connection with the administrative dispute over
the surplus. The petitioners and petitioner-intervenors argue
that these orders also violated the FPA and the APA. In Part III,
we set forth the facts relevant to this petition and explain our
reasons for remanding the orders in question to FERC for
reconsideration.

I

In the mid-1990s, federal electricity policy took a
competitive turn. Prior to that time, “utilities were vertically
integrated monopolies; electricity generation, transmission,
and distribution for a particular geographic area were generally
provided by and under the control of a single regulated entity.”
Midwest ISO Transmission Owners v. FERC, 373 F.3d 1361,
1363-64 (D.C. Cir. 2004). Since then, those vertical
monopolies have broken apart and, in many regions, systems
utilizing electricity trading markets have sprung up in their
place. Generators sit at one end of the regional transmission
process; at the other end sit local utility companies. The
markets help coordinate and allocate electricity from the
generators to the local utilities. The market operator involved
in this case, PJM Interconnection, LLC, is an RTO that uses
markets to determine pricing and to schedule the transmission
of electricity across the massive territory in which it operates.
See Atl. City Elec. Co. v. PJM Interconnection, LLC, 115
FERC ¶ 61,132, 61,473 (2006). PJM operates two markets
relevant to this portion of the case: a “Day-Ahead Market” and
a “Real-Time Market.”

The vast majority of electricity traded in the PJM markets
is traded in the Day-Ahead Market, in which traders bid on
 4
electricity to be transmitted the next day. See Black Oak
Energy, LLC v. PJM Interconnection, LLC, 125 FERC
¶ 61,042, 61,146 (2008). (Since electricity cannot be
effectively stored, delivery must be timely.) The Day-Ahead
market “derives a market-clearing price from the sellers’ and
buyers’ price and quantity indications for the next day; sales
are then made at the market-clearing price.” Edison Mission
Energy, Inc. v. FERC, 394 F.3d 964, 965 (D.C. Cir. 2005).
PJM then produces a transaction schedule in advance of actual
production and distribution. See FERC OFFICE OF
ENFORCEMENT, ENERGY PRIMER: A HANDBOOK OF ENERGY
BASICS 101 (2012) [hereinafter ENERGY PRIMER]. “The
day-ahead market allows market participants to . . . hedge
against price fluctuations that can occur in real-time” due to
problems such as generator outages, weather events, and
unforeseen congestion. Id.

Not all electricity is purchased in advance, however.
Various risk factors upset sellers’ and buyers’ projections of
supply and demand as manifested in the Day-Ahead schedule.
In PJM’s Real-Time Market, participants correct for these
changes by trading electricity at prices quoted for sale and
delivery within five-minute intervals. See id. at 102. PJM
calculates these prices based on grid operating conditions and
submitted bids. See id. PJM then coordinates the supply and
distribution chain “to meet the instantaneous demand for
electricity.” Id.

In the Day-Ahead and Real-Time markets, PJM calculates
prices according to the method of Locational Marginal Pricing
(LMP), which is used by electricity market operators across the
country. See Sacramento Mun. Util. Dist. v. Cal. Indep. Sys.
Operator Corp., 616 F.3d 520, 524-26 (D.C. Cir. 2010) (per
curiam); Wis. Pub. Power, Inc. v. FERC, 493 F.3d 239, 250-51
(D.C. Cir. 2007) (per curiam). Under LMP, the price any given
 5
buyer pays for electricity reflects a collection of costs attendant
to moving a megawatt of electricity through the system to a
buyer’s specific location on the grid.

As we have explained in the past,

[w]ith an LMP-based rate structure, prices are designed to
reflect the least-cost of meeting an incremental
megawatt-hour of demand at each location on the grid, and
thus prices vary based on location and time. [In an LMP
system, each price] consists of three components: (i) the
cost of generation; (ii) the cost of congestion; and (iii) the
cost of transmission losses.

Sacramento Mun. Util. Dist., 616 F.3d at 524 (citation
omitted). The cost of generation can be thought of as the
“baseline cost” of serving electricity (known in the industry as
“load”) to another location on the system in a hypothetical,
congestion-free environment. Id. Congestion, in turn, drives up
costs because it requires PJM to dispatch more expensive
generators to meet demand. See ENERGY PRIMER at 65. The
cost of congestion results in different prices at different nodes
of the system, depending on how congested the wires leading
to those nodes are. Wis. Pub. Power, Inc., 493 F.3d at 250-51.
At issue in this case is the cost of “transmission losses,” which
refers to “the amount of electric energy lost when electricity
flows across a transmission system . . . .” Sithe/Independence
Power Partners, L.P. v. FERC, 285 F.3d 1, 2 (D.C. Cir. 2002).
The losses are a function of “the amount of the current flowing
on the wire[,] . . . the resistance it encounters,” and the distance
it travels. Id. Thus, all else equal, at peak demand times, there
are higher losses, and at low demand times, there are lower
losses. PJM charges every buyer of electricity to cover these
transmission losses; we will call that charge the “transmission
loss component” of an LMP price.
 6

