NACS v. Board of Governors of the Federal Reserve System

United States Court of Appeals

FOR THE DISTRICT OF COLUMBIA CIRCUIT

Argued January 17, 2014 Decided March 21, 2014

No. 13-5270

NACS, FORMERLY KNOWN AS NATIONAL ASSOCIATION OF

CONVENIENCE STORES, ET AL.,

APPELLEES

v.

BOARD OF GOVERNORS OF THE FEDERAL RESERVE SYSTEM,

APPELLANT

Appeal from the United States District Court

for the District of Columbia

(No. 1:11-cv-02075)

Katherine H. Wheatley, Associate General Counsel, Board

of Governors of the Federal Reserve System, argued the cause

for appellant. With her on the briefs were Richard M. Ashton,

Deputy General Counsel, Yvonne F. Mizusawa, Senior Counsel,

and Joshua P. Chadwick, Counsel.

Seth P. Waxman argued the cause for amici curiae The

Clearing House Association, L.L.C., et al. in support of neither

party. With him on the brief were Albinas Prizgintas, Noah A.

Levine, and Alan Schoenfeld.

2

Shannen W. Coffin argued the cause for appellees. With him

on the brief was Linda C. Bailey.

Andrew G. Celli Jr., Ilann M. Maazel, and O. Andrew F.

Wilson were on the brief for amicus curiae The Retail Litigation

Center, Inc. in support of appellees.

Jeffrey I. Shinder was on the brief for amici curiae

7-Eleven, Inc., et al. in support of appellees.

David A. Balto was on the brief for amicus curiae United

States Senator Richard J. Durbin in support of appellees.

Before: TATEL, Circuit Judge, and EDWARDS and

WILLIAMS, Senior Circuit Judges.

Opinion for the Court filed by Circuit Judge TATEL.

TATEL, Circuit Judge: Combining features of credit cards

and checks, debit cards have become not just the most popular

noncash payment method in the United States but also a source

of substantial revenue for banks and companies like Visa and

MasterCard that own and operate debit card networks. In 2009

alone, debit card holders used their cards 37.6 billion times,

completing transactions worth over $1.4 trillion and yielding

over $20 billion in fees for banks and networks. Concerned that

these fees were excessive and that merchants, who pay the fees

directly, and consumers, who pay a portion of the fees indirectly

in the form of higher prices, lacked any ability to resist them,

Congress included a provision in the Dodd-Frank financial

reform act directing the Board of Governors of the Federal

Reserve System to address this perceived market failure. In

response, the Board issued regulations imposing a cap on the

per-transaction fees banks receive and, in an effort to force

3

networks to compete for merchants’ business, requiring that at

least two networks owned and operated by different companies

be able to process transactions on each debit card. Merchant

groups challenged the regulations, seeking lower fees and even

more network competition. The district court granted summary

judgment to the merchants, concluding that the rules violate the

statute’s plain language. We disagree. Applying traditional tools

of statutory interpretation, we hold that the Board’s rules

generally rest on reasonable constructions of the statute, though

we remand one minor issue—the Board’s treatment of so-called

transactions-monitoring costs—to the Board for further

explanation.

I.

Understanding this case requires looking under the hood—

or, more accurately, behind the teller’s window—to see what

really happens when customers use their debit cards. After

providing some background about debit cards and the debit card

marketplace, we outline Congress’s effort to solve several

perceived market failures, the Board’s attempt to put Congress’s

directives into action, and the district court’s rejection of the

Board’s approach.

A.

We start with the basics. For purposes of this case, the term

“debit card” describes both traditional debit cards, which allow

cardholders to deduct money directly from their bank accounts,

and prepaid cards, which come loaded with a certain amount of

money that cardholders can spend down and, in some cases,

replenish. Debit card transactions are typically processed using

what is often called a “four party system.” The four parties are

the cardholder who makes the purchase, the merchant who

accepts the debit card payment, the cardholder’s bank (called the

“issuer” because it issues the debit card to the cardholder), and

4

the merchant’s bank (called the “acquirer” because it acquires

funds from the cardholder and deposits those funds in the

merchant’s account). In addition, each debit transaction is

processed on a particular debit card “network,” often affiliated

with MasterCard or Visa. The network transmits information

between the cardholder/issuer side of the transaction and the

merchant/acquirer side. Issuers activate certain networks on

debit cards, and only activated networks can process transactions

on those cards.

Virtually all debit card transactions fall into one of two

categories: personal identification number (PIN) or signature.

PIN and signature transactions employ different methods of

“authentication”—a process that establishes that the cardholder,

and not a thief, has actually initiated the transaction. In PIN

authentication, the cardholder usually enters her PIN into a

terminal. In signature authentication, the cardholder usually

signs a copy of the receipt. Most networks can process either

PIN transactions or signature transactions, but not both.

Signature networks employ infrastructure used to process credit

card payments, while PIN networks employ infrastructure used

by ATMs. Only about one-quarter of merchants currently accept

PIN debit. Some merchants have never acquired the terminals

needed for customers to enter their PINs, while others believe

that signature debit better suits their business needs. More about

this later. And merchants who sell online generally refuse to

accept PIN debit because customers worry about providing PINs

over the Internet. Merchants who do accept both PIN and

signature debit often allow customers to select whether to

process particular transactions on a PIN network or a signature

network.

Whether PIN or signature, a debit card transaction is

processed in three stages: authorization, clearance, and

5

settlement. Authorization begins when the cardholder swipes her

debit card, which sends an electronic “authorization request” to

the acquirer conveying the cardholder’s account information and

the transaction’s value. The acquirer then forwards that request

along the network to the issuer. Once the issuer has determined

whether the cardholder has sufficient funds in her account to

complete the transaction and whether the transaction appears

fraudulent, it sends a response to the merchant along the network

approving or rejecting the transaction. Even if the issuer

approves the transaction, that transaction still must be cleared

and settled before any money changes hands.

Clearance constitutes a formal request for payment sent

from the merchant on the network to the issuer. PIN transactions

are authorized and cleared simultaneously: because the

cardholder generally enters her PIN immediately after swiping

her card, the authorization request doubles as the clearance

message. Signature transactions are first authorized and

subsequently cleared: because the cardholder generally signs

only after the issuer has approved the transaction, the merchant

must send a separate clearance message. This difference between

PIN and signature processing explains why certain businesses,

including car rental companies, hotels, and sit-down restaurants,

often refuse to accept PIN debit. Car rental companies authorize

transactions at pick-up to ensure that customers have enough

money in their accounts to pay but postpone clearance to allow

for the possibility that the customer might damage the vehicle or

return it without a full tank of gas. Hotels authorize transactions

at check-in but postpone clearance to allow for the possibility

that the guest might trash the room, order room service, or

abscond with the towels and robes. And sit-down restaurants

authorize transactions for the full amount of the meal but

postpone clearance to give diners an opportunity to add a tip.

6

The final debit card payment processing step, settlement,

involves the actual transfer of funds from the issuer to the

acquirer. After settlement, the cardholder’s account has been

debited, the merchant’s account has been credited, and the

transaction has concluded. Rather than settle transactions one-

by-one, banks generally employ companies that determine each

bank’s net debtor/creditor position over a large number of

transactions and then settle those transactions simultaneously.

Along the way, and central to this case, the parties charge

each other various fees. The issuer charges the acquirer an

“interchange fee,” sometimes called a “swipe fee,” which

compensates the issuer for its role in processing the transaction.

