See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/342763666

The Competitive Issues in Credit Card Markets

Article · July 2020

CITATIONS READS
0 384

1 author:

Hongru Tan
Sichuan University
8 PUBLICATIONS   14 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

No Surcharge Rule View project

All content following this page was uploaded by Hongru Tan on 13 July 2020.

The user has requested enhancement of the downloaded file.

 The Competitive Issues in Credit Card Markets∗

Hongru Tan†

July 13, 2020

Abstract

This paper considers competitive issues raised in credit card markets. Among others, the
issues are mainly regarding to welfare effects of no surcharge rule (NSR) clauses or more general
non-discriminatory provisions (NDPs), and the regulation of interchange or merchant fees. This
paper develops a theory of platform competition tailored to credit card markets, and finds
abundant meaningful results. First of all, we find that the welfare implication of NDPs hinges
on relative distortions induced by market power of networks and merchant internalization (MI).
When MI overwhelms market power, the ban is likely to increase social welfare, and vice versa.
Another impressive finding is that both cardholder consumers and merchant consumers obtain
zero benefit from using a credit card when NDPs is in place. This is because platforms are able
to extract all of consumer surplus via an inflated price of product which embodies excessive
merchant fees. Lifting the NDPs will unambiguously increases surplus of both cardholders and
merchants. Moreover, we find that retaining the NSR and applying tourist test regulation will
yield the identical results as when surcharges are allowed and optimally capped.
Keywords: Credit Card Markets; No Surcharge Rule; Non Discriminatory Provisions.
JEL classification: G21, G28, L52

∗
We would like to thank Julian Wright for helpful discussion and comments. Hongru Tan gratefully acknowledges
the National Social Science Foundation for Young Scholars of China (No. XXXXXXX). All errors are ours.
†
School of Economics, Sichuan University, China; [email protected]

1
1 Introduction

This paper presents a theory of platform competition to discuss the competitive effect of no-

surcharge rule (NSR) clause or more general non-discriminatory provisions (NDPs), and related

regulatory issues in credit card markets. The NSR is a business clause which prohibits merchants

from imposing extra charges on payments via a particular credit card than any other methods of

payments, including another credit card, debit cards or cash. The NSR is enacted and enforced

by credit card companies (Visa or MasterCard) and is one of the prerequisites that merchants

adopt the credit card services. An even stronger constraint is anti-steering restraint or NDPs in

case of American Express (AmEx) which forbid merchants from encourage customers to use any

other payment means by monetary and non-monetary methods, including revealing the very cost

of merchants.

Figure 1: closed and open payment systems

The often called credit card industry is precisely referred to as the market of General purchase

credit and charge (GPCC) cards.1 Typically, two types of networks provides the GPCC payment

service. They are closed system as AmEx and open system as Visa or MasterCard, which is shown in

Figure 1. In a closed system, AmEx directly serves both consumers and merchants, while Visa in an

open system indirectly serves the both end-user groups via issuers (or issuing banks) and acquirer (or

acquiring banks). The fee collection is also different. AmEx directly charges customer and merchant

fee ( or often called merchant discounts in practice) to consumers and merchants respectively.

Visa charges two network service fee to issuers and acquirers. Merchants and consumers pay the
1
See definition in the file U.S. v. American Exp. Co., 88 F. Supp. 143,204 (2015).

2
merchant fee and customer fee to acquirers and issuers respectively. Finally, acquirers pay issuers

an interchange fee (IF). Moreover, the open systems also experience a structure transition. Both

Visa and MasterCard went for Initial pubic offerings (IPOs) during the decade legislation,2 which

turns themselves from not-for-profit associations into for-profit companies. Before IPOs, Visa

or MasterCard collects no fees and the IF was collectively determined by their member banks.

Afterwards, they make a profit from the two network service fees.

The NSR clauses are at issue in the lawsuit against Visa and MasterCard – In re Payment Card

Interchange Fee and Merchant Discount Antitrust Litigation; and NDPs are the centre dispute

in lawsuit against AmEx – State of Ohio et al., v. American Express, 585 U.S. [2018]. The

non-monetary NDPs have also been applied by Visa or MasterCard before 2010. However, both

networks agreed to remove all the non-monetary NDPs, and merely AmEx insisted to keep them

and was willing to confront a litigation. In the case of Visa and MasterCard, merchants accused

the two credit card companies and banks of colluding to inflate their interchange fees and cause the

high fees borne by merchants and prohibiting from using surcharges to steer consumers towards

other payment of methods. In the case of AmEx, the plaintiffs including Department of Justice

(DOJ) and several states accused that NDPs “short-circuit the ordinary price-setting mechanism in

the network services by removing the competitive ‘reward’ for networks offering merchants a lower

price for acceptance services”, which leads to an inflated merchant fee.

The litigation of AmEx attracts great attention because its seminally applied the canonical

“rule of reason” to a two-sided platform, and explicitly treated two-sidedness as the key in analysis.

After years of hearing, AmEx finally won the case, and the courts conclude that plaintiffs have not

dismissed the burden to show anti-competitive effects of NDPs in the first step of rule of reason

analysis, given the relevant market is defined as the market of card transactions. Alternatively,

the courts think plaintiffs failed to show the anti-competitive effects of NDPs when both sides

of market are considered. Standing in contrast with the court’s decision, theory of this article

shows the opposite: when two sides of the markets are taken into account, NDPs adversely affects

competition and leads to surplus losses to both consumers and merchants. The same conclusion

also applies to the NSR clause of Visa or MasterCard when two networks competes and two sides

of market are considered.

2
The case In re Payment Card Interchange Fee and Merchant Discount Antitrust Litigation lasts for decades. We
will introduce the relevant cases in the background section.

3
 In particular, this article considers a two-sided monopolistic platform model to mimic the case

of AmEx, and oligopolistic platform competition model for the case of Visa and MasterCard. Each

platform provides the intermediation service to both consumers (or cardholders) and merchants. In

the case of multiple networks, either provides identical card services to both sides, and compete in

quantity as Cournot competition. Both consumers and merchants benefit from the consummated

transaction via card payments. The merchant is a monopolist in selling a product (or goods).

Consumers decide to buy a product or not, and the use of a particular payment means. MI prevails

among merchants when NDPs3 are in place. It indicates merchants, out of strategic consideration,

have to take and accept card payments despite bearing excessive merchant fees – the “must take”

phenomenon as in Vickers (2005). These excessive fees are eventually borne by consumers when they

are built into the price of products. The consumers, who pay with other alternatives, also pay part

of the fees. Alternatively, it indicates merchants internalize the consumer surplus when decide to

take the card payments or not. So, it is called merchant internalization. MI incurs an over-provision

of transaction services because it leads to consumers who do not enjoy the service cross-subsidize

those who use it. Notably, NDPs, which handicap merchants from truthfully revealing the costs

and renders a uniform price of product to all consumers, are the prerequisites to the existence of

MI. When these clauses are removed, MI goes away.

Our model finds abundant meaningful results. First, we find that the welfare implication of

NDPs hinges on relative distortions induced by market power of networks and MI. When the market

power is strong, it more than offsets the effect of MI and under-provision equilibrium occurs. In this

case, a ban of NDPs withdrawing the over-provision is likely to incur an under-provision equilibrium

with a declined social welfare. When market power is weak, the distortion of MI will dominate

and lead to a massive over-provision. When ban the NDPs, a less under-provision is realized since

only distortion of market power exists. Consequently, a positive welfare impact is inclined to occur.

Another impressive finding is that both consumers and merchant obtain zero benefit from using a

credit card when NDPs is in place. This is because platforms are able to extract all of consumer

surplus via an inflated price of product which embodies excessive merchant fees. Lifting the NDPs

will unambiguously increases surplus of both consumers and merchants.