For our purposes, there are two relevant ways to calculate
the transmission loss component of an LMP price: “average
loss pricing” and “marginal loss pricing.” Whereas average
loss pricing charges buyers the average cost of transmission
losses, marginal loss pricing charges buyers the higher,
marginal cost of transmission losses. (Confusingly, marginal
loss pricing and LMP are not the same thing. Marginal loss
pricing is a method for calculating the transmission loss
component of LMP.) PJM, for a time, used average loss
pricing, but FERC eventually determined that the method
inequitably charged long-distance buyers too little, and
short-distance buyers too much. See Wis. Pub. Power, Inc., 493
F.3d at 252 (noting the problem with average loss pricing); Atl.
City Elec. Co., 115 FERC ¶ 61,132, 61,473-74 (noting that
PJM was using the average loss method). FERC therefore
ordered PJM to implement the marginal loss pricing method.
See id. at 61,478. Marginal loss pricing “recovers transmission
losses on a transaction-by-transaction basis by . . . treat[ing]
every transmission as if it were the last (marginal) transmission
on the system.” Wis. Pub. Power, Inc., 493 F.3d at 252. This
method charges each buyer for the last, most problematic load
transmission during any given time period.

Under the marginal loss method, the effect of losses on the
marginal cost of delivering energy is factored into the
energy price (i.e., the . . . LMP) at each location. Other
things being equal, customers near generation centers pay
prices that reflect smaller marginal loss costs while
customers far from generation centers pay prices that
reflect higher marginal loss costs. In addition, under the
marginal loss method (and unlike under the current
average loss system), PJM . . . consider[s] the effects of
losses in determining which generators to dispatch in
order to serve load at least cost.
 7

Atl. City Elec. Co., 115 FERC ¶ 61,132, 61,474. At the time of
its adoption, a commenter submitted that the systemic cost
savings of this method would amount to $100 million a year.
See id. at 61,478. 2

2
FERC has explained the effect of marginal loss pricing on
incentives in the following way:

When prices at each location reflect the full marginal cost of
delivery, (i.e., energy, congestion and losses), customers can
make efficient choices among suppliers at different locations.
The full marginal cost of delivering electricity to a customer at
one location includes the marginal cost of the losses in moving
the energy from the generator to the customer’s
location. . . . For example, if the marginal losses to deliver
energy from a remote generator to a customer at another
location are 10 percent, then in order to deliver 1 MWh to the
customer, the remote generator must produce 10 percent more,
or 1.1 MWh of energy. If the remote generator’s marginal cost
to produce 1 MWh is $50, then the marginal cost of delivering 1
MWh of energy to the customer is $55 (i.e., the marginal cost of
producing 1.1 MWh). Suppose that the customer could be
served with energy either from the remote generator or from a
local generator whose losses would be de minimus and whose
marginal production cost is $53/MWh. If the buyer fails to
consider, and is not required to pay for, losses, the remote
generator would appear to be cheaper, since its marginal
production cost (of $50/MWh) would be lower than the
$53/MWh marginal production cost of the nearby generator.
However, when marginal losses are considered, the nearby
generator would be the more efficient source. That is because
the marginal cost of delivering energy to the customer from the
nearby generator would be about the same as the marginal
production cost of $53/MWh (since losses would be de
minimus), while the full marginal cost to deliver energy from
the remote generator would be higher, i.e., $55/MWh. Thus, in
determining what supply sources can most efficiently serve
 8

But the marginal loss pricing scheme creates an
administrative challenge. Because “transmission losses
increase with the amount of current in the system, treating
every transmission as the marginal transmission produces
revenue in excess of actual losses . . . .” Wis. Pub. Power, Inc.,
493 F.3d at 252. That is, marginal losses always exceed
average losses. By charging everyone as if they were
responsible for the last, most problematic transmission on the
system, PJM ends up collecting more money – much more
money – than the amount it actually takes to cover the cost of
the transmission losses. The resulting surplus, a large pot of
money held by PJM, has no clear owner.

This case is a dispute over the obvious question: Who
should get the money? By statute, PJM takes the first crack at
an answer because it must file a tariff describing its rates and
terms of service, one component of which is its plan for
distributing the transmission loss surplus money. See 16 U.S.C.
§ 824d(c); Sithe/Independence Power Partners, L.P., 285 F.3d
at 4-5 (describing a surplus distribution system as an integral
part of a tariff). In turn, PJM’s tariffs are subject to approval by
FERC. See 16 U.S.C. § 824d(a). Though the parties do not
dispute FERC’s approval of the marginal loss pricing approach
itself, they dispute its approval of PJM’s system for
distributing the transmission loss surplus.

customers, the cost of marginal losses should be considered.
Failure to consider marginal losses – or to understate marginal
loss costs – can inefficiently inflate the total cost of serving
load.