The network charges both the issuer and the acquirer “network

processing fees,” otherwise known as “switch fees,” which

compensate the network for its role in processing the

transaction. Finally, the acquirer charges the merchant a

“merchant discount,” the difference between the transaction’s

face value and the amount the acquirer actually credits the

merchant’s account. Because the merchant discount includes the

full value of the interchange fee, the acquirer’s portion of the

network processing fee, other acquirer and network costs, and a

markup, merchants end up paying most of the costs acquirers

and issuers incur. Merchants in turn pass some of these costs

along to consumers in the form of higher prices. In contrast to

credit card fees, which generally represent a set percentage of

the value of a transaction, debit card fees change little as price

increases. Thus, a bookstore might pay the same fees to sell a

$25 hardcover that Mercedes would pay to sell a $75,000 car.

Before the Board promulgated the rules challenged in this

case, networks and issuers took advantage of three quirks in the

debit card market to increase fees without losing much business.

First, issuers had complete discretion to decide whether to

7

activate certain networks on their cards. For instance, an issuer

could limit payment processing to one Visa signature network, a

Visa signature network and a Visa PIN network, or Visa and

MasterCard signature and PIN networks. Second, networks had

complete discretion to set the level of interchange and network

processing fees. Finally, Visa and MasterCard controlled most of

the debit card market. According to one study entered into the

record, in 2009 networks affiliated with Visa or MasterCard

processed over eighty percent of all debit transactions. Steven C.

Salop, et al., Economic Analysis of Debit Card Regulation

Under Section 920, Paper for the Board of Governors of the

Federal Reserve System 10 (Oct. 27, 2010). Making things

worse for merchants, these companies imposed “Honor All

Cards” rules that prohibited merchants from accepting some but

not all of their credit cards and signature debit cards. Merchants

were therefore stuck paying whatever fees Visa and MasterCard

chose to set, unless they refused to accept any Visa and

MasterCard credit and signature debit cards—hardly a realistic

option for most merchants given the popularity of plastic.

Exercising this market power, issuers and networks often

entered into mutually beneficial agreements under which issuers

required merchants to route transactions on certain networks that

generally charged high processing fees so long as those networks

also set high interchange fees. Many of these agreements were

exclusive, meaning that issuers agreed to activate only one

network or only networks affiliated with one company.

Networks and issuers also negotiated routing priority

agreements, which forced merchants to process transactions on

certain activated networks rather than others. By 2009,

interchange and network processing fees had reached, on

average, 55.5 cents per transaction, including a 44 cent

interchange fee, a 6.5 cent network processing fee charged to the

issuer, and a 5 cent network processing fee charged to the

8

acquirer. Debit Card Interchange Fees and Routing, Notice of

Proposed Rulemaking (“NPRM”), 75 Fed. Reg. 81,722, 81,725

(Dec. 28, 2010).

B.

Seeking to correct the market defects that were contributing

to high and escalating fees, Congress passed the Durbin

Amendment as part of the 2010 Dodd-Frank Wall Street Reform

and Consumer Protection Act, Pub. L. No. 111-203, 124 Stat.

1376 (2010). The amendment, which modified the Electronic

Funds Transfer Act (EFTA), Pub. L. No. 95-630, 92 Stat. 3641

(1978), contains two key provisions. The first, EFTA section

920(a), restricts the amount of the interchange fee. Specifically,

it instructs the Board of Governors of the Federal Reserve

System to promulgate regulations ensuring that “the amount of

any interchange transaction fee . . . is reasonable and

proportional to the cost incurred by the issuer with respect to the

transaction.” 15 U.S.C. § 1693o-2(a)(3)(A); see also id. §

1693o-2(a)(6)–(7)(A) (exempting debit cards issued by banks

that, combined with all affiliates, have assets of less than $10

billion and debit cards affiliated with certain government

payment programs from interchange fee regulations). To this

end, section 920(a)(4)(B), in language the parties hotly debate,

requires the Board to “distinguish between . . . the incremental

cost incurred by an issuer for the role of the issuer in the

authorization, clearance, or settlement of a particular debit

transaction, which cost shall be considered . . . , [and] other

costs incurred by an issuer which are not specific to a particular

electronic debit transaction, which costs shall not be

considered.” Id. § 1693o-2(a)(4)(B)(i)-(ii). Like the parties, we

shall refer to the costs of “authorization, clearance, and

settlement” as “ACS costs.” In addition, section 920(a) “allow[s]

for an adjustment to the fee amount received or charged by an

issuer” to compensate for “costs incurred by the issuer in

9

preventing fraud in relation to electronic debit transactions

involving that issuer,” so long as the issuer “complies with the

fraud-related standards established by the Board.” Id. § 1693o-

2(a)(5)(A).

The second key provision, EFTA section 920(b), prohibits

certain exclusivity and routing priority agreements. Specifically,

it instructs the Board to promulgate regulations preventing any

“issuer or payment card network” from “restrict[ing] the number

of payment card networks on which an electronic debit

transaction may be processed to . . . 1 such network; or . . . 2 or

more [affiliated networks].” Id. § 1693o-2(b)(1)(A). It also

directs the Board to prescribe regulations that prohibit issuers

and networks from “inhibit[ing] the ability of any person who

accepts debit cards for payments to direct the routing of

electronic debit transactions for processing over any payment

card network that may process such transactions.” Id. § 1693o-

2(b)(1)(B). Congress anticipated that these prohibitions would

force networks to compete for merchants’ business, thus driving

down fees.

C.

In late 2010, the Board proposed rules to implement

sections 920(a) and (b). As for section 920(a), the Board

proposed allowing issuers to recover only “incremental” ACS

costs and interpreted “incremental” ACS costs to mean costs

that “vary with the number of transactions” an issuer processes

over the course of a year. NPRM, 75 Fed. Reg. at 81,735. Issuers

would thus be unable to recover “costs that are common to all

debit card transactions and could never be attributed to any

particular transaction (i.e., fixed costs), even if those costs are

specific to debit card transactions as a whole.” Id. at 81,736. The

Board “recognize[d]” that this definition would “impose[] a

burden on issuers by requiring issuers to segregate costs that

10

vary with the number of transactions from those that are largely

invariant to the number of transactions” and “that excluding

fixed costs may prevent issuers from recovering through

interchange fees some costs associated with debit card

transactions.” Id. The Board nonetheless determined that other

definitions of “incremental cost” “do not appropriately reflect

the incremental cost of a particular transaction to which the

statute refers.” Id. at 81,735. Limiting the interchange fee to

average variable ACS costs, the Board proposed allowing

issuers to recover at most 12 cents per transaction—considerably

less than the 44 cents issuers had previously received on

average. Id. at 81,736–39.

After evaluating thousands of comments, the Board issued a

Final Rule that almost doubled the proposed cap. The Board

abandoned its proposal to define “incremental” ACS costs to

mean average variable ACS costs, deciding instead not to define

the term “incremental costs” at all. Debit Card Interchange Fees

and Routing, Final Rule (“Final Rule”), 76 Fed. Reg. 43,394,

43,426–27 (July 20, 2011). Observing that “the requirement that

one set of costs be considered and another set of costs be

excluded suggests that Congress left to the implementing agency

discretion to consider costs that fall into neither category to the

extent necessary and appropriate to fulfill the purposes of the

statute,” the Board allowed issuers to recover all costs “other

than prohibited costs.” Id. Thus, in addition to average variable

ACS costs, issuers could recover: (1) what the proposed rule had

referred to as “fixed” ACS costs; (2) costs issuers incur as a

result of transactions-monitoring to prevent fraud; (3) fraud

losses, which are costs issuers incur as a result of settling

fraudulent transactions; and (4) network processing fees. Id. at

43,429–31. The Board prohibited issuers from recovering other

costs, such as corporate overhead and debit card production and

delivery costs, that the Board determined were not incurred to

11

process specific transactions. Id. at 43,427–29. Accounting for

all permissible costs, the Board raised the interchange fee cap to

21 cents plus an ad valorem component of 5 basis points (.05

percent of a transaction’s value) to compensate issuers for fraud

losses. Id. at 43,404.