In addition, we find that retaining the NDPs and applying tourist test regulation will yield the
3
This is also the case with NSR clause. Since NDPs includes NSR clauses, we will only mention NDPs for brevity
if it can be applied to both.

4
identical results as when allow surcharges and impose an optimal surcharge cap. This is to say, the

two scenarios will lead to same levels of surplus to all interested parties. The impact of total social

welfare of each policy could go both directions, hinging on both distortions mentioned above. We

also mimic the competition between AmEx and Visa (or MasterCard), and their different business

models. We manage to establish an equilibrium where AmEx focuses on a sector with higher

evaluation of transaction services, and with consumers spend heavily (as for charge cards), and

Visa serve both this high value sector and another low value sector, but could earn a profit of

interests from outstanding balance. Interestingly, the asymmetrical business models leaves some

consumer surplus under the NDPs. Other comparison are similar to the above analysis.

This article adds to the literature on antitrust and regulatory issues in credit card markets.

Rochet and Tirole (2002) is the pioneer study on the NSR clause. They first draw comparison be-

tween distortions due to market power and MI, but in a model with a single not-for-profit network

and Hotelling merchants. Competition between networks which should have been discussed but

not in their article. The assumption of Hotelling merchants also has a drawback – no excessive sur-

charges occur when the clauses are removed. In reality, excessive surcharge behavoirs are robust as

observed in Australia. Tan and Deng (2020) provide an explanation of these surcharging behavoirs.

We update the model by assuming competing for-profit networks to reflect the governance change

in the industry. This article also assume monopolistic merchants which could capture the excessive

surcharging behavoirs when the clauses are removed. Wright (2003) discusses the NSR clause, but

does not consider “must take” argument or effect of MI, yet the competition between networks.

Schwartz and Vincent (2006) suggest that competitive effect of NSR clauses hinges on the propor-

tion of cash and card using consumers, and it is determined exogenously in the model. Carlton

and Winter (2018) examine the NSR clauses in a traditional market approach. They regard it as a

vertical most-favored-nation (vMFN) restraint, and find that it suppresses competition upstream,

and it is a mechanism to expropriate consumer surplus. They also suggest that the litigation on

a two-sided market issue should follow the practices in a conventional case. Schwartz and Vincent

(2019) apply a model with competing networks, but do not manage to provide an answer to the

competitive effect. As claimed by authors the article rather provides a basis to understand the

competition between two platforms, and mechanism in setting fees. Edelman and Wright (2015)

generalize the analysis of NSR clause into broad issue of price coherence in two-sided markets.

5
They focus on the incentive to improve consumer side benefit. Their main finding is that “this

(price coherence) leads to inflated retail prices, excessive adoption of the intermediaries’services,

over-investment in benefits to buyers, and a reduction in consumer surplus and sometimes welfare.”

Another stream of related literature is on regulatory issues. Baxter (1983) first provides a

theory and justification in setting an interchange fees (IF). He finds that setting IF according

to banking costs could internalize the negative externality from consumers onto merchants under

certain condition, and therefore resolve a market failure. Schmalensee (2002) demonstrates that

the IF is a mechanism to balance the fees charged to both sides, but does not find a workable

theory to guide policy enforcement. Rochet and Tirole (2011) invent the tourist test to regulate

the IF, and find that the tourist test will maximize the social welfare or at least consumer surplus

when merchants are homogeneous. Wright (2012) establishes that tourist test is valid in a broader

setting including merchants are heterogeneous. Bourguignon et al. (2019) show that tourist test is

necessary and optimal with NSR clauses, and no regulation is needed when surcharges are allowed

and subject to an optimal limit. Aurazo and Vasquez (2019) consider how tax evasion advantage

with cash payments affect the threshold of tourist test in developing countries. Reisinger and Zenger

(2019) examine how tourist test affect the incentives of investment, finding that the test will yield

a too low incentive to invest under a total welfare standard, but an approximately optimal level of

incentive under a total user surplus standard. Bourreau and Verdier (2019) discuss how IFs affects

investment incentive of competing issuing banks to improve consumer services.

All these literature assume networks as not-for-profit associations and a pass-through logic in

setting IFs. That is to say, a linear relationship exists between the fees, and IFs work as a hook to

the fees borne by end users. An increase in the IF will rise the merchant fee and lower the customer

fee accordingly. However, Visa and MasterCard become for-profit companies post IPOs, and the

mechanism in setting IF charges as well. We modify the model to reflect this change and revisit the

very issue on IFs. More generally, Rysman and Wright (2014) survey the literature on credit card

markets and note that there is no study explaining the various surcharging behaviors in Australia

and other countries. Tan and Wright (2018) and Tan and Wright (2020) discuss optimal pricing

structures in general two-sided markets and its application to credit card markets.

6
2 Model

This section introduces the models. We start with a monopoly platform model to mimic the situa-

tion of AmEx. In doing so, we will explain the intuition regarding NDPs. We then consider duopoly

platform competition and the more general case of oligopoly platforms for the case of Visa and Mas-

terCard. Since we apply a Cournot model unlike most of previous literature, I shall first elaborate

on its plausibility. “Cournot” means that the two platforms provide identical transaction services

and compete in quantity. In practice, consumers can hardly tell the difference between a Visa card

and a MasterCard card, except for their names or logos. Both brands historically charge the same

rate of interchange fees,4 and offer similar rewards scheme to consumers. We also think AmEx

provides the identical transaction services as Visa or MasterCard. AmEx argued that it provides

a higher class of services than Visa and MasterCard, by offering better consumers reward program

with greater benefit, but at the expense of a significant higher swipe fee to merchants. However,

the fundamental service provided by AmEx is nothing more than a means of payments, and the

only difference with Visa and MasterCard is the price level and structure charged to consumers

and merchants. Price level means the total prices borne by end-user groups, and price structure

means the decomposition of price level. In short, the transaction services that AmEx provides is

still the same as that of Visa and MasterCard. So, Cournot model is a proper representative of

competition between credit card companies.

We assume both consumers and the merchant have their own transaction benefits, bB and bS .

Platforms charge customer fee pB and merchant fee pS to consumers and merchants. Consumers

have an option to choose their method of payments. Merchants also have a choice regarding accept

cards or not. The platforms provide the infrastructure that enables consumers and merchants

making transactions with cards. By accepting credit cards rather than cash, merchants enjoy

various transaction benefits, including fraud protection, accounting facilities, time savings at the

counter, and transaction enablement through credit or float. Likewise, consumers enjoy their own

transaction benefits by paying with credit card rather than cash, i.e., conveniences related to

eliminating the need to carry cash, the need to count cash, and the risk of losing cash, and etc.

The assumption on merchants is critical, and it shall capture two elements of realities in credit
4
See David S. Evans, 2011, Interchange Fees: The Economics and Regulation of What Merchants Pay for Cards
p.63

7
card markets. First, the must-take argument as in Vickers (2005) and MI as in Rochet and Tirole

(2002) and Farrell (2006) should prevail under NDPs. Second, excessive surcharges should be the

case as the observations in various counties, which are documented in the report “Review of Card

Surcharging: A Consultation Document” June 2011, Reserve Bank of Australia. To meet the two

requirements, we adopt the monopoly merchant setting as in Wright (2012). Hotelling (including

the Hotelling-Lerner-Salop) differentiated product competition also satisfies the two features, if we

assume the transportation cost is sufficiently large, t > vB (pB )DB (pB ), as introduced immediately.

For exposition purpose, we focus on the case with monopoly merchant.