Cal. Indep. Sys. Operator Corp. Pub. Util. Providing Serv. in Cal.
Under Sellers’ Choice Contracts, 107 FERC ¶ 61,274, 62,269
(2004) (emphasis omitted).
 9
In its communications with PJM about its tariff, FERC
was adamant about what PJM should not do when distributing
the surplus. FERC explained in multiple orders that PJM was
forbidden from using the money to “reimburse” market
participants for the initial transmission loss payments. See,
e.g., Black Oak Energy, LLC v. PJM Interconnection, LLC,
122 FERC ¶ 61,208, 62,184-85 (2008). Traders are smart:
when they know that their marginal loss payments are going to
be partially refunded, they will treat the LMP as a mere sticker
price that masks the true, post-rebate price of each trade,
distorting the incentives marginal loss pricing is supposed to
create. To some extent, any system that PJM adopts will alter
the incentives that traders face, but the more direct the relation
between the LMP price calculation and the surplus
disbursement calculation, the more completely the system will
erode LMP’s incentive structure. To prevent this, FERC
required PJM to divorce the surplus allocations from the
amount that market participants pay into the surplus in the first
place.

Along the road to marginal loss pricing, PJM identified
several methods for distributing the surplus while complying
with FERC’s “no reimbursements” constraint. See Atl. City
Elec. Co. v. PJM Interconnection, LLC, 117 FERC ¶ 61,169,
61,860-61 (2006). Eventually, FERC approved a system in
which the surplus would be allocated to market participants
based on the amount they pay for the fixed costs of the
transmission grid. See Black Oak Energy, LLC, 122 FERC
¶ 61,208, 62,185; Black Oak Energy, LLC, 125 FERC
¶ 61,042, 61,145-48. This system garnered the support of the
majority of PJM market participants, see Atl. City Elec. Co.,
117 FERC ¶ 61,169, 61,860, but also had its detractors, some
of whom filed the initial administrative complaint giving rise
to the orders at issue in this case. See generally FERC Docket
No. EL08-14 (Dec. 7, 2007).
 10

In Part II, we address the petition for review of the orders
that approve this system and deny requests for reconsideration
of the approval. (Collectively, we call them the “Surplus
Orders.”) The parties bringing the petition are a set of
electricity traders active on the PJM market. They are variously
referred to in the record and the briefs as “virtual marketers,”
“financial marketers,” and “arbitrageurs.” We use the term
“virtual marketers.” Whatever the name, the salient factor that
distinguishes them from all others who participate in the PJM
market is that they never actually transmit or take delivery of
electricity. Rather, their trades are offsetting: when they are
done trading, they neither owe, nor are they owed, any
electricity. Instead, they have either profited or lost based on
price fluctuations in the time between their purchases and their
sales. The virtual marketers pay none of the fixed costs of the
grid. 3 As a result, under the system FERC approved, the
virtual marketers receive no surplus allocation. They petition
for review of FERC’s orders approving that outcome. We deny
their petition for review. 4

II

The virtual marketers argue that FERC’s orders selecting a
transmission loss surplus allocation system violate 16 U.S.C.
§ 824d, which requires that “all rules and regulations affecting
or pertaining to . . . rates or charges shall be just and

3
This was not always the case. See discussion infra note 6 and
accompanying text.
4
Because we so hold, we need not address the petition for
review of FERC’s denial of the virtual marketers’ Second
Complaint, which concerned the scope of potential refunds. See
EPIC Merchant Energy NJ/PA, L.P. v. PJM Interconnection, LLC,
131 FERC ¶ 61,130 (2010).
 11
reasonable,” § 824d(a), and prohibits FERC from approving a
tariff that grants “undue preference or advantage to any person
or subject[s] any person to any undue prejudice or
disadvantage, or . . . maintain[s] any unreasonable difference in
rates . . . as between classes of service,” § 824d(b). In
reviewing each challenge, we apply the familiar arbitrary and
capricious standard to FERC’s actions. See Sacramento Mun.
Util. Dist., 616 F.3d at 528, 533-35 (applying the
arbitrary-and-capricious framework to § 824 review); W. Area
Power Admin. v. FERC, 525 F.3d 40, 57-58 (D.C. Cir. 2008)
(describing § 824 review as arbitrary-and-capricious review).
Under this “highly deferential” standard, see Sacramento Mun.
Util. Dist., 616 F.3d at 528 (citation omitted), we hold that the
Surplus Orders meet the requirements of § 824d.