In response to section 920(b), the Board’s proposed rule

outlined two possible approaches. Under “Alternative A,”

issuers would have to activate at least two unaffiliated networks

on each debit card regardless of method of authentication.

NPRM, 75 Fed. Reg. at 81,749. For example, an issuer could

activate a Visa signature network and a MasterCard PIN

network. Under “Alternative B,” issuers would have to activate

at least two unaffiliated networks for each method of

authentication. Id. at 81,749–50. For example, an issuer could

activate both Visa and MasterCard signature and PIN networks.

In the Final Rule the Board chose Alternative A.

Acknowledging that “Alternative A provides merchants fewer

routing options,” the Board reasoned that it satisfied statutory

requirements and advanced Congress’s desire to enhance

competition among networks without excessively undermining

the ability of cardholders to route transactions on their preferred

networks or “potentially limit[ing] the development and

introduction of new authentication methods.” Final Rule, 76

Fed. Reg. at 43,448.

D.

Upset that the Board had nearly doubled the interchange fee

cap (as compared to the proposed rule) and had selected the less

restrictive anti-exclusivity option, several merchant groups,

including NACS, the organization formerly known as the

National Association of Convenience Stores, filed suit in district

court. The merchants argued that both rules violate the plain

12

terms of the Durbin Amendment: the interchange fee cap

because the statute allows issuers to recover only average

variable ACS costs, not “fixed” ACS costs, transactions-

monitoring costs, fraud losses, or network processing fees; and

the anti-exclusivity rule because the statute requires that all

merchants—even those who refuse to accept PIN debit—be able

to route each debit transaction on multiple unaffiliated networks.

Several financial services industry groups, which during

rulemaking had urged the Board to set an even higher

interchange fee cap and adopt an even less restrictive anti-

exclusivity rule, participated as amici curiae in support of

neither party.

The district court granted summary judgment to the

merchants. The court began by observing that “[a]ccording to

the Board, [the statute contains] ambiguity that the Board has

discretion to resolve. How convenient.” NACS v. Board of

Governors of the Federal Reserve System, 958 F. Supp. 2d 85,

101 (D.D.C. 2013). Rejecting this view, the district court

determined that the Durbin Amendment is “clear with regard to

what costs the Board may consider in setting the interchange fee

standard: Incremental ACS costs of individual transactions

incurred by issuers may be considered. That’s it!” Id. at 105. The

district court thus concluded that the Board had erred in

allowing issuers to recover “fixed” ACS costs, transactions-

monitoring costs, fraud losses, and network processing fees. Id.

at 105–09. The court also agreed with the merchants that section

920(b) unambiguously requires that all merchants be able to

route every transaction on at least two unaffiliated networks. Id.

at 109–14. The Board’s final anti-exclusivity rule, the district

court held, “not only fails to carry out Congress’s intention; it

effectively countermands it!” Id. at 112. Concluding that “the

Board completely misunderstood the Durbin Amendment’s

statutory directive and interpreted the law in ways that were

13

clearly foreclosed by Congress,” the district court vacated and

remanded both the interchange fee rule and the anti-exclusivity

rule. Id. at 114. But because regulated parties had already “made

extensive commitments” in reliance on the Board’s rules, the

district court stayed vacatur to provide the Board a short period

of time in which to promulgate new rules consistent with the

statute. Id. at 115. Subsequently, the district court granted a stay

pending appeal.

The Board now appeals, arguing that both rules rest on

reasonable constructions of ambiguous statutory language.

Financial services amici, urging reversal but still ostensibly

appearing in support of neither party, filed a brief and

participated in oral argument—though we have considered only

those arguments that at least one party has not disavowed. See

Eldred v. Reno, 239 F.3d 372, 378 (D.C. Cir. 2001) (noting that

arguments “rejected by the actual parties to this case” are “not

properly before us”); Eldred v. Ashcroft, 255 F.3d 849, 854

(D.C. Cir. 2001) (Sentelle, J., dissenting from denial of

rehearing en banc) (“Under the panel’s holding, it is now the law

of this circuit that amici are precluded both from raising new

issues and from raising new arguments.”). In a case like this, “in

which the District Court reviewed an agency action under the

[Administrative Procedures Act], we review the administrative

action directly, according no particular deference to the

judgment of the District Court.” In re Polar Bear Endangered

Species Act Listing and Section 4(d) Rule Litigation, 720 F.3d

354, 358 (D.C. Cir. 2013) (internal quotation marks omitted).

Because the Board has sole discretion to administer the Durbin

Amendment, we apply the familiar two-step framework set forth

in Chevron U.S.A. Inc. v. Natural Resources Defense Council,

Inc., 467 U.S. 837 (1984). At Chevron’s first step, we consider

whether, as the district court concluded, Congress has “directly

spoken to the precise question at issue.” Id. at 842. If not, we

14

proceed to Chevron’s second step where we determine whether

the Board’s rules rest on “reasonable” interpretations of the

Durbin Amendment. Id. at 844.

Before addressing the parties’ arguments, we think it worth

emphasizing that Congress put the Board, the district court, and

us in a real bind. Perhaps unsurprising given that the Durbin

Amendment was crafted in conference committee at the eleventh

hour, its language is confusing and its structure convoluted. But

because neither agencies nor courts have authority to disregard

the demands of even poorly drafted legislation, we must do our

best to discern Congress’s intent and to determine whether the

Board’s regulations are faithful to it.

II.

We begin with the interchange fee. Recall that section

920(a)(4)(B)(i) requires the Board to include “incremental

cost[s] incurred by an issuer for the role of the issuer in the

authorization, clearance, or settlement of a particular electronic

debit transaction,” and that section 920(a)(4)(B)(ii) prohibits the

Board from including “other costs incurred by an issuer which

are not specific to a particular electronic debit transaction.”

Echoing the district court, the merchants argue that the two

sections unambiguously permit issuers to recover only

“incremental” ACS costs. “The plain language of the Durbin

Amendment,” the merchants insist, “does not grant the Board

the discretion it claims to consider costs beyond those delineated

in Section 920(a)(4)(B).” Appellees’ Br. 26; see also NACS, 958

F. Supp. 2d at 100 (noting that the district court had “no

difficulty concluding that the statutory language evidences an

intent by Congress to bifurcate the entire universe of costs

associated with interchange fees”). Alternatively, the merchants

briefly argue that even if section 920(a)(4)(B) is ambiguous, the

Board’s resolution of that ambiguity was unreasonable—though

15

they acknowledge that this argument essentially rehashes their

Chevron step one argument. See Appellees’ Br. 44 (“Many of

the same arguments discussed above also demonstrate the

unreasonableness of the interchange fee standard.”). The Board

also thinks the Durbin Amendment is unambiguous, though it

argues that the statute clearly establishes a third category of

costs: those that are not “incremental” ACS costs but are

specific to a particular transaction. See Final Rule, 76 Fed. Reg.

at 43,426 (“[T]here exist costs that are not encompassed in

either the set of costs the Board must consider under Section

920(a)(4)(B)(i), or the set of costs the Board may not consider

under Section 920(a)(4)(B)(ii).”). Relying on the requirement

that the interchange transaction fee be “reasonable and

proportional to the cost incurred by the issuer with respect to the

transaction,” 15 U.S.C. § 1693o-2(a)(2), (a)(3)(A), the Board

concludes that it may but need not allow issuers to recover costs

falling within this third category, subject of course to other

statutory constraints. Like the merchants, the Board also offers a

Chevron step two argument. See Appellant’s Br. 71 (“Even

assuming for the sake of argument that the district court offered

a possible reading, the statute does not unambiguously foreclose

the Board’s construction . . . .”).