In particular, the merchant is a monopolist in selling a product to consumers who are located

in an unit distance interval. Consumers are uniformly spread over the interval. Each consumer

obtains utility u from purchase the product, and incurs a transportation cost t per unit distance

to merchant’s store. The consumer only realizes her benefit bB from using cards after going into

the store. bB is drawn from the uniform distribution over [0, 1]. We assume the full coverage of

the product market. With NDPs, the monopoly merchant will set the price of product p equal

to p = u + vB (pB )DB (pB ) − t—the utility of consuming the product u plus the consumer surplus

from using cards minus the transportation cost of the consumer who is located at x = 1. The term

average consumer surplus vB (pB ) is defined as vB (pB ) = E (bB | bB ≥ pB ) − pB .5 vB (pB )DB (pB )

is the total consumer surplus of consumers. When NDPs are removed and surcharges are allowed,

a monopoly merchant sets the price of product purchased in cash as pcash = u − t, and sets the

price of product purchased with credit cards as pcard = u − t + s. The term s is the level of

surcharge. Consumer’s demand for transaction services is given by DB (pB ) = 1 − pB with NDPs,

and DB (pB ) = 1 − pB − s when otherwise.

MI implies the merchant is able to extract all consumer surplus of using cards vB (pB )DB (pB )

by inflating the price of product. However, consumers with realized benefit less than the price

bB < pB would not purchase the product, even with cash. These consumers (consider the one

located at x ) will obtain net utility from purchase with cash is u − p − tx, where the price of

product p = u + vB (pB )DB (pB ) − t. Then, they obtain net utility −vB (pB )DB (pB ) + t − tx. If

instead of not purchasing, they would incur a transportation cost of tx or net utility of −tx. These

consumers will purchase if condition t > vB (pB )DB (pB ) holds. Consequently, MI also holds. It
R∞
5 pB (bB −pB )dF (bB )
Mathematically, the term is given by vB (pB ) = E (bB | bB ≥ pB ) − pB = R∞ , where F (bB ) is the
pB dF (bB )
cumulative distribution function of bB .

8
is worthy mentioning that the full coverage of the market implies that the demand of product is

inelastic. This is to say, the inflated price of product does not affect the demand. Condition of MI

can also be met when the demand is elastic as explained in the appendix A of Wright (2012). For

exposition purpose, we confine discussion to the case with inelastic demand.

The timing of platform is given as follows. First, platforms or networks set the per transaction

fees pB and pS to consumers and the merchants, and determine whether to adopt the NDPs. In a

closed system, AmEx sets pB and pS directly. In an open system, the IFs and two network services

fees are set by credit card companies and their affiliated banks. Once the three fees are set, pB

and pS are determined consequently. Second, merchants makes a choice to accept the card or not,

given pS . Third, each consumer first goes to a store and realizes her transaction benefits bB . If

she finds the merchant store accepts card payment, she then decides to use a card or not given

pB .6 We use backward induction to solve the game. In the first stage of game, the platform sets

pB and pS ; In the second stage, the merchant makes an option to card services and set the prices

of product, and consumers decide to purchase and the means of payments. Notably, there are two

markets. One is the market of the product, and another is the market for credit card transaction

services. Consumers have their preferences over both the product and transaction service. Under

our assumption, there is no interdependence between the two markets, i.e., the price of transaction

service does not affect the consumption of product, and vice versa. In this way, we could focus on

credit card markets, without losing the practical background of product market.

2.1 The case of monopoly platform

This section discusses the case with monopoly platform. We start by discussing the merchant’s

decision of card adoption and pricing behaviors. The equilibrium condition for a monopoly merchant

under the NSR is well introduced and proposed in the proposition 1 of article Rochet and Tirole

(2011) and the appendix A in Wright (2012). For brevity, I shall use their results directly and

explain the intuitions behind each equation. For a given pS , the merchant has two options – either
6
Notably, we omit the card adoption decision of consumers, which reflects the majority of consumers do carry
a card in the real world. Alternatively, we can understand as the consumers compare both membership and usage
fees with excepted realized benefits for each transaction, and membership fees are evened out for each purchase.
Moreover, membership fees, which account for around 1.5 percent of networks and banks revenue, is insignificant
compared to 15 percent for usage fees. Although, networks may incur some cost to acquire an consumer, but the
cost is still neglected, compared to an average $ 400 yearly revenue from each household. For the above reasons, we
assume every consumer has a certain type of credit cards.

9
accept the NDPs and card transactions, or decline cards and merely accept cash.

If the merchant accepts the NSR clause, she will set the uniform price of product as

p = u + vB (pB )DB (pB ) − t (1)

Intuitively, u − t is maximum price the merchant can charge for the product to ensure the consumer

located at furthest spot purchase the product. The merchant will charge vB (pB )DB (pB ) more on

top of “pure” product price to fully extract the consumer surplus. If the merchant declines card

payments, she will set the price of product as

p=u−t (2)

The merchant will accept cards if

bS + vB (pB ) ≥ pS (3)

The merchant will obtain transaction benefit of bS and extra average consumer surplus vB (pB ) for

each transaction, and pay a cost of merchant fee pS if accepts cards. If the net benefit is greater

than the cost, the merchant will accept. The merchant could also refuse card payments, and the

ability to decline is referred to as merchant resistance. For a given pS , some merchants will accept,

while other turn it down. The extent of merchant resistance hinges on the amount of consumer

surplus can be expropriated, and her own transaction benefit bS . The equation (3) is the condition

of MI, implying the merchant fully consider the consumer surplus when accept a card payment.

Taking into account the merchant’s reaction, the platform will set the highest merchant fee pS such

that all merchants accept the card, i.e.,

pS = bS + vB (pB ) (4)

In doing so, the platform ensures merchant weakly prefer to accept cards, and at the mean time

seizes all consumer surplus extracted by the merchant. Without loss of generality, we assume

merchants accept cards when she is indifferent between payments by cards and cash. Therefore,

10
the monopoly platform profit maximization is given by

max π N = (pB + pS − c) DB (5)

DB

where pB = 1 − DB and pS is given by equation (4). Alternatively, the problem is

max π N = (1 − DB + bS + vB (1 − DB ) − c) DB (6)
DB

When NDPs are removed and surcharges are allowed, the merchant will accept cards regardless,

because she cannot only pass through the cost onto consumers but also impose excessive surcharges.

In this situation, the merchant will set cash price as pcash = u−t, and card price as pcard = u−t+s.

The profit maximization of merchant is given by

π M = (pcash − d + bS − pS ) DB + (pcash − d) (1 − DB ) (7)

After substituting pcash = u − t, DB = 1 − pB − s, the problem is

maxπ M = u − t − d + (bS + s − pS ) (1 − pB − s) (8)

s

Given pB and pS , the merchant select a surcharge to maximize her profit, and the solution is given

by
1 − bS − pB + pS
s= (9)
2

The inverse demand function is given by

1
pB + s = 1 − DB = (1 − bS + pB + pS ) (10)
2

By rearrange equation (10), we obtain the price level as

pB + pS = 1 + bS − 2DB (11)

Notably, since the merchant always pass the cost onto consumers, the fee structure becomes neutral

as discussed in Gans and King (2003), and only the fee level matters. This point can be verified

11
in equation (10), in which the final price borne by consumer pB + s is determined by the fee level

pB + pS , regardless of the decomposition. Anticipating the response function of the monopoly

merchant, the platform’s profit maximization problem is:

max π N = (pB + pS − c)DB = (1 + bS − 2DB − c) DB (12)

DB

The solutions to platform’s problems under either case, i.e., (6) and (12), can be easily obtained,

which are given as follows. With the NDPs :7

1
DB = 1 + bS − c, pB = c − bS , pS = bS + (1 + bS − c) , π M = 0 (13)
2

1 1
πN = (1 + bS − c)2 , CSCard = 0, T SCard = (1 + bS − c)2 (14)
2 2

When surcharges are allowed:

1 1 1
DB = (1 + bS − c) , pB + pS = (1 + bS + c) , π M = (1 + bS − c)2 (15)
4 2 16

1 1 7
πN = (1 + bS − c)2 , CSCard = (1 + bS − c)2 , T SCard = (1 + bS − c)2 (16)
8 32 32