A

The virtual marketers argue that FERC violated
§ 824d(a)’s requirement of “just and reasonable” rates because
the surplus allocation system the Commission selected runs
afoul of the “cost-causation principle.” That principle requires
that “all approved rates reflect to some degree the costs
actually caused by the customer who must pay them.” E. Ky.
Power Coop., Inc. v. FERC, 489 F.3d 1299, 1303 (D.C. Cir.
2007) (internal quotation marks omitted). The cost-causation
principle has its roots in monopoly rate regulation, where rates
are required to “be based on the costs of providing service . . .
plus a just and fair return on equity.” Ala. Elec. Coop. v. FERC,
684 F.2d 20, 27 (D.C. Cir. 1982). In the context of monopoly
regulation, this principle helps ensure that utilities “produce
revenues from each class of customers which match, as closely
as practicable, the costs to serve each class or individual
customer.” K N Energy, Inc. v. FERC, 968 F.2d 1295, 1300-01
(D.C. Cir. 1992) (internal quotation marks omitted) (emphasis
omitted). That is, we scrutinize a utility’s rates to ensure a
 12
match between cost-causation and cost-responsibility. In the
context of a market, we do the same, and our object of scrutiny
is the operator’s method of fixing a market price, coupled with
its system for disbursing any surpluses accumulated because of
the LMP method. See Sithe/Independence Power Partners,
L.P., 285 F.3d at 4-5 (holding that both aspects of the tariff are
subject to review).

Indeed, we have analyzed other market operators’ surplus
allocation schemes for compliance with the cost-causation
principle. See id.; Sacramento Mun. Util. Dist., 616 F.3d at
534-35. In Sithe, we held that FERC had failed to justify the
imposition of marginal loss pricing under that principle, but we
left the door open to clarification and explanation. See 285 F.3d
at 4-5. That explanation was forthcoming in Sacramento
Municipal Utility District (Sacramento). And though the
Sacramento court upheld a pro rata surplus allocation system,
616 F.3d at 535, whereas we are asked to review a system in
which the petitioners receive no share, the Sacramento court’s
reasoning still guides us here. Indeed, the reasoning offered in
that case demonstrates why the virtual marketers’
cost-causation challenge fails.

In Sacramento, the California Independent System
Operator proposed to distribute its transmission loss surplus
“to transmission customers on a pro rata basis by using those
revenues to uniformly reduce the cost of each megawatt-hour
purchased on the system.” Id. at 534 (citation omitted). In
concluding that this system complied with cost-causation
principles, the Sacramento court observed that it is impossible
to tease out causal responsibility for transmission losses in an
LMP-based market system at any given point in time. Who is
“causing” the first increment of current to flow through the
system? Who is “causing” the marginal increment to flow?
Since it is impossible to identify a “first” or a “marginal”
 13
increment, it is impossible to say who is causing which to flow.
As the Sacramento court held, “it is not possible to determine a
cost below marginal cost that any individual [customer] caused
as a result of that customer’s [demand for] electricity.” Id. at
534 (internal quotation marks omitted). Or, as we explained in
a similar context, “for purposes of marginal cost pricing, all
customers cause the incurrence of the costs associated with
coincident peak load, whether by adding or merely continuing
their usage.” Nat’l Ass’n of Regulatory Util. Comm’rs v.
FERC, 475 F.3d 1277, 1285 (D.C. Cir. 2007) (citing Town of
Norwood, Mass. v. FERC, 962 F.2d 20, 24 n.1 (D.C. Cir.
1992)). This means that any individual market participant
deserves no share of the surplus under cost-causation, as each
is equally the customer who “caused” the marginal
transmission loss. See Sacramento Mun. Util. Dist., 616 F.3d at
535 (“No customer is less deserving than another of being
treated as the marginal customer . . . .”). Because FERC is
treating the virtual marketers in this case “as the marginal
customer,” they are being treated consistently with
cost-causation principles.

The virtual marketers argue against the application of
Sacramento’s view of cost causation in this case. They explain
that they should not be treated “as the marginal customer”
because, as they put it, “[v]irtual transactions by definition are
purely financial and do not cause the physical flow of power
over transmission lines.” Pet’rs’ Br. 30. It is true that the virtual
marketers “submit bids for purely financial purchases or sales
of energy, which do not entail physical generation or
consumption of energy.” New York Indep. Sys. Operator, Inc.,
98 FERC ¶ 61,282, 62,216 (2002). But if physical activity were
the measure of cost causation, then PJM would not be allowed
to charge the virtual marketers at all, since they do not place
real demands on the transmission system.
 14
Of course, this would be preposterous. The virtual
marketers buy and sell contracts for electricity like all the other
market participants. Even though their trades are purely
financial, they depend on the existence of a market for actual
electricity. And their activities, though “virtual,” contribute to
the fluctuation of the market price, which in turn influences
whether load-serving entities (the technical name for market
participants who actually traffic in electricity) will purchase
electricity at a given time. Just as a wheat-trading arbitrageur
must trade wheat at the market price even though she does not
take delivery of the wheat, an electricity-trading arbitrageur
must trade electricity at the locational marginal price even
though she, in some sense, does not “cause the physical flow of
power over transmission lines.” Their trades must be treated as
if they impose costs on the system just like the trades of all
other participants. Sacramento established the principle that
each customer who pays a locational marginal price is equally
deserving of treatment as the marginal customer. Thus, each
customer is entitled to no set share of the resulting surplus. Just
as this principle applied to the transmission customers
petitioning for review in Sacramento, 616 F.3d at 534-35, it
applies to the virtual marketers in this case.