The parties’ competing arguments present us with two

options. Were we to agree with the merchants that the statute

allows recovery only of “incremental” ACS costs, we would

have to invalidate the rule without considering the particular

categories of costs the merchants challenge given that the Board

expressly declined to define the ambiguous statutory term

“incremental,” let alone determine whether those particular types

of costs qualify as “incremental” ACS costs. See Securities &

Exchange Commission v. Chenery Corp., 318 U.S. 80, 87 (1943)

(“The grounds upon which an administrative order must be

judged are those upon which the record discloses that its action

16

was based.”). Were we to determine that the Board’s reading of

section 920(a)(4)(B) is either compelled by the statute or

reasonable, we would have to go on to consider whether the

statute allows recovery of “fixed” ACS costs, transactions-

monitoring costs, fraud losses, and network processing fees. We

must therefore first decide whether section 920(a)(4)(B)

bifurcates the entire universe of costs the Board may consider, or

whether the statute allows for the existence of a third category of

costs that falls outside the two categories specifically listed.

A.

The Board may well have been able to interpret section

920(a)(4)(B) as the merchants urge. Such a reading could rely on

the statutory mandate to “distinguish between” one set of costs

and “other costs,” and could interpret section 920(a)(4)(B)(i) as

referring to variable costs and section 920(a)(4)(B)(ii) as

referring to fixed costs. But contrary to the merchants’ position,

and consistent with the Board’s Chevron step two argument, we

certainly see nothing in the statute’s language compelling that

result. The merchants’ preferred reading requires assuming that

the phrase “incremental cost incurred by the issuer for the role of

the issuer in the authorization, clearance, and settlement of a

particular electronic debit transaction” describes all issuer costs

“specific to a particular electronic debit transaction.” For several

reasons, however, we believe that phrase could just as easily, if

not more easily, be read to qualify the language of section

920(a)(4)(B)(i) such that it encompasses a subset of costs

specific to a particular transaction, leaving other costs specific to

a particular transaction unmentioned.

To begin with, as the Board pointed out in the Final Rule,

the phrase “incremental cost” has a several possible definitions,

including marginal cost, variable cost, “the cost of producing

some increment of output greater than a single unit but less than

17

the entire production run,” and “the difference between the cost

incurred by a firm if it produces a particular quantity of a good

and the cost incurred by the firm if it does not produce the good

at all.” Final Rule, 76 Fed. Reg. at 43,426–27. As a result,

depending on how these terms are defined, the category of

“incremental” costs would not necessarily encompass all costs

that are “specific to a particular electronic debit transaction.” See

infra at 26 (noting the parties’ agreement that the “specific to a

particular electronic debit transaction” phrase should not be read

to limit issuers to recovering only the marginal cost of each

particular transaction).

Second, the phrase “incurred by an issuer for the role of the

issuer in the authorization, clearance, or settlement of a

particular electronic debit transaction” limits the class of

“incremental” costs the Board must consider. So even if the

word “incremental” were read to include all costs specific to a

particular transaction, Congress left unmentioned incremental

costs other than incremental ACS costs. See Final Rule, 76 Fed.

Reg. at 43,426 n.116 (“The reference in Section 920(a)(4)(B)(i)

requiring consideration of the incremental costs incurred in the

‘authorization, clearance, or settlement of a particular

transaction’ and the reference in section 920(a)(4)(B)(ii)

prohibiting consideration of costs that are ‘not specific to a

particular electronic debit transaction,’ read together, recognize

that there may be costs that are specific to a particular electronic

debit transaction that are not incurred in the authorization,

clearance, or settlement of that transaction.”). For example, in

the proposed rule the Board determined that “cardholder rewards

that are paid by the issuer to the cardholder for each transaction”

and “costs associated with providing customer service to

cardholders for particular transactions” are “associated with a

particular transaction” but “are not incurred by the issuer for its

role in authorization, clearing, and settlement of that

18

transaction.” NPRM, 75 Fed. Reg. at 81,735. Moreover, in the

Final Rule the Board explained that fraud losses “are specific to

a particular transaction” because they result from the settlement

of particular fraudulent transactions, but are not incurred by the

issuer for the role of the issuer in the authorization, clearance, or

settlement of particular transactions. Final Rule, 76 Fed. Reg. at

43,431 (describing fraud losses as “the result of an issuer’s

authorization, clearance, or settlement of a particular electronic

debit transaction that the cardholder later identifies as

fraudulent”); see also Appellant’s Br. 67 (defending the Board’s

decision to allow issuers to recover some fraud losses on the

ground that fraud losses fall outside section 920(a)(4)(B)).

Third, as the Board pointed out, had Congress wanted to

allow issuers to recover only incremental ACS costs, it could

have done so directly. See Final Rule, 76 Fed. Reg. at 43,426.

For instance, in section 920(a)(3)(A) Congress could have

instructed the Board to “promulgate regulations ensuring that

interchange fees are reasonable and proportional to the

incremental costs of authorization, clearance, and settlement that

an issuer incurs with respect to a particular electronic debit

transaction.” Instead, in section 920(a)(3)(A) Congress required

the Board to promulgate regulations ensuring that interchange

fees are “reasonable and proportional to the cost incurred by the

issuer with respect to the transaction” and separately instructed

the Board, when determining issuer costs, to “distinguish

between” incremental ACS costs, which the Board must

consider, 15 U.S.C. § 1693o-2(a)(4)(B)(i), and “other costs . . .

which are not specific to a particular electronic debit

transaction,” which the Board must not consider, id. § 1693o-

2(a)(4)(B)(ii).

The merchants advance several arguments in support of the

opposite conclusion. They first assert that the “which” clause in

19

the phrase “other costs incurred by an issuer which are not

specific to a particular electronic debit transaction” should be

read descriptively rather than restrictively. As their labels

suggest, descriptive clauses explain, while restrictive clauses

define. To illustrate, consider a simple sentence: “the cars which

are blue have sunroofs.” Read descriptively, the clause “which

are blue” states a fact about the entire class of cars, which also

happen to have sunroofs. Read restrictively, the clause defines a

particular class of cars—blue cars—all of which have sunroofs.

Although often subtle, the distinction between descriptive and

restrictive clauses makes all the difference in this case. Here’s

why.

We have thus far assumed that section 920(a)(4)(B)(ii)’s

“which” clause should be read restrictively. On this reading (the

Board’s), the clause defines the class of “other costs” issuers are

precluded from recovering. As explained above, based on this

restrictive reading the Board reasonably concluded that the

statute establishes three categories of costs. But if the clause

should instead be read descriptively, then it would describe a

characteristic of “other costs” without limiting the meaning of

“other costs.” On this reading (the merchants’), the statute

bifurcates the entire universe of costs, requiring the Board to

define the statutory term “incremental cost incurred by an issuer

for the role of the issuer in the authorization, clearance, or

settlement of a particular electronic debit transaction” as

including all costs other than costs “not specific to a particular

electronic debit transaction.”

Normally, writers distinguish between descriptive and

restrictive clauses by setting the former but not the latter aside

with commas and by introducing the former with “which” and

the latter with “that.” Here, Congress introduced the clause at

issue with the word “which” but failed to set it aside with

20

commas. Word choice thus suggests a descriptive reading of the

clause, while punctuation suggests a restrictive reading. In

support of a descriptive reading, the merchants rely on a ninety-

year-old Supreme Court case for the proposition that

“[p]unctuation is a minor, and not a controlling, element in

interpretation.” Barrett v. Van Pelt, 268 U.S. 86, 91 (1925); see

also NACS, 958 F. Supp. 2d at 102 (calling Congress’s failure to

use commas a “red herring”). This decision provides the

merchants little help. Not only was it written long before the

development of modern approaches to statutory interpretation,

see U.S. National Bank of Oregon v. Independent Insurance

Agents of America, Inc., 508 U.S. 439, 454–55 (1993) (noting

that although reliance on punctuation must not “distort[] a

statute’s true meaning,” “[a] statute’s plain meaning must be

enforced, of course, and the meaning of a statute will typically

heed the commands of its punctuation”), but it addressed

statutory language that, unlike here, contained a clearly

misplaced comma, Barrett, 268 U.S. at 88 (interpreting a statute

“so inapt and defective that it is difficult to give it a construction

that is wholly satisfactory” without ignoring its comma).