With NDPs, we observe that consumer surplus is zero. This is because the inflated price of

product extracts all of the surplus, and the extraction depends on the assumption of inelas-

tic demand of product, and significant sunk cost of transportation. Once a consumer visits

one store, she will purchase either with a card or with cash, due to a sufficient high reserve

price u and significant transportation cost. Cash-using consumers also pay the inflated price

p = u + vB (pB )DB (pB ) − t compared to the otherwise price p = u − t. In this way, cash users

cross-subsidize card users, at the expense of their benefit from consuming product by amount of

(1 − DB (pB ))vB (pB )DB (pB ). Card users obtain total transaction benefits as vB (pB )DB (pB ), and

they pay the total extra fee via inflated price as DB (pB )vB (pB )DB (pB ). The net benefit of card

users gained is (1 − DB (pB ))vB (pB )DB (pB ), which is exactly equal to the loss of cash users. As

a result, the total surplus of all consumers is zero. Expecting the merchant’s full expropriation

of consumer surplus, the platform charges a merchant fee such that all the surplus is transferred
7
In fact, the profit of merchant is given by π M = u − t − d, which is the total profit from both product and card
market. For illustrative purpose, we only list the profit in card market in this article.

12
into her profit. Therefore, the merchant earns no benefit from the card market. All the surplus is

accrued to platform’s profit. Proposition 1 characterizes the results as follows:

Proposition 1. When the NSR clause is in place and MI prevails, a monopolistic platform will

inflate the price of product by vB (pB )DB (pB ), such that cash users cross subsidize card users by

amount of (1 − DB (pB ))vB (pB )DB (pB ), which results in the platform seizes all surplus in card

market, rendering zero benefits for both consumers and merchant.

Interestingly, the price pB = c − bS effectively implies a first best solution (i.e., summed benefit

of marginal users equals the cost bB + bS = pB + bS = c) and social welfare is maximized. The

card service is therefore just provided, and neither under- nor over- provision exists. The existence

of MI profitably lowers the price to pB = c − bS , compared to price pB = 12 (1 + bS − c) absent MI

condition. The platform does not bear the infra marginal loss when increases the output, and she

is able to profitably lower price pB beyond the monopoly pricing.

The caveat is that some rents are transferred from product market into the card transaction

market. Given consumer’s insensitive product demand response to the inflated prices, the NDPs

and excessive merchant fees render a shift of welfare of consumer’s surplus (cash users in particular)

from products into the rent of card transaction market which is ultimately absorbed by networks.

The shift of the surplus subsidizes the infra marginal loss due to the decrease in price pB , and the

platform does not suffer a loss of infra margin, when expand its output.

It is counter intuitive that a monopolist will profitably lowering price to the cost and expand-

ing output to social optimal level. But, it is the case in a two-sided credit card market where

NDPs short-circuit the ordinary pricing setting mechanism. This countering intuition is not well

understood among academia, let along policy makers. Proposition 2 summarizes this finding.

Proposition 2. When NDPs are in place and MI prevails, a monopolistic platform will profitably

lower price to the cost and expand output to social optimal level, and social welfare is maximized.

When NDPs are removed, MI goes away since the merchant always accept cards and is able

to impose a surcharge. In this situation, the two-sided credit card market turns into a one-sided

one. The fee structure is neutral. The platform does not concern about how to adjust the fee

structure to balance two-sides of the markets, and all it need to worry about is the fee level, which

determines its profit. The model is thus turning into a simple vertical supply chain story. We

13
can see the point from the fact that inverse demand pB + s is solely determined by the price level

pB + pS . By rearranging the profit functions of merchant and platform (e.g., plugging equations

(9) and (10) into equations (8) and (12)), we can see both functions are determined by the price

level as well.

The results without NDPs show that consumers obtains positive surplus, both platform and

merchant earns a positive profit from the card market. It is a simple double marginalization story of

a vertical supply chain. The platform is the upstream supplier and the merchant is the downstream

seller, and each impose a margin onto the cost of card services. Although the surcharge cannot

be solved due to indetermination of fee structure, we can still obtain an expression of excessive

surcharge as

1 − bS − pB + pS 1 − bS − pB − pS 1 − bS − 12 (1 + bS + c) 1 − 3bS − c
s−pS = −pS = = = (17)
2 2 2 4

3
The extra charge is part of fee borne by consumers: the total fee borne is pB + s = 4 (1 + bS − c),

and total fee borne minus the extra charge equals price level, i.e., pB + s − (s − pS ) = pB + pS . The

sign of extra charge is ambiguous, but with a higher likelihood to be positive when the maximum

transaction benefit of consumers is greater than the transaction benefit of merchants. After banning

the NDPs, we can see that: (a) the card demand decreases; (b) the price level remains the same;

(c) merchant’s profit increases; (d) platform’s profit decreases; (e) Consumer surplus from using

cards increase; (f) total surplus, including the consumer surplus, merchant’s and platform’s profits,

decreases. The results are proposed as follow:

Proposition 3. When a monopolistic platform provides card transaction services, the prohibition

of NDPs will decrease card usage, level off price level of platform, leading to an increase of consumer

surplus and merchant’s profit, and a drop of platform profit and social welfare.

The removal of NDPs eliminates the concern of price structure, and paralyzes one side of the

market. It disables platform’s perfect expropriation of consumer surplus via an excessive high

merchant fee and hence the inflated price of product. At the mean time, it allows the merchant

to impose a surcharge. No matter how a second margin adversely affect the efficiency, it will not

absorb all the consumer surplus. Although total surplus decreases afterwards, both consumers and

the merchant gain benefits from the injunctive relief. The conflict between platform and merchant

14
is obvious: the merchant’s profit increase afterwards, while the profit of platform decreases. The

conflict in profitability can be manifested from lawsuits in the U.S., whereby merchants are on

plaintiff side, and platforms are defendant. It is the conflict in profitability, not the claimed social

welfare or consumer surplus, that brings about the past or current lawsuits in practice.

2.2 Application to AmEx case

The findings of the monopolistic platform can be applied to analyse AmEx’ NDPs. Specifically,

the findings give full theoretical support to plaintiffs in State of Ohio et al., v. American Express,

585 U.S. [2018] . In this lawsuit, both Appeal and Supreme Courts conclude that plaintiffs have

not dismissed the burden to show anti competitive effects of NDPs in the first step of rule of reason

analysis, given the relevant market is defined as the market of card transactions.

The Supreme court decides in favor of AmEx primarily relies on the following grounds. First

of all, it claimed that the district court ” wrongly focuses on only one side of the two-sided credit-

card market”. Second, it pointed that plaintiffs failed to provide preponderant evidence that the

price of card transactions was higher than competitive level. Third, the experienced expanding

output and quality counter proved the existence of market power, and that the NDPs had stifled

competition. However, our model shows, in a two-sided market scenario, NDPs harm both merchant

and cardholder consumers, despite the fact that they boost the demand of card transactions, which

is one of primary arguments Supreme Court counts on to reach the decision.

To fit the model into discussion of NDPs of AmEx, we could regard AmEx as a monopoly in

providing premium charge card transaction service as in travel and entertainment (T & E ) or

corporate cards sectors, and regards the alternative means as the Visa or MasterCard cards or

other payment means. The merchant fees charged by the alternative platforms are normalized to

zero. AmEx managed to charge a premium over its rivals’ all-in prices to merchants, and it is

called premium pricing strategy. The model shows that this strategy would inflate the price of

goods overall, resulting in a full extraction of consumer surplus on cardholder side. The alternative

cardholders cross subsidize the AmEx cardholders. Likewise, the merchants obtain no benefit from

card transaction service, despite that total social welfare in maximized. Nevertheless, nothing is

enjoyed by the consumers of both sides, even in the events that card usage is expanded.