It must be noted that the petitioners in Sacramento were
dissatisfied with a pro rata share of the surplus, whereas the
petitioners here are dissatisfied with a zero share. 5 This puts
this case on different footing from Sacramento in some crucial
respects, requiring careful analysis of whether the surplus
allocation system unduly discriminates against the virtual
marketers.

5
Again, we note that the virtual marketers did not always
receive a zero share. See discussion infra note 6 and accompanying
text.
 15
B

There is no question that the surplus allocation system
selected by FERC discriminates against virtual marketers.
They receive none of the surplus, while the entities that pay the
fixed costs of the grid receive significant disbursements even
though, as a matter of cost causation, they do not deserve any
particular amount of surplus, either. The virtual marketers
argue that this discrimination is undue, in violation of
§ 824d(b). They also argue that FERC lacked substantial
evidence to back up its supposed justifications for approving
the discriminatory system, and that those justifications were
arbitrary and capricious.

We accept disparate treatment between ratepayers only if
FERC “offer[s] a valid reason for the disparity.” Electricity
Consumers Resource Council v. FERC, 747 F.2d 1511, 1515
(D.C. Cir. 1984) (per curiam) (internal quotation marks
omitted); see also Ark. Elec. Energy Consumers v. FERC, 290
F.3d 362, 367 (D.C. Cir. 2002) (“A rate is not unduly
preferential or unreasonably discriminatory if the utility can
justify the disparate effect.” (internal quotation marks
omitted)). FERC identifies valid reasons by pointing to
differences between parties that are relevant to the
achievement of permissible policy goals. See Transmission
Agency of N. Cal. v. FERC, 628 F.3d 538, 549 (D.C. Cir. 2010)
(“The court will not find a Commission determination to be
unduly discriminatory if the entity claiming discrimination is
not similarly situated to others.” (citation omitted)). In this
case, the Surplus Orders sufficiently justified the approval of a
discriminatory system on the grounds that virtual marketers
perform different roles from load-serving entities within the
market, and that the system will limit virtual marketers’
incentives to engage in market manipulation. Therefore, we
 16
hold that the Commission’s action did not run afoul of
§ 824d(b) or the APA.

FERC reasonably determined that the virtual marketers
are not similarly situated to the rest of PJM’s market
participants. The virtual marketers are distinguishable from
other market participants because “unlike load[-serving
entities], arbitrageurs balance each purchase transaction with a
sales transaction.” Black Oak Energy, LLC, 125 FERC
¶ 61,042, 61,145-46. That is, unlike entities that traffic in
electricity, the virtual marketers have a purely financial interest
in the markets. See Black Oak Energy, LLC, 122 FERC
¶ 61,208, 62,185. They do not participate as producers or
distributors of electricity, but rather as speculators and
risk-takers. Thus, they play a very different role within the
system than do load-serving entities. From FERC’s policy
perspective, the virtual marketers serve a useful purpose: they
spot and exploit inefficiencies, driving prices closer to an
accurate reflection of fundamental value. See, e.g., Black Oak
Energy, LLC, 125 FERC ¶ 61,042, 61,146 (stating that the
virtual marketers should “make transactions that reduce price
divergence between the Day-Ahead and Real-Time markets”).
This sets them apart from load-serving entities, and FERC
reasonably acts on this difference when it sets policy.

But their unique position within the marketplace animates
FERC’s concern over whether virtual marketers will have a
beneficial effect on the functioning of the markets. Since their
business interests are purely speculative, FERC explained, the
virtual marketers pose a threat as potential market
manipulators. FERC reasonably approved the surplus
allocation system because it promoted a policy of preventing
market manipulation of a certain stripe. We defer to FERC’s
policy priorities, so this explanation is adequate under arbitrary
and capricious review. See Alcoa Inc. v. FERC, 564 F.3d 1342,
 17
1347 (D.C. Cir. 2009) (“Issues of rate design are fairly
technical and, insofar as they are not technical, involve policy
judgments that lie at the core of the regulatory mission.”
(internal quotation marks and citations omitted)). As FERC
explained, any formula that disburses surplus to the virtual
marketers according to trading volume will create incentives
for them to focus on increasing their surplus disbursements by
increasing their trading volume. See Black Oak Energy, LLC,
122 FERC ¶ 61,208, 62,185. FERC put it this way:

Paying excess loss charges to [virtual marketers] . . . is
inconsistent with the concept of arbitrage itself. The
benefits of arbitrage are supposed to result from trading
acumen in being able to spot divergences between markets
. . . . If [virtual marketers] can profit from the volume of
their trades, they are not reacting only to perceived price
differentials in LMP or congestion, and may make trades
that would not be profitable based solely on price
differentials alone.