The idea that we should entirely ignore punctuation would

make English teachers cringe. Even if punctuation is sometimes

a minor element in interpreting the meaning of language,

punctuation is often crucial—a reader might appropriately gloss

over a comma mistakenly inserted between a noun and a verb

yet pay extra attention to a comma or semicolon setting off

separate items in a list. Following the merchants’ advice and

stuffing punctuation to the bottom of the interpretive toolbox

would run the risk of distorting the meaning of statutory

language. After all, Congress communicates through written

language, and one component of written language is grammar,

including punctuation. As Strunk and White puts it, “the best

writers sometimes disregard the rules of rhetoric. When they do

21

so, however, the reader will usually find in the sentence some

compensating merit, attained at the cost of the violation. Unless

he is certain of doing as well, he will probably do best to follow

the rules.” WILLIAM STRUNK, JR. & E.B. WHITE, THE ELEMENTS

OF STYLE xvii–xviii (4th ed. 2000) (internal quotation marks

omitted). Put another way, “all our thoughts can be rendered

with absolute clarity if we bother to put the right dots and

squiggles between the words in the right places.” LYNN TRUSS,

EATS, SHOOTS & LEAVES 201–02 (2003).

In this instance, the absence of commas matters far more

than Congress’s use of the word “which” rather than “that.”

Widely-respected style guides expressly require that commas set

off descriptive clauses, but refer to descriptive “which” and

restrictive “that” as a style preference rather than an ironclad

grammatical rule. As The Chicago Manual of Style explains:

A relative clause that is restrictive—that is, essential to

the meaning of the sentence—is neither preceded nor

followed by a comma. But a relative clause that could

be omitted without essential loss of meaning (a

nonrestrictive clause) should be both preceded and (if

the sentence continues) followed by a comma.

Although which can be used restrictively, many careful

writers preserve the distinction between restrictive that

(no commas) and descriptive which (commas).

THE CHICAGO MANUAL OF STYLE 250 (14th ed. 2003). Compare

STRUNK & WHITE at 3–4 (“Nonrestrictive relative clauses are

parenthetic. . . . Commas are therefore needed.”), and WILSON

FOLLETT, MODERN AMERICAN USAGE: A GUIDE 69 (Erik

Wensberg ed., 1998) (same), with STRUNK & WHITE at 59 (“The

use of which for that is common in written and spoken language.

. . . Occasionally which seems preferable to that . . . But it would

be a convenience to all if these two pronouns were used with

22

precision.”), and FOLLETT at 293 (“The alert reader will notice

that quite a few excellent authors decline to use that and which

in precisely the ways that late-twentieth-century grammar books

recommend.”).

In fact, elsewhere in the Durbin Amendment Congress

demonstrated that it is among those writers who ignore the

distinction between descriptive “which” and restrictive “that.” In

section 920(b)(1)(A), for example, Congress instructed the

Board to prevent networks and issuers from activating on a debit

card only one network or “2 or more such networks which are

owned, controlled, or otherwise operated by” the same company.

15 U.S.C. § 1693o-2(b)(1)(A)(i)-(ii) (emphasis added). Even

though Congress used the word “which” to introduce this clause,

the clause is clearly restrictive. A descriptive reading would

require that the Board prevent issuers and networks from ever

activating “one network” or “2 or more such networks.” In other

words, a descriptive reading would prevent the activation of any

networks at all, rendering debit cards useless chunks of plastic.

Cf. Barnhart v. Thomas, 540 U.S. 20, 24 (2003) (finding a

restrictive clause in the statutory phrase “any other kind of

substantial gainful work which exists in the national economy”).

By contrast, in the Durbin Amendment Congress set aside every

clearly descriptive clause with commas. See, e.g., 15 U.S.C. §

1693o-2(a)(4)(B)(ii) (“other costs incurred by an issuer which

are not specific to a particular electronic debit transaction, which

costs shall not be considered under paragraph (2)” (emphasis

added)).

The merchants also emphasize Congress’s use of the terms

“distinguish between,” 15 U.S.C. § 1693o-2(a)(4)(B), and “other

costs,” id. § 1693o-2(a)(4)(B)(ii). According to the merchants,

the term “distinguish between” suggests that Congress required

the Board to “differentiate [between] the two categories of

23

costs,” and “the very use of the term ‘other costs’—as opposed

to simply ‘costs’—indicates the entire universe of costs that is

remaining after consideration of includable costs.” Appellees’

Br. 28. As noted above, these terms might provide some textual

support for the merchants’ preferred reading of the statute. But

given the Board’s reasonable determination that issuers incur

costs, other than incremental ACS costs, that are “specific to a

particular transaction,” the terms “distinguish between” and

“other costs” hardly compel the conclusion that the Board must

interpret section 920(a)(4)(B) as encompassing all costs that

issuers incur. Imagine that you make a deal to hand over part of

your baseball card collection and to distinguish between rookie

cards, which you must hand over, and other cards less than five

years old, which you must not. Although it would probably

make little financial sense, you could certainly hand over a 1960

Harmon Killebrew Topps card without violating the terms of the

deal.

Next, the merchants assert that the Board, by inferring the

existence of a third category of costs, improperly reads a

delegation of authority into congressional silence. According to

the merchants, “Congress would not delineate with specificity

the characteristics of includable costs (e.g., incremental) if it

intended, by its silence, to allow the Board to consider and

include their opposite (e.g., nonincremental).” Appellees’ Br.

31; accord American Petroleum Institute v. Environmental

Protection Agency, 198 F.3d 275, 278 (D.C. Cir. 2000) (“[I]f

Congress makes an explicit provision for apples, oranges and

bananas, it is most unlikely to have meant grapefruit.”). But

section 920(a)(3)(A) clearly grants the Board authority to

promulgate regulations ensuring that interchange fees are

reasonable and proportional to costs issuers incur. The question

then is how section 920(a)(4)(B) limits the Board’s discretion to

define the statutory term “cost incurred by the issuer with

24

respect to the transaction,” not whether that section affirmatively

grants the Board authority to allow issuers to recover certain

costs.

Finally, in a footnote the merchants point to section

920(a)(3)(B)’s requirement that the Board disclose certain ACS

cost information and to section 920(a)(4)(A)’s requirement that

the Board “consider the functional similarity between electronic

debit transactions and checking transactions that are required

within the Federal Reserve bank system to clear at par.” The

district court relied heavily on these provisions, concluding that

Congress’s decisions to limit disclosure “to the same costs

specified in section (a)(4)(B)(i)” and to direct the Board to

consider similarities, but not differences, between checks and

debit cards support the merchants’ interpretation of the statute.

NACS, 958 F. Supp. 2d at 103–04. But even assuming the

disclosure provision mirrors section 920(a)(4)(B)(i)’s reference

to incremental ACS costs—the word “incremental” appears

nowhere in the disclosure provision—the statute also allows the

Board to collect “such information as may be necessary to carry

out the provisions of this section,” not just information about

incremental ACS costs. 15 U.S.C. § 1693o-2(a)(3)(B). Similarly,

Congress’s instruction to the Board to “consider the functional

similarity between electronic debit transactions and checking

transactions” hardly precludes the Board from considering

differences as well. Doing just that, the Board decided that it

could allow banks to recover some costs in the debit card

context that they are unable to recover in the checking context.