The Supreme court also thinks that the higher merchant fees collected by AmEx could be used

15
to fund the innovation to improve cardholder’s benefit. Our theory indicates that whether the

NDPs affects innovation on cardholder benefit is a pseudo-question. So long as all the created

benefits of cardholders are taken away, it does not matter how much benefits are realized from

innovation. The courts also discuss the issue of entry barriers in the industry. They recognized

that “the credit-card industry continues to be characterized by formidable barriers to entry. These

barriers arise because of the nature of the industry and requirements a network must fulfill before

entering it.” But what they do not understand is that NDPs are the vital premise to foster the

nature of formidable barriers. In particular, NDPs bestows the credit card markets the nature of

two-sidedness. An entrant enters the market has to accord both sides, and conquer the critical mass

or chichen & egg problem. Once NDPs are prohibited, merchants always accept card payments.

The market degenerates into a one-sided one, and entry is relative easier. So, NDPs do increase

the barriers to entry.

2.3 The case of duopoly platforms

In this section we will apply a model of Cournot platforms to mimic the competition between Visa

and MasterCard. Since it is the NSR clauses at issues in case of Visa or MasterCard, we will only

mention NSR clauses in this section. We first consider the case where issuing and acquiring banks

are both perfect competitive, and then relax this later on. Both type of banks then play no real

role in the game, and the game of platform competition is similar to that of monopolistic platform.

The only difference is that platform competition in quantity in first stage nails down the price

borne by consumers. In particular, under the NSR clauses, the two platforms set the merchants fee

according to equation (4), The monopoly merchant accepts cards and sets the price of the product

as equation (1). On the consumer side, the two platforms compete according to Cournot game.

Take into account the effect of MI , the profit maximization of two platforms, denoted by I and J,

are given by

max π I = 1 − q I − q J + bS + vB (pB ) − c) q I


(18)
qI

max π J = 1 − q I − q J + bS + vB (pB ) − c q J


(19)
qJ

where q I and q J are the quantities of card service provided by each platform. q I + q J is the total

industry sale, which is equal to DB (pB ). Likewise, the platform’s problems under surcharging

16
regime are given by

max π I = 1 + bS − 2 q I + q J − c q I



(20)
qI

max π J = 1 + bS − 2 q I + q J − c q J



(21)
qJ

We can easily solve problems expressed by equation (18) to (21), and the results under NSR clauses

are given as follows:

4 4 1 2
DB = (1 + bS − c), pB = (c − bS ) − , pS = bS + (1 + bS − c) , π M = 0 (22)
3 3 3 3

4 4
πI + πJ = (1 + bS − c)2 , CSCard = 0, T SCard = (1 + bS − c)2 (23)
9 9

and the results when surcharges are allowed are given by:

1 1 1
DB = (1 + bS − c) , pB + pS = (1 + bS + 2c) , π M = (1 + bS − c)2 (24)
3 3 9

1 1 5
πI + πJ = (1 + bS − c)2 , CSCard = (1 + bS − c)2 , T SCard = (1 + bS − c)2 (25)
9 18 18

The results show that the consumer surplus and merchant’s profit from card market remain zero

under the NSR clauses. All the surplus are accrued to the two platforms equally. Interestingly, it

can be easily verified that the zero surplus result for both consumers and merchant holds regardless

of how many platforms compete. Due to the ability in extraction of consumer surplus, each platform

knows she will not bear the infra marginal loss due to an increase in output, and competes in a more

aggressive way compared to the situation otherwise. The degree of competition among platforms

will not change the fact of depletion in users surplus, so long as MI prevails. We formally conclude

in the following proposition.

Proposition 4. When the NSR clauses is in place and MI prevails, competing platforms absorb

all the surplus from credit card markets.

Compared to the case with a monopoly platform, less total surplus are realized, because the

competition between platforms is unable to internalize the negative externality from one’s sale to

another. The card service is over provided, exceeding first best solution. In the monopoly case, the

over provision distortion from MI is fully offset by the under provision distortion due to market

power of platform. In the duopoly case, the effect of MI dominates that of market power, which

17
leads to an over provision result.

Under the surcharging regime, it is still a double marginalization story. Compared to the case

of monopoly platform, more total surplus is realized, both consumer surplus and merchant’s profit

increase, and the total platform profits decrease. The competition upstream suppresses down the

whole sale price, and both merchant and consumers gain a benefit from it. The welfare implication

discussion remains the same as in last section. The collusive equilibrium of this Cournot game is

equivalent to those of monopolistic platform setting, and welfare implication remains unchanged.

So, we have the following:

Proposition 5. When duopoly platforms provide card transaction services, the prohibition of NSR

clause will decrease card usage leading to an increase of consumer surplus and merchant’s profit,

and a drop of platform profit and social welfare.

The analysis can be easily extended to oligopolistic competition between platforms. The com-

petitive effects for consumers and merchant remain the same: lifting NSR clauses will unambigu-

ously benefits both consumers and the merchant. However, the social welfare may go up when the

competition between platforms is sufficiently vigorous. It can be proved that when N ≥ 4, ban of

NSR clause will increase social welfare. This is because MI strongly overwhelms the market power

which declines as the number of platforms increases, and it incurs a much higher over provision in

output compared to the amount of under provision if MI is curbed.

2.4 Issuing margins

This section will relax the assumption of perfect issuing competition and discuss its implication. In

particular, we assume the issuing banks retain a constant margin m for each transaction. That is, a

consumer will bear a final price of pB +m with NSR clauses, and the demand of card services is then

given by DB (pB ) = 1−pB −m. Under surcharging regime, the demand is DB (pB ) = 1−pB −s−m.

The duopoly platforms will set their fees, taking into account the issuing margin. Their problems

under either scenario are given by

max π I = 1 − m − q I − q J + bS + vB (pB ) − c) q I


(26)
qI

max π J = 1 − m − q I − q J + bS + vB (pB ) − c q J


(27)
qJ

18
where q I and q J are the quantities of card service provided by each platform. q I + q J is the total

industry sale, which is equal to DB (pB ). Likewise, the platform’s problems under surcharging

regime are given by

max π I = 1 + bS − 2m − 2 q I + q J − c q I



(28)
qI

max π J = 1 + bS − 2m − 2 q I + q J − c q J



(29)
qJ

The solution with NSR clauses is given as follows:

4 4 1 m 2
DB = (1 − m + bS − c), pB = (c − bS ) − + , pS = bS + (1 − m + bS − c) , (30)
3 3 3 3 3

4
π M = 0, π I + π J = (1 − m + bS − c)2 , (31)
9
4m
CSCard = 0, π Issuer = (1 − m + bS − c) , (32)
3
4 4m
T SCard = (1 − m + bS − c)2 + (1 − m + bS − c) (33)
9 3

From the results, we can see that the markup of issuing banks increases the customer fee, decrease

card provision, and brings down the merchant fee. Neither consumers nor the merchant benefits

from the card services. All the surplus is still accrued to platforms and issuing banks. The effect

of markup on the total surplus is ambiguous. It depends on to what extend the distortion due

to the issuing markup corrects the over provision due to the MI. As we can see that, the markup

indeed weakens the distortion of MI, which is manifested from the lowered merchant fee. The

markup increases customer fee borne by consumers and lowers the realized consumer surplus, and

platforms have to lower the merchant fee when they expect the lower surplus can be expropriated

via the inflated price of product. The effect of issuing markup when surcharges are allowed is

straightforward and the solution are given as below. From the results, we can see that issuing

markup lowers the demand, consumer surplus, profits of merchant and platforms and total surplus.