Id.; see also Black Oak Energy, LLC, 125 FERC ¶ 61,042,
61,145 n.46 (“[U]sing a pure load ratio share calculation would
provide an incentive for the arbitrageurs to conduct trades
simply to receive a larger [surplus allocation].”). That
increased trading could distort prices and destabilize the
electricity markets, and such activity would place the virtual
marketers far afield of their intended role within a competitive
energy system. FERC is well within its powers when it
promotes a policy of limiting market participants’ incentives to
speculate to the detriment of the efficient functioning of the
market.

The virtual marketers argue that FERC lacks substantial
evidence in the record to support its view that the system it
selected will help prevent market manipulation. True, FERC’s
 18
analysis is not based on retrospective data. But given the
circumstances, there is no way that it could be, because PJM
had not implemented the proposed system when FERC had to
act, and we defer to reasonable and cogent explanations of
predictable economic outcomes, even in the absence of
retrospective data. See FCC v. WNCN Listeners Guild, 450
U.S. 582, 594-95 (1981) (approving of the FCC’s predictions
about the effects of market forces). FERC’s economic
reasoning also finds support in the submissions of PJM itself.
See Black Oak Energy, LLC, 122 FERC ¶ 61,208, 62,180.
These comments corroborate FERC’s reasonable economic
predictions.

In response, the virtual marketers present a parade of
horribles. They predict that the surplus allocation system will
deter virtual marketers from participating in the PJM markets,
“repress [efficient] price signals” to load-serving entities, and
generally reduce the efficiency of the PJM market. See Pet’rs’
Br. 31, 33, 35. But none of these possibilities – and as far as we
know, they are only possibilities – demonstrates the
irrationality of FERC’s decisions. First of all, when raising the
specter of decreased market participation by virtual traders, the
petitioners fail to distinguish between good participation and
bad participation. It is within FERC’s discretion to deter virtual
marketers from making certain kinds of trades while leaving in
place the background incentives to engage in
efficiency-promoting arbitrage. Regarding the repression of
efficient price signals to the load-serving entities and the
supposed threats to the efficiency of the market, the virtual
marketers point to no evidence supporting their view. FERC
sufficiently explained why the system it chose was, in the
Commission’s view, conducive to the production of efficient
price signals. See Black Oak Energy, LLC, 122 FERC
¶ 61,208, 62,184-86. At the very least, FERC determined that
the surplus allocation system was better than available
 19
alternatives at fostering an efficient marketplace. Id. The
arbitrary and capricious standard is a deferential one, and the
virtual marketers’ speculative claims are not sufficient to
overcome FERC’s explanation.

III

As discussed above, PJM’s surplus disbursement system
ties distributions to the payment of the fixed costs of the grid.
Though the virtual marketers pay none of those costs now, they
once did when they traded on a market called the Up-To
Congestion Market. 6 Even so, the virtual marketers now
receive no share of the surplus. Eventually, they filed a petition
with FERC objecting to their disparate treatment, and in
September 2009, the Commission ordered PJM to refund the
virtual marketers for the surplus allocations to which they were
entitled, amounting to $37 million. Black Oak Energy, LLC v.
PJM Interconnection, LLC, 128 FERC ¶ 61,262, 62,222
(2009).

But in July 2011, FERC took another look at the matter of
refunds and changed its view, effectively ordering PJM to

6
The Up-To Congestion Market allows traders to specify a cap
on the price they are willing to pay for the congestion component of
an LMP price between two points on the grid. See Issue Details:
Up-To Congestion Transactions, www.PJM.com (last visited July
25, 2013), http://www.pjm.com/committees-and-groups/issue-track
ing/issue-tracking-details.aspx?Issue={A1D2CD14-012A-47E0-84
56-A76BDB97BA6C}. Until September 17, 2010, whenever virtual
marketers made trades on the Up-To Congestion Market, they
acquired “transmission reservations,” which included a component
that paid for the fixed costs of the grid; since September 17, 2010,
Up-To Congestion trades have not involved payment of grid fixed
costs. See Pet’rs’ Br. 16 n.9 (citing PJM Interconnection Inc., LLC,
132 FERC ¶ 61,244 (2010)).
 20
recoup the refunds it had paid the virtual marketers. See Black
Oak Energy, LLC v. PJM Interconnection, LLC, 136 FERC
¶ 61,040, 61,163-64 (2011). The virtual marketers objected,
arguing that FERC failed to provide proper notice that it might
reconsider the decision to order refunds. In reply, FERC issued
an order in May 2012 explaining that the virtual marketers
should have been on notice and affirming its July 2011
decision not to order refunds. See Black Oak Energy, LLC v.
PJM Interconnection, LLC, 139 FERC ¶ 61,111, 61,780-82
(2012).