Given the Durbin Amendment’s ambiguity as to the

existence of a third category of costs, we must defer to the

Board’s reasonable determination that the statute splits costs into

three categories: (1) incremental ACS costs, which the Board

must allow issuers to recover; (2) costs specific to a particular

25

transaction, other than incremental ACS costs, which the Board

may, but need not, allow issuers to recover; and (3) costs not

specific to a particular transaction, which the Board may not

allow issuers to recover. See Chevron, 467 U.S. at 843

(“Sometimes the legislative delegation to an agency on a

particular question is implicit rather than explicit. In such a case,

a court may not substitute its own construction of a statutory

provision for a reasonable interpretation made by the

administrator of an agency.”).

B.

Because the Board reasonably interpreted the Durbin

Amendment as allowing issuers to recover some costs in

addition to incremental ACS costs, we must now determine

whether the Board reasonably concluded that issuers can recover

the four specific types of costs the merchants challenge: “fixed”

ACS costs, network processing fees, fraud losses, and

transactions-monitoring costs. Much like agency ratemaking,

determining whether issuers or merchants should bear certain

costs is “far from an exact science and involves policy

determinations in which the [Board] is acknowledged to have

expertise.” Time Warner Entertainment Co. v. Federal

Communications Commission, 56 F.3d 151, 163 (D.C. Cir.

1995) (internal quotation marks omitted). We afford agencies

special deference when they make these sorts of determinations.

See, e.g., BNSF Railway Co. v. Surface Transportation Board,

526 F.3d 770, 774 (D.C. Cir. 2008) (“In the rate-making area,

our review is particularly deferential, as the Board is the expert

body Congress has designated to weigh the many factors at issue

when assessing whether a rate is just and reasonable.”); Time

Warner, 56 F.3d at 163. With that caution in mind, we address

each category of costs.

26

“Fixed” ACS Costs

Microeconomics textbooks draw a clear distinction between

“fixed” and “variable” costs: fixed costs are incurred regardless

of transaction volume, whereas variable costs change as

transaction volume increases. E.g., N. GREGORY MANKIW,

PRINCIPLES OF MICROECONOMICS 276–77 (3d ed. 2004). The

merchants, noting that the statute precludes recovery of costs

“not specific to a particular . . . transaction,” 15 U.S.C. § 1693o-

2(a)(4)(B)(ii), argue that the Board’s Final Rule improperly

allows recovery of fixed costs such as “equipment, hardware and

software.” Appellees’ Br. 35. “By definition,” the merchants

declare, “fixed costs are not ‘specific’ to any ‘particular’

transaction and fall squarely within the statute’s excludable costs

provision.” Id. at 39. The merchants therefore urge us to require

the Board to return to something along the lines of its proposed

rule, under which merchants could only recover average variable

ACS costs.

The merchants’ argument certainly has some persuasive

power. One might think it a stretch if a shoe store claimed that

the rent it pays its landlord is somehow “specific” to a

“particular” shoe sale. But the merchants have never argued that

issuers should be allowed to recover only costs incurred as a

result of processing individual, isolated transactions. See NPRM,

75 Fed. Reg. at 81,736 (requesting comment about whether

“costs should be limited to the marginal cost of a transaction”);

Final Rule, 76 Fed. Reg. at 43,427 n.118 (noting that “[t]he

Board did not receive comments regarding the use of marginal

cost”). Indeed, the Board’s proposed rule, which the merchants

seem to endorse, would have allowed recovery of costs that are

variable over the course of a year but could not be traced to any

one particular transaction.

27

We think the Board reasonably declined to read section

920(a)(4)(B) as preventing issuers from recovering “fixed”

costs. As the Board pointed out, the distinction the merchants

urge between what they refer to as non-includable “fixed” costs

and includable “variable” costs depends entirely on whether, on

an issuer-by-issuer basis, certain costs happen to vary based on

transaction volume in a particular year. For example, in any

given year one issuer might classify labor as an includable cost

because labor costs happened to vary based on transaction

volume over that year, while another issuer might classify labor

as a non-includable cost because such costs happened to remain

fixed over that year. See Final Rule, 76 Fed. Reg. at 43,427.

Moreover, the Board pointed out, the distinction between

variable and fixed ACS costs depends in some instances on

whether an issuer “performs its transactions processing in-

house” or “outsource[s] its debit card operations to a third-party

processor that charge[s] issuers a per-transaction fee based on its

entire cost.” Id. In any event, the Board concluded, requiring

issuers to segregate includable “variable” costs from excludable

“fixed” costs on a year-by-year basis would prove “exceedingly

difficult for issuers . . . [because] even if a clear line could be

drawn between an issuer’s costs that are variable and those that

are fixed, issuers’ cost-accounting systems are not generally set

up to differentiate between fixed and variable costs.” Id. The

Board therefore determined that any distinction between fixed

and variable costs would prove artificial and unworkable.

Instead, pointing out that the statute requires interchange

fees to be “reasonable and proportional” to issuer costs, the

Board interpreted section 920(a)(4)(B) as allowing issuers to

recover costs they must incur in order to effectuate particular

electronic debit card transactions but precluding them from

recovering other costs too remote from the processing of actual

transactions. “This reading interpret[s] costs that ‘are not

28

specific to a particular electronic debit transaction,’ and . . .

cannot be considered by the Board, to mean those costs that are

not incurred in the course of effecting any electronic debit

transaction.” Id. at 43,426. In our view, the Board reasonably

distinguished between costs issuers could recover and those they

could not recover on the basis of whether those costs are

“incurred in the course of effecting” transactions. Id. For

instance, the Board’s rule allows issuers to recover equipment,

hardware, software, and labor costs since “[e]ach transaction

uses the equipment, hardware, software and associated labor,

and no particular transaction can occur without incurring these

costs.” Id. at 43,430. By contrast, the rule precludes issuers from

recovering the costs of producing and distributing debit cards

because “an issuer’s card production and delivery costs . . . are

incurred without regard to whether, how often, or in what way

an electronic debit transaction will occur.” Id. at 43,428. Given

the Board’s expertise, we see no basis for upsetting its

reasonable line-drawing. See ExxonMobil Gas Marketing Co. v.

Federal Energy Regulatory Commission, 297 F.3d 1071, 1085

(D.C. Cir. 2002) (“We are generally unwilling to review line-

drawing . . . unless a petitioner can demonstrate that lines drawn

. . . are patently unreasonable, having no relationship to the

underlying regulatory problem.” (internal quotation marks

omitted)).

Network Processing Fees

This is easy. Network processing fees, which issuers pay on

a per-transaction basis, are obviously specific to particular

transactions. The merchants argue that allowing issuers to

recover network processing fees through the interchange fee

would run afoul of section 920(a)(8)(B), which requires the

Board to ensure that “a network fee is not used to directly or

indirectly compensate an issuer with respect to an electronic

29

debit transaction.” Perhaps signaling that even the merchants are

not entirely confident about this argument, they present it only in

a footnote. The merchants should have left it out entirely. As the

Board points out, section 920(a)(8)(B) is designed to prevent

issuers and networks from circumventing the Board’s

interchange fee rules, not to prevent issuers from recovering

reasonable network processing fees through the interchange fee.

Final Rule, 76 Fed. Reg. at 43,442 (“[Section 920(a)(8)(B)]

authorizes the Board to prescribe rules to prevent circumvention

or evasion of the interchange transaction fee standards.”).