1 1
DB = (1 − m + bS − c) , pB + pS = (1 − m + bS + 2c) , (34)
3 3

1 m
πM = (1 − m + bS − c)2 , π Issuer = (1 − m + bS − c) , (35)
9 3
1 1
πI + πJ = (1 − m + bS − c)2 , CSCard = (1 − m + bS − c)2 , (36)
9 18

19
 5 m
T SCard = (1 − m + bS − c)2 + (1 − m + bS − c) (37)
18 3

In the presence of market power of issuing banks, thee distortions exist under NSR clauses,

including the market powers of platforms and issuing banks, and the MI. The presence of issuing

market power will weaken the dominant effect of MI, and decrease the likelihood that ban will

increase social welfare. The results are formally presented as follows:

Proposition 6. When duopoly platforms provide card transaction services and issuing banks pos-

sess market power, the prohibition of NSR clause will decrease card usage leading to an increase of

consumer surplus and merchant’s profit, and a drop of platform profit and social welfare.

2.5 Merchant resistance

The merchant resistance describes the extent to which merchants could decline the card services.

The higher degree of resistance, the less intention to “must-take” a card. Alternatively, it refers

to the ability of merchants to refuse the card services. The resistance hinges on two factors –

the transaction benefit and the degree of MI. The higher value of transaction benefit, the less

likelihood for merchants to turn down card services for a given merchant fee. Similarly, the more

consumer surplus from card services can be extracted through an inflated price of product, the less

willingness for a merchant to decline card services. Since the transaction benefit is fixed, the major

source of merchant resistance arises from MI. Next, we will examine how the degree of MI affects

card provision. Suppose, the merchant can merely expropriate α portion of consumer surplus.

This can be the case for a monopoly merchant when transportation cost is not sufficient high, i.e.,

t = αvB (pB )DB (pB ). For Hotelling merchants, this situation arise when consumers are not fully

informed of card acceptance policy as in Rochet and Tirole (2002). When α portion of consumer

surplus can be extracted, the maximum merchant fee can be charged is

pS = bS + αvB (pB ) (38)

Solving the problem yields the results as follows:

2(1 + bS − c) 2(1 + bS − c)
DB = α , pB = 1 − , πM = 0 (39)
3(1 − 2 ) 3(1 − α2 )

20
 2 1
CSCard = 0, T SCard = π I + π J = (1 + bS − c)2 (40)
9 (1 − α2 )

From the results we can see that the demand, merchant fee, platform’s profits and the total surplus

decrease compared to the fully MI case. The low degree of MI does not yields any “leftover” surplus

to either consumers or merchants. Both groups still earn zero surplus from card transaction. The

platforms will extract any surplus they spot in credit card markets, and it is does not matter how

much they have spotted. Since MI does not exist under the surcharging regime, the results remain

unchanged.

2.6 Regulation of interchange and merchant fees

The regulation of interchange fees (IF) in open systems and merchant fees in closed system is

another critical issue in policy enforcement. Two main theories exist on this regulatory issue. First

is the cost-based approach initiated in Baxter (1983) and reinforced in Tan (2020), by which the

interchange (or merchant) fees should be set according to issuing and acquiring costs. Baxter points

out an appropriate level of IF can solve a market failure due to negative externality imposed by

consumers onto merchants. Tan (2020) advances Baxter’s idea with heterogeneous consumers and

merchants. It also points out that tourist test is an invalid approach when regulate the IFs or

merchant fees in a closed system with heterogeneous merchants. These two pieces discuss the very

issue in a model with single network, without considering the case with compete networks.

The second is tourist test approach introduced by Rochet and Tirole (2011), which is aimed

at capping IFs such that merchants do not have to bear excessive fees because of strategic con-

sideration. It is in merchant’s interest to accept the cards ex ante to attract customers to their

stores, and steer to other payments method ex post to avoid an excessive high merchant fee due

to business strategic consideration. The tourist test insures merchants have the same incentive to

take cards both ex ante and ex post. Bourguignon et al. (2019) suggests that it would be optimal

to either keep the NSR clause and regulate IFs according to tourist test, or abandon the NSR and

free IFs. In these two literature, the networks are assumed to be a not-for-profit association, which

maximizes the transaction volume through its network. While, we assume networks are for-profit

companies. In what follows, we will examine the optimality of IFs and merchant fees with compet-

ing networks and under either the NSR or surcharging regime in attempt to see the optimal policy

intervention.

21
 For illustrative purpose, we will discuss in context of duopoly platforms model. First of all,

the regulation of fees under surcharging regime is not an issue since the price structure is neutral.

The only regulatory issue is with NSR clauses. Intuitively, the outcomes when fees are limited

according to tourist test with NSR clause is identical to the outcomes when surcharges are allowed

and capped by net merchant’s cost – merchant fee minus transaction benefit. This is because, in

either case, only the distortion due to networks’ market power exists, which leads to an identical

provision and social welfare in the card market. Therefore, we have the following proposition:

Proposition 7. Regulating the interchange (or merchant ) fees by tourist test under the NSR

clause, yields the identical outcomes in consumer surplus, merchant profit, and platform’s profits

as when surcharges are allowed and limited to merchant net cost.

Since 2013, surcharges are allowed in U.S. within both Visa and MasterCard networks, with a

cap equal to the least of average merchant’s fees and a 4% of transaction value. In Canada 2017, Visa

and MasterCard agreed a settlement to modify their restraint of NSR clause, and allow merchants

to surcharge. Likewise, the excessive surcharges were banned since 2017 in Australia. No standard

of the cap has been stipulated (like 4% in the U.S.). Instead, the guidelines of Competition and

Consumer Amendment (Payment Surcharges) Act of 2016 suggests that the cap should be based on

merchant’s cost to accept card payments. Theory suggests that a cap equal to net merchant’s cost

is optimal. In practice, the cap is set according to gross merchant’s cost to accept card payments,

which does not net out merchant’s benefit. So, the mandatory cap leave some profit margin to

merchants, and they will make a positive profit under the surcharging regime.

As mentioned in theory, consumers earn no benefit with NSR clauses, and ban of it will unam-

biguously improve consumer surplus, regardless of curbing surcharges or not. Merchant also make

no profit under the NSR, and make a positive under otherwise. The cap on surcharges will lower

its profit but improve the consumer surplus. The profit of networks decrease after the ban. Total

surplus could go both directions. According to the above proposition, the conclusion on welfare

implication of NSR clause also applies to the impact of tourist test. Specifically, tourist test in-

creases consumer surplus in certain. Merchants, however, gains nothing from it, and always make

zero profit. The network’s profits decrease. The total surplus remains ambiguous. Therefore, we

have the proposition:

Proposition 8. The regulation of the interchange (or merchant ) fees with tourist test under NSR

22
clauses will increase consumer surplus, decrease platform’s profits, and level off merchant’s profit.

The total welfare effect is ambiguous.

Whether application of tourist test (equivalently ban of NSR clauses and impose optimal sur-

charge cap ) improves the total social welfare hinges on the relative forces between market power of

networks and merchant internalization. This point is clearly demonstrated in the model of oligopoly

networks. We can easily extend the model of duopoly competition into oligopoly competition with

N networks. For brevity, we omit the process of derivation, rather states some of results. when

N = 1, the monopoly network will supply Q = 1+bS −c unit of services and the price is pB = c−bS ,

which is the social cost to provide the service. In this case, the market power distortion of network

fully offsets the distortion of MI, and the card provision is optimal and social welfare is optimized. If
(1+bS −c)
the fees are curbed by tourist test, the quantity is Q = 2 and the price has pB > c − bS . The

tourist test curbs the over-provision of MI and renders an under provision result. When N = 2, the

market power distortion is weaker than that of MI, and card services is over provided with industry
4(1+bS −c)
level quantity as Q = 3 , and price being pB < c − bS . After applying tourist test, the
2(1+bS −c)
quantity is given by Q = 3 , and price has pB > c − bS , which indicates an under provision.