The virtual marketers subject to the recoupment now seek
review of the July 2011 and May 2012 orders. 7 (Collectively,
we call these the “Recoupment Orders.”) They argue that they
lacked proper notice that their refunds might be recouped, and
that the Recoupment Orders were, in any event, arbitrary,
capricious, and contrary to the FPA’s prohibitions on unjust,
unreasonable, and unduly discriminatory rates. We hold that
FERC gave the virtual marketers reasonable notice that their
refunds were under reconsideration, but that FERC’s orders
were arbitrary and capricious because they were insufficiently
justified.

A

FERC possesses sua sponte statutory authority to
reconsider its orders under certain conditions:

Until the record in a proceeding shall have been filed in a
court of appeals, . . . the Commission may at any time,
upon reasonable notice and in such manner as it shall

7
This group includes those who brought the petition for review
addressed in Parts I and II of this opinion, along with a group of
similarly situated petitioner-intervenors.
 21
deem proper, modify or set aside, in whole or in part, any
finding or order made or issued by it under the provisions
of this chapter.

16 U.S.C. § 825l(a). The virtual marketers argue that FERC
lacked authority to change its course on the refunds because it
gave them no “reasonable notice” that the issue was on the
table. We give Chevron deference to FERC’s view of what
constitutes “reasonable notice” even though it comes in this
case not explicitly, as a statement of law, but implicitly, as a
fact-bound determination. See INS v. Cardoza-Fonseca, 480
U.S. 421, 447-48 (1987) (applying the Chevron framework to
the “concrete meaning [given] through a process of
case-by-case adjudication” to the statutory term “well-founded
fear”); see also Nat’l R.R. Passenger Corp. v. Boston & Me.
Corp., 503 U.S. 407, 420 (1992) (deferring to the ICC’s
implicit interpretation of the statutory term “required” even
though the ICC “did not in so many words articulate” it).

In its May 2012 order, FERC reasoned that the virtual
marketers were put on notice that their refunds were at risk by
two prior docket entries. First, a group of electricity exporters
filed a request for rehearing of the September 2009 refund
order arguing that FERC precedent barred retroactive
alteration of the treatment of their surplus allocations. 8
Second, responding in April 2010 to that rehearing request and
others, FERC filed an order that significantly expanded upon
the scope of the exporters’ rehearing request. We need not
decide whether the exporters’ rehearing request provided
“reasonable notice” to the virtual marketers that their refunds
were being reconsidered because the April 2010 order did.

8
Electricity exporters conduct transactions that ship power
from within the PJM system into neighboring systems.
 22
The April 2010 order responded to arguments raised in the
exporters’ request for rehearing of the September 2009 order,
and expanded beyond them. The exporters contended that
retroactive alteration of the treatment of their surplus
allocations was contrary to FERC precedent. Their arguments
were equally applicable to the virtual marketers’ refunds. As a
non-profit, PJM lacks “corporate funds of its own to pay
refunds, and it would have to acquire such funds either through
surcharges or through an up-lift charge to all members.” Black
Oak Energy, 136 FERC ¶ 61,040, 61,164 n.42. Thus, PJM’s
membership must pay for any refund that FERC orders PJM to
pay. According to the exporters, members’ confidence in the
marketplace was shaken by having to pay for the refunds. See
Black Oak Energy, LLC, 139 FERC ¶ 61,111, 61,777,
61,780-81. This logic applied to any surplus allocation refund
made by PJM, and the April 2010 order expanded the scope of
reconsideration to include all the refunds ordered in September
2009.

The broad inquiry FERC initiated in the April 2010 order
should have made it clear to the virtual marketers that their
refunds were subject to reconsideration. That order gave all
parties “45 days from the date of PJM’s filing to brief any
issues with respect to refunds . . . .” See Black Oak Energy,
LLC v. PJM Interconnection, LLC, 131 FERC ¶ 61,024,
61,171-72 (2010) (emphasis added). In other words, the
September 2009 refund order was not final. The April 2010
order also directed PJM to submit a “detailed refund report,”
which would identify all parties burdened or benefited by the
refunds and would explain why PJM conducted the refund as it
had. Id. (The report was designed to update and clarify a refund
report that FERC required of PJM in September 2009. Id.) It is
reasonable for FERC to hold that the scope of the April 2010
order placed the virtual marketers on notice that their refunds
might be reconsidered.
 23

B

The virtual marketers argue that the Recoupment Orders
are unjust, unreasonable, and unduly discriminatory because
they “reinstitute” a tariff that FERC itself had found unlawful
in September 2009. Pet’rs’ Br. 46-47 (citing Black Oak
Energy, LLC, 128 FERC ¶ 61,262, 62,221-22 (holding that
PJM’s treatment of Up-To Congestion trades ran afoul of
§ 824d)). But the Recoupment Orders did not “reinstitute” an
unlawful tariff; they merely modified the remedy that FERC
ordered in September 2009. That order imposed a prospective
remedy, banning PJM from mistreating virtual marketers who
contribute to the fixed costs of the grid, and a retrospective
remedy, effectively ordering PJM to pay refunds. With the
revisions in the Recoupment Orders, PJM continues to be
bound by the ban on mistreatment of virtual marketers who
contribute to fixed costs. The revisions nullified only the
retrospective feature of the September 2009 remedy; they did
not reinstate an unjust tariff. Thus, the virtual marketers’
“unjust, unreasonable, and unduly discriminatory” argument
fails.