Fraud Losses

The merchants nowhere challenge the Board’s conclusion

that fraud losses, which result from the settlement of particular

fraudulent transactions, are specific to those transactions. The

only question is whether a separate provision of the Durbin

Amendment—section 920(a)(5)’s fraud-prevention adjustment,

which allows issuers to recover fraud-prevention costs if those

issuers comply with the Board’s fraud-prevention standards—

precludes the Board from allowing issuers to recover fraud

losses as part of section 920(a)(2)’s “reasonable and

proportional” interchange fee. The merchants claim that it does.

First, noting that Congress intended the fraud-prevention

adjustment to be the only “fraud-related adjustment of the

issuer,” 15 U.S.C. § 1693o-2(a)(5)(A)(ii)(I), the merchants argue

that the Board should have allowed issuers to recover fraud-

related costs only through the fraud-prevention adjustment. We

disagree. The Board determined—reasonably in our view—that

because fraud losses result from the failure of fraud-prevention,

they do not themselves qualify as fraud-prevention costs. See

Final Rule, 76 Fed. Reg. at 43,431 (“An issuer may experience

losses for fraud that it cannot prevent and cannot charge back to

the acquirer or recoup from the cardholder.”). And nothing in the

30

statute suggests that Congress used the word “adjustment” to

describe the process of determining which costs issuers should

be allowed to recover directly through the interchange fee.

Rather, when discussing the fraud-prevention adjustment,

Congress empowered the Board to “allow for an adjustment to

the fee amount received or charged by an issuer under paragraph

(2).” 15 U.S.C. § 1693o-2(a)(5)(A). Paragraph (2), in turn,

requires that the interchange fee be “reasonable and

proportional” to costs incurred by issuers. Id. § 1693o-2(a)(2).

Thus, Congress used the word “adjustment” to describe a bonus

over and above the “reasonable and proportional” interchange

fee.

The merchants next maintain that allowing issuers to

recover fraud losses through the interchange fee “irrespective of

any particular bank’s efforts to reduce fraud” would undermine

Congress’s decision to condition receipt of the fraud-prevention

adjustment on compliance with the Board’s fraud-prevention

standards. Appellees’ Br. 43. Even assuming the merchants’

policy argument has some merit—allowing recovery of fraud

losses regardless of compliance with fraud-prevention standards

might well decrease issuers’ incentives to invest in fraud

prevention—the Board rejected it, reasoning that “[i]ssuers will

continue to bear the cost of some fraud losses and cardholders

will continue to demand protection against fraud.” Final Rule,

76 Fed. Reg. at 43,431. Such policy judgments are the province

of the Board, not this Court. See Village of Barrington, Illinois

v. Surface Transportation Board, 636 F.3d 650, 666 (D.C. Cir.

2011) (“As long as the agency stays within [Congress’s]

delegation, it is free to make policy choices in interpreting the

statute, and such interpretations are entitled to deference.”

(internal quotation marks omitted) (alterations in original)).

31

Transactions-Monitoring Costs

The Board acknowledged in the Final Rule that

transactions-monitoring costs, unlike fraud losses, are the

paradigmatic example of fraud-prevention costs. Final Rule, 76

Fed. Reg. at 43,397 (“The most commonly reported fraud

prevention activity was transaction monitoring.”). The Board

then distinguished between “[t]ransactions monitoring systems

[that] assist in the authorization process by providing

information to the issuer before the issuer decides to approve or

decline the transaction,” which the Board placed outside the

fraud-prevention adjustment, and “fraud-prevention activities . .

. that prevent fraud with respect to transactions at times other

than when the issuer is effecting the transaction”—for instance

the cost of sending “cardholder alerts . . . inquir[ing] about

suspicious activity”—which the Board determined should be

“considered in connection with the fraud-prevention

adjustment.” Id. at 43,430–31. Challenging this distinction, the

merchants think it “preposterous to suggest that Congress would

specifically address the costs associated with fraud prevention in

a separate provision of the statute, condition the recovery of

those costs on an issuer’s compliance with fraud prevention

measures, and then . . . permit recovery of those very same

costs” whether or not an issuer complies with fraud-prevention

standards. Appellees’ Br. 41.

As an initial matter, we agree with the Board that

transactions-monitoring costs can reasonably qualify both as

costs “specific to a particular . . . transaction” (section

920(a)(4)(B)) and as fraud-prevention costs (section 920(a)(5)).

Thus, the Board may have discretion either to allow issuers to

recover transactions-monitoring costs through the interchange

fee regardless of compliance with fraud-prevention standards or

to preclude issuers from recovering transactions-monitoring

32

costs unless those issuers comply with fraud-prevention

standards. That said, “an agency must cogently explain why it

has exercised its discretion in a given manner.” Motor Vehicle

Manufacturers Association of the United States v. State Farm

Mutual Automobile Insurance Co., 463 U.S. 29, 48 (1983). We

agree with the merchants that the Board has fallen short of that

standard.

The Board insists that the distinction it drew between fraud-

prevention costs falling outside the fraud-prevention adjustment

and fraud-prevention costs falling within it reflects the

distinction between, on the one hand, section 920(a)(4)(B)’s

focus on a single transaction and, on the other, section

920(a)(5)(A)(i)’s focus on “electronic debit transactions

involving that issuer.” According to the Board, Congress

“intended the . . . fraud-prevention adjustment to take into

account an issuer’s fraud prevention costs over a broad spectrum

of transactions that are not linked to a particular transaction.”

Appellant’s Br. 66–67. But as noted above, the Board

interpreted the term “specific to a particular . . . transaction” as

in fact allowing recovery of many costs not literally “specific” to

any one “particular” transaction. See supra at 26–28. The costs

of hardware, software, and labor seem no more “specific” to one

“particular” transaction than many of the fraud-prevention costs

the Board determined fall within the fraud prevention

adjustment. The Board’s own interpretation of the statute thus

undermines its justification for concluding that Congress

established a fraud-prevention adjustment, conditioned receipt of

that adjustment on compliance with fraud-prevention standards,

yet allowed issuers to recover the paradigmatic example of

fraud-prevention costs—transactions-monitoring costs—whether

or not issuers comply with those standards.

33

All that said, the Board may well be able to articulate a

reasonable justification for determining that transactions-

monitoring costs properly fall outside the fraud-prevention

adjustment. But the Board has yet to do so. “If the record before

the agency does not support the agency action, if the agency has

not considered all relevant factors, or if the reviewing court

simply cannot evaluate the challenged agency action on the

basis of the record before it, the proper course, except in rare

circumstances, is to remand to the agency for additional

investigation or explanation.” Florida Power & Light Co. v.

Lorion, 470 U.S. 729, 744 (1985) (emphasis added). We shall do

so here. Because the interchange fee rule generally rests on a

reasonable interpretation of the statute, because the Board may

well be able to articulate a sufficient explanation for its

treatment of fraud-prevention costs, and because vacatur of the

rule would be disruptive—the merchants seek an even lower

interchange fee cap, but vacating the Board’s rule would lead to

an entirely unregulated market, allowing the average interchange

fee to once again reach or exceed 44 cents per transaction—we

see no need to vacate. See Heartland Regional Medical Center

v. Sebelius, 566 F.3d 193, 198 (D.C. Cir. 2009) (noting that

remand without vacatur is warranted “[w]hen an agency may be

able readily to cure a defect in its explanation of a decision” and

the “disruptive effect of vacatur” is high); see also, e.g.,

Environmental Defense Fund v. Environmental Protection

Agency, 898 F.2d 183, 190 (D.C. Cir. 1990) (instructing that

courts should ordinarily remand without vacatur when vacatur

would “at least temporarily defeat” the interests of the party

successfully seeking remand).

III.