The test eliminates MI and over corrects the distortion causing over provision, leading to a result

of under provision. The social welfare remains the same due to negative and positive displacements

away from the optimal quantity are the same. When N = 3, the test will unambiguously improve

social welfare. This is because MI strongly overwhelms the market power which declines as the

number of networks increases, and it incurs a much higher provision quantity compared to the un-
6(1+bS −c) 3(1+bS −c)
der provision if MI is curbed. The quantities are Q = 4 without a cap, and Q = 4
1+bS −c
with a tourist cap. The former suggests an over provided quantity of Q = 2 , and the latter
1+bS −c
indicates an under provided quantity of Q = 4 . So, we see an increase in social welfare.

2.7 Application to case of Visa and MasterCard

More than 30 countries and regions either had conducted or are conducting antitrust and regula-

tory investigations against Visa and MasterCard on NSR clauses. For example, the Reserve Bank

of Australia (RBA) removed NSR clauses on both networks since 2003, and imposed a limit onto

surcharges as of 2017, which is mandated in the Competition and Consumer Amendment (Payment

Surcharges) Act of 2016. In Canada, the Competition Bureau challenged NSR clauses in 2010, but

23
failed in the case – The Commissioner of Competition v. Visa Canada Corporation and Master-

Card International Incorporated, 2013 Comp.Trib.10. But, the plaintiffs consisting of independent

retailers in five major provinces won and successfully removed the NSR clause in a 2011 class-action

case. Moreover, the Department of Finance imposed a reduction onto interchange fees which takes

effected in 2020.

The issue of NSR clauses in the U.S. is brought about in a separate class-action case in 2005

– In re Payment Card Interchange Fee and Merchant Discount Antitrust Litigation. Merchants

accused the two credit card companies and banks of colluding to inflate their interchange fees and

cause the high fees borne by merchants, and prohibiting from using surcharges to steer consumers

towards other payment of methods. Around 12 million merchants joined in the plaintiff side of the

case, and it lasts more than a decade. In 2013, Visa and MasterCard along with banks (including

JP Morgan Chase & Co, Citigroup and Bank of America) initially agreed to a historical record high

$ 7.25 billion settlement. However, a federal appeals court vetoed the deal in 2016 on the ground

that, among others, the prospective injunctive relief is not met the requirement on behalf of the

representative merchants. The case was finally settled in 2019. In the settlement, both companies

agreed to allowing merchants to impose surcharges, but subject to a limit. In the decade long

process, a number of merchants refused to reach the deal, and some others (like Amazon.com) even

opted out the case and pursuit their own litigation.

Our theory suggests that NSR clauses are detrimental to both cardholder consumers and mer-

chant consumers; and the optimal scheme should be either removing NSR clauses and imposing an

optimal cap on surcharges, or keeping the clauses and imposing tourist test on merchant fees. Our

findings provides solid and yet updated theoretical basis for antitrust practices in various countries.

Although the total welfare implication is ambiguous, the damages to both groups of consumers is

sufficient to pin down the court decision. The only issue remained is to determine the optimal cap

onto interchange and merchant fees, and it is a practical issue.

3 Conclusion

NDPs and NSR clauses have been widely used in a number of markets, for example online shopping,

but received particular attention in credit card markets. The credit card industry is of economic

significant. For example, AmEx in the U.S. cards were accepted at about 6.4 million merchants’

24
stores, compared to the number around 9.4 million for Visa and MasterCard cards(U.S. v. American

Exp. Co., 88 F. Supp. 143, 204, 2015). Likewise, according to the National Retail Federation report,

merchants together with consumers pay $ 50 billion swipe fees a year, amounting to a $ 400 charge

for each household. Globally, around $ 10. 8 trillion value transactions were paid by credit cards,

which accounts more than 10 percent of the world’s gross domestic product (HSN Consultants

2016, p.9).

To understand the competitive effects of these practices is critical among the discussion of public

policies, especially when the ruling in State of Ohio et al., v. American Express, 585 U.S. [2018]

which seminally applies the theory of two-sided markets into the canonical rule of reason analysis

will generate profound influence to similar litigation in other two-sided markets. Unfortunately,

our theory goes against the court’s decision in favor of AmEx. We find that when two sides of the

markets are taken into account, NDPs unambiguous stifle competition and leads to surplus losses

to both cardholder consumers and merchant consumers. The same conclusion also applies to NSR

clauses of Visa or MasterCard when two-sided networks competes.

Our model findings are profound. First of all, we find that the welfare implication of NDPs

hinges on relative distortions induced by market power of networks and merchant internalization

(MI). Previous literature have also identified these two effects. But, we move one step forwards to

characterize the determinants of comparison between theses two effects, especially in an updated

model. We also find both consumers and merchant obtain zero benefit from using a credit card

when NDPs is in place. Lifting the NDPs will unambiguously increases surplus of both consumers

and merchants.

In addition, we find that retaining the NDPs and applying tourist test regulation will yield the

identical results as when allow surcharges and impose an optimal surcharge cap. We also mimic

the competition between AmEx and Visa (or MasterCard), and their different business models

which are shown in the appendix. We manage to establish an equilibrium where AmEx focuses

on a sector with higher evaluation of transaction services, and with consumers spend heavily (as

for charge cards), and Visa serve boht this high value sector and another low value sector, but

could earn a profit of interests from outstanding balance. Interestingly, the asymmetrical business

models leaves some consumer surplus under the NDPs. Other comparison are similar to the above

analysis.

25
References

Aurazo, J. and Vasquez, J. (2019). Merchant’s card acceptance: An extension of the tourist test

for developing countries. Technical report, Graduate Institute of International and Development

Studies Working Paper.

Baxter, W. F. (1983). Bank interchange of transactional paper: Legal and economic perspectives.

The Journal of Law & Economics, 26(3):541–588.

Bourguignon, H., Gomes, R., and Tirole, J. (2019). Shrouded transaction costs: must-take cards,

discounts and surcharges. International Journal of Industrial Organization, 63:99–144.

Bourreau, M. and Verdier, M. (2019). Interchange fees and innovation in payment systems. Review

of Industrial Organization, 54(1):129–158.

Carlton, D. W. and Winter, R. A. (2018). Vertical most-favored-nation restraints and credit card

no-surcharge rules. The Journal of Law and Economics, 61(2):215–251.

Edelman, B. and Wright, J. (2015). Price coherence and excessive intermediation. The Quarterly

Journal of Economics, 130(3):1283–1328.

Farrell, J. (2006). Efficiency and competition between payment instruments. Review of Network

Economics, 5(1).

Gans, J. S. and King, S. P. (2003). The neutrality of interchange fees in payment systems. Topics

in Economic Analysis & Policy, 3(1).

Reisinger, M. and Zenger, H. (2019). Interchange fee regulation and service investments. Interna-

tional Journal of Industrial Organization.

Rochet, J.-C. and Tirole, J. (2002). Cooperation among competitors: Some economics of payment

card associations. Rand Journal of Economics, pages 549–570.

Rochet, J.-C. and Tirole, J. (2011). Must-take cards: Merchant discounts and avoided costs.

Journal of the European Economic Association, 9(3):462–495.

Rysman, M. and Wright, J. (2014). The economics of payment cards. Review of Network Economics,

13(3):303–353.

26
Schmalensee, R. (2002). Payment systems and interchange fees. The Journal of Industrial Eco-

nomics, 50(2):103–122.

Schwartz, M. and Vincent, D. R. (2006). The no surcharge rule and card user rebates: Vertical

control by a payment network. Review of Network Economics, 5(1).

Schwartz, M. and Vincent, D. R. (2019). Platform competition with cash-back rebates under no

surcharge rules. Journal of Industrial Economics, Forthcoming.

Tan, H. (2020). The regulation of mercant fees in credit card markets. Journal of Regulatory

Economics, 57(3):258–276.

Tan, H. and Deng, Z. (2020). The no-surcharge rule and surcharging behaviors in credit card

markets. Australian Economic Papers (forthcoming).