But we agree with the virtual marketers that the
Commission acted arbitrarily and capriciously when it
effectively ordered PJM to recoup the refunds. FERC justified
the recoupment on the ground that it brought the remedies for
PJM’s unjust distribution of the surplus into alignment with
Commission precedent. According to FERC, its policy is to
deny refunds where revenues were accurately collected, and
rates are being changed on a prospective basis. See Black Oak
Energy, LLC, 136 FERC ¶ 61,040, 61,163-64 (citing orders).
FERC argued here that PJM had accurately collected revenues
according to its LMP tariff, but that the system needed to be
altered. By this reasoning, had FERC followed its precedent in
 24
the first instance, there would have been no $37 million refund.
FERC would have required PJM to comply prospectively and
left it at that. The Recoupment Orders were FERC’s effort at
correcting this mistake. FERC admits that it “belatedly”
reached its conclusion that no refund should have been ordered
in the first place, thus “compelling PJM to recoup refunds it
previously made,” but argues that when it reached this correct
conclusion “is not of legal consequence.” Resp’ts Br. 41
(citation omitted). We disagree.

There is a significant distinction between denying refunds
and recouping them. As the virtual marketers argued in their
request for rehearing of the July 2011 order, recoupment may
reduce the confidence of participants in the smooth functioning
of the market in a way that straightforward denial of refunds
does not. Yet, in its Recoupment Orders, FERC repeatedly
obscured the fact that it was effectively ordering PJM to claw
back money that has already been paid out. Instead of
justifying recoupment, the Commission wrote as if it were
denying the refunds outright. The order stated, “denying
refunds . . . is the fairest approach,” and “refunds should not be
required.” Black Oak Energy, LLC, 139 FERC ¶ 61,111,
61,782-83. True enough, but there is more to this case than
that, for the refunds at issue were already out the door. In
addition to explaining why it should have denied the refunds in
the first place, FERC must explain why recouping is
warranted. Because FERC failed to explain how it analyzed
this crucial aspect of the case, we hold that the Commission
acted arbitrarily and capriciously. See, e.g., Motor Vehicle
Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43
(1983). It may well be that FERC’s policy reasons for
effectively ordering recoupment outweigh its negative effects,
but FERC must analyze that question, not ignore it. For that
reason, we remand.
 25
Although we remand, we do so without vacating the
Recoupment Orders. The decision to vacate depends on two
factors: the likelihood that “deficiencies” in an order can be
redressed on remand, even if the agency reaches the same
result, and the “disruptive consequences” of vacatur.
Allied-Signal v. Nuclear Regulatory Comm’n, 988 F.2d 146,
150-51 (D.C. Cir. 1993). We find it plausible that FERC can
redress its failure of explanation on remand while reaching the
same result. See, e.g., Lone Mountain Processing, Inc. v. Sec’y
of Labor, 709 F.3d 1161, 1164 (D.C. Cir. 2013) (“The
Commission may well arrive at the same result it reached
originally, but it must do so with more clarity than it showed in
the first instance.” (citation omitted)). And vacatur in this case
would certainly be disruptive because it would prompt yet
another refund, which would require yet another charge on
uninvolved market participants. As we have noted, because
PJM is a non-profit, the only way it can obtain funds to pay out
a refund is by charging its market participants to cover them.
See Black Oak Energy, LLC, 139 FERC ¶ 61,111, 61,783. If
FERC, considering all the factors, ultimately concludes that
recoupment was the proper path, the whole cycle would repeat
itself, imposing significant transaction costs on PJM, its
members, and the virtual marketers themselves. Faced with
those prospects, we deem it better to preserve the status quo as
FERC reconsiders its Recoupment Orders. However, we
emphasize that FERC’s opportunity to reconsider is not an
invitation to do nothing. See In re Core Commc’ns, Inc., 531
F.3d 849, 862 (D.C. Cir. 2008) (Griffith, J., concurring). The
Commission may not obtain the result it seeks through inaction
when it has failed to justify that result with reasoning.

IV

For the foregoing reasons, we deny the petition for review
of the Surplus Orders and grant the petition for review of the
 26
Recoupment Orders. We remand the matter of the recoupment
to the Commission for reconsideration consistent with this
opinion.

So ordered.