Having resolved the merchants’ challenges to the

interchange fee rule, we turn to the anti-exclusivity rule. As

explained above, see supra at 9, section 920(b) requires the

34

Board to promulgate regulations preventing “an issuer or

payment card network” from “restrict[ing] the number of

payment card networks on which an electronic debit transaction

may be processed” to a single network, or to networks affiliated

with one another. In the proposed rule, the Board outlined two

alternatives: require issuers and networks to activate two

unaffiliated networks or two unaffiliated networks for each

method of authentication. In the Final Rule, the Board chose the

former, requiring activation of two unaffiliated networks on each

debit card regardless of method of authentication.

The merchants believe that the Durbin Amendment

unambiguously requires that all merchants have multiple

unaffiliated network routing options for each debit transaction.

See NACS, 958 F. Supp. 2d at 109–12 (accepting this argument).

Arguing that the Board’s rule flunks this requirement, the

merchants emphasize two undisputed facts. First, given that

most merchants refuse to accept PIN debit, many transactions

can currently be processed only on signature debit. Second,

cardholders, not merchants, often have the ability to select

whether to process transactions on signature networks or PIN

networks. As a result, the merchants emphasize, under the

Board’s rule many merchants will still lack the ability to choose

between unaffiliated networks when deciding how to process

particular transactions. Disputing none of this, the Board points

out that all merchants could accept PIN debit even if some

choose not to and emphasizes that the statute is silent about

“restrictions imposed by merchants or consumers that limit

routing choice.” Appellant’s Br. 22. Given the parties’

agreement that under the Board’s rule some merchants will lack

routing choice for particular transactions, we must determine

whether the statute requires that all merchants—even those who

voluntarily choose not to accept PIN debit—have the ability to

decide between unaffiliated networks when routing transactions.

35

The merchants have a steep hill to climb. Congress directed

the Board to issue rules that would accomplish a particular

objective, leaving it to the Board to decide how best to do so,

and the Board’s rule seems to comply perfectly with Congress’s

command. Under the rule, “issuer[s] and payment card

network[s]” cannot “restrict the number of payment card

networks on which an electronic debit transaction may be

processed” to only affiliated networks—exactly what the statute

requires. 15 U.S.C. § 1693o-2(b)(1)(A).

Undaunted, the merchants emphasize one largely

conclusory textual argument and allude to another. First, relying

on the statutory phrase “electronic debit transaction,” id. §

1693o-2(b)(1)(A), they maintain that the statute plainly “requires

the Board to ensure that merchants be afforded a choice of

networks for each debit transaction.” Appellees’ Br. 45. But

context matters. Relying on the statute’s reference to “issuer[s]

and payment card network[s],” the Board reasonably read the

“electronic debit transactions” phrase to prevent issuers and

networks, prior to instigation of any particular debit transaction,

from limiting the number of networks “on which an electronic

debit transaction may be processed” to only affiliated networks.

15 U.S.C. § 1693o-2(b)(1)(A) (emphasis added).

In a footnote, the merchants repeat, though they seem not to

embrace, a textual argument on which the district court relied.

Looking to the statutory definitions of “electronic debit

transaction” (“a transaction in which a person uses a debit card”)

and of “debit card” (“any card . . . issued or approved for use

through a payment card network to debit an asset account . . .

whether authorization is based on signature, PIN, or other

means”), id. § 1693o-2(c)(2), (c)(5), the district court ruled that

the statutory term “electronic debit transaction” requires that

issuers and networks activate multiple unaffiliated networks for

36

each transaction “whether authorization is based on signature,

PIN, or other means,” NACS, 958 F. Supp. 2d at 110–11. But we

think it quite implausible that Congress engaged in a high-stakes

game of hide-and-seek with the Board, writing a provision that

seems to require one thing but embedding a substantially

different and, according to financial services amici, much more

costly requirement in the statute’s definitions section. Cf.

Whitman v. American Trucking Association, 531 U.S. 457, 468

(2001) (“Congress . . . does not . . . hide elephants in

mouseholes.”).

The merchants also argue that the Board’s rule runs afoul of

the Durbin Amendment’s purpose. Pointing out that Congress

intended network competition to drive down network processing

fees, the merchants insist that the Board has undermined this

competitive market because “merchants will be deprived of

network choice for a substantial segment of debit transactions in

the marketplace today.” Appellees’ Br. 47. But the Board

thought differently. As it explained in the Final Rule,

“merchants that currently accept PIN debit would have routing

choice with respect to PIN debit transactions in many cases

where an issuer chooses to participate in multiple PIN debit

networks.” Final Rule, 76 Fed. Reg. at 43,448. Indeed, the Board

presents uncontested evidence demonstrating that its rule has, as

predicted, substantially increased network competition.

According to the Board, as a result of the rule over 100 million

debit cards were activated on new networks, and “[Visa], which

had previously accounted for approximately 50-60% of the [PIN

debit] market, lost roughly half that share.” Appellant’s Br. 37 &

n.6 (internal quotation marks omitted).

Of course, as the Board acknowledges, the merchants’

preferred rule would result in more competition. But in its Final

Rule the Board explained the policy considerations that led it to

37

reject that approach. For one thing, cardholders might prefer to

route transactions over certain networks, perhaps because they

believe those networks to have better fraud-prevention policies.

Final Rule, 76 Fed. Reg. at 43,447–48. Also, the merchants’

preferred rule “could potentially limit the development and

introduction of new authentication methods” since issuers would

be unable to compel merchants to accept new authentication

techniques. Final Rule, 76 Fed. Reg. at 43,448. The merchants

ignore these reasonable concerns. Given that the Board’s rule

advances the Durbin Amendment’s purpose, we decline to

second-guess its reasoned decision to reject an alternative option

that might have further advanced that purpose.

Next, the merchants emphasize the interaction between

section 920(b)’s two key components: the anti-exclusivity and

routing priority provisions. According to the merchants, the

Board’s anti-exclusivity rule renders the routing priority

provision meaningless, since merchants will often lack the

ability to choose between multiple unaffiliated routing options.

But as the Board points out, the merchants misunderstand the

routing priority provision. Recall that it prohibits issuers and

networks from requiring merchants to process transactions over

certain activated networks rather than others. Far from rendering

the routing priority provision a nullity, the Board’s anti-

exclusivity provision would be ineffective without it. Absent the

routing priority provision, issuers and networks could, for

instance, activate two PIN networks and a signature network

affiliated with one of the PIN networks and then require

merchants to route transactions over the PIN network affiliated

with the signature network rather than over the other PIN

network.

Finally, the merchants question the Board’s premise that it

is they, not issuers and networks, who restrict routing options for

38

transactions under the Board’s Final Rule. To this end, they

assert that issuer and network rules arbitrarily prevent merchants

from processing PIN transactions on signature networks and vice

versa, suggesting that the Board could comply with the statute

by eliminating the distinction between PIN and signature debit.

But even if issuers and networks are responsible for maintaining

this distinction—a point they strongly dispute—merchants, not

issuers or networks, limit their own options when they refuse to

accept PIN debit, and cardholders, not issuers or networks, limit

merchants’ options when given the ability to choose how to

process transactions. “The principal fallacy with the Merchants’

argument,” the Board aptly explains, “is that they selectively

view transactions only from their own perspective and only after

the point at which the merchant itself or the consumer may have

elected to restrict certain routing options,” whereas “section

920(b) speaks only in terms of issuer and payment card network

restrictions” imposed prior to initiation of any particular debit

card transaction. Reply Br. 2–3.

In sum, far from summiting the steep hill, the merchants

have barely left basecamp. We therefore defer to the Board’s

reasonable interpretation of section 920(b) and reject the

merchants’ challenges to the anti-exclusivity rule.

IV.

For the foregoing reasons, we reverse the district court’s

grant of summary judgment to the merchants and remand for

further proceedings consistent with this opinion.

So ordered.