Tan, H. and Wright, J. (2018). A price theory of multi-sided platforms: Comment. The American

Economic Review, 108(9):2758–2760.

Tan, H. and Wright, J. (2020). Pricing distortions in multi-sided platforms. Working Paper.

Verdier, M. (2011). Interchange fees in payment card systems: A survey of the literature. Journal

of Economic Surveys, 25(2):273–297.

Vickers, J. (2005). Public policy and the invisible price: competition law, regulation and the

interchange fee. Competition Law Journal, 4(1):5.

Wright, J. (2003). Optimal card payment systems. European Economic Review, 47(4):587–612.

Wright, J. (2012). Why payment card fees are biased against retailers. The RAND Journal of

Economics, 43(4):761–780.

Appendix: Asymmetrical business model

Thus far, we have discussed the monopoly and symmetrical duopoly cases. This subsection will

apply an asymmetrical competition model to mimic the differentiated business strategies between

AmEx card and Visa (or MasterCard). Historically, AmEx distinguishes itself with its “spend-

centric” business model, in which consumers spending heavily are attracted to its network, and the

27
main source of revenue is collected from the merchant discount fees. A majority of AmEx products

are charge cards, and it has built up an reputation with its best known Green, Gold, and Platinum

charge cards. Charge cards enable consumers to spend up to a limit, and pay back the balance in

full after a grace period with no interests. Even if passes a due day of the repayment, a consumer

will be mostly received a reminder of bill payment, and without worrying about monetary penalty.

AmEx cards’ spending limits vary from cards, and also hinges on consumer’s personal financial

status. Carrying an AmEx card is a social status symbol. On the contrary, Visa and MasterCard

carry a “lend-centric” model, with which around half revenue of networks and their issuing banks

are collected from interests of outstanding balance. Although AmEx’s merchant discount fees vary

from sectors, i.e. airlines, hotels, grocery stores, online retailers and among others, AmEx charges

the same level of merchant discount fees to every type of its cards for transaction in a particular

sector. Merchant discount fees of Visa (or MasterCard) credit cards varies across different types of

cards, with a higher merchant discount fees charged to cards with a greater reward program.

In practice, both AmEx’ charge cards and Visa’s credit cards are categorized into a single

relevant market in the antitrust enforcement. In the State of Ohio et al., v. American Express,

585 U.S. [2018] , the relevant market is defined as the market of General purchase credit and

charge (“GPCC”) cards. The district court agreed with the two-sidedness of GPCC card market,

and regarded it consists of at least two separate, but highly interdependent, markets. The two

separate markets include a card issuing market, in which card networks compete over expenditure

basis of cardholders, and a network services market, in which networks compete over merchant

card acceptance. However, the district court applies the rule of reason analysis only on the network

services market. The Appeal and Supreme court think the relevant market should be the card

transactions market, which is two-sided market and compose of cardholder and merchant consumers.

So, we will investigate the competition between AmEx and Visa(or MasterCard) in two-sided

market setting, in an attempt to provide an explanation to adoption of business models by networks,

and discuss the competitive effects of business clause in question and consequence of enforcement

practices.

Practically, we observe that AmEx charges the average merchant fee around 3 percent transac-

tion value as of August 2019, about 50 percentage higher than Visa (or MasterCard) does.8 Most
8
Data sources: Visa USA Interchange Reimbursement Fees published on April 13, 2019; MasterCard 2019–2020
U.S. Region Interchange Program and Rates; and Wells Fargo Payment Network Qualification Matrix published on

28
consumers having an AmEx card also carry a Visa or MasterCard card, and around 6.4 million

merchants accept AmEx cards in the U.S., despite that around 3 million merchants, who accept

Visa and MasterCard, chooses to decline AmEx cards.9 To mimic the practice, we assume two

distinct merchants with different transaction benefits, i.e., bL H L H

S and bS with bS < bS , representing

two separate sectors, sell products to two groups of consumers. In the sector bH
S , consumers also

have a higher evaluation of card service, and their transaction benefit bH

B is a draw from uniform

distribution over [0, B]. In the sector bL

S , consumer’s transaction benefit is drawn from uniform

distribution over [0, 1]. In sector bH

S , consumers merely utilize the cards as a payment means,

and pay back the outstanding balance at the end of each grace period, as with charge cards. So,

networks can only collect revenue from transaction fees. In sector bL

S , Visa is also able to collect

revenue from its lending business with its credit cards, but AmEx is unable to do so because it

mainly issues charge cards which do not carry a credit function. In this setting, each network has

two options. First, set a low merchant fee and serve two sectors. Second, set a high merchant fee

and serve one sector. The trade off between the two options is a higher margins in sector bH
S , and

a loss of consumers in sector bL

S . Assume this extra profit that Visa could collect from its lending

business is sufficient high such that Visa is always willing to serve the sector bL
S , and hence both

sectors. Notably, this portion profit out of interest does not explicitly enters into the model. But it

will work as an valve in discussion below to sustain an asymmetrical equilibrium, in which AmEx

strategically selects a higher merchant fee and therefore only serve the sector bH
S , and Visa chooses

to serve both sectors. In this situation, AmEx and Visa will charge a merchant fee respectively as

AmEx : pS = pH H H H


S = bS + vB pB (41)

V isa : pS = pH L L L L


S = pS = bS + vB pB (42)

That is to say, Visa charges the merchant fee according to the acceptance condition in sector bL
S,

and AmEx hinging on that in sector bH L

S . To solve the problem, we first consider sector bS . In this

sector, only visa serves and acts as a monopolist. The solutions with the NSR clause are given as

the results in section 2.1, and relisted as in Table 1. In the sector bH

S , AmEx and Visa compete

in quantity. AmEx sets a merchant fee internalizing the consumer surplus in sector bH
S , while Visa

April 12, 2019.

9
U.S. v. American Exp. Co., 88 F.Supp. 143, 204 (2015)

29
has to set a merchant fee accounting for the consumer surplus from the sector bL
S . Their profit

maximizing problems are then given by

AmEx : max π I = B(1 − q I − q J ) + bH H H

I
S + vB p B − c q (43)
qI

V isa : max π J = B(1 − q I − q J ) + bL L L

J
S + vB p B − c q (44)
qJ

where
L

1
vB pL L
B = (1 + bS − c) (45)
2
H

B I
vB pH
B = (q + q J ) (46)
2

The solutions are listed in column one of Table 1. AmEx charges a higher merchant fee than Visa
H pH = 1 (1 + 6B + 4bH + 3bL − 7c), and it


does, and this can be seen from results below where vB B 12 S S
L pL = 1 (1 + bL − c) given B > 1 and bH > bL . However, due to the no steering


is greater than vB B 2 S S S

clause imposed by AmEx, the merchant in sector bH

S cannot price discriminate across networks,

and will set a uniform price of product

H
pH H H L L L H




p = u − t + DB B (σvB pB + (1 − σ)(vB pB + bS − bS )) (47)

qI
Where σ = q I +q J
is the ratio of AmEx’s demand to total demand. As the no steering constraint

clauses imposed, the merchant is unable to price differently steering purchases towards the least

costly payment means, and she has to charge a uniform price of product such that the extra cost
H pH for its sale with


is evened out for each transaction. The merchant will bear extra cost of vB B
L pL + bL − bH for the sale with Visa cards. Therefore, she


AmEx cards, and bear extra cost of vB B S S

spreads out the cost to each unit demand of product.

AmEx, expecting that the merchant’s acceptance, charges a merchant fee such that the consumer

surplus is fulled absorbed. But the Visa charges a lower merchant fee to secure the customers in

sector bL
S , and it makes the fully extraction not occur. AmEx is in a better position to absorb the

surplus from merchant side. This implies that it has lower cost in providing services on consumer

side, and in turn suggests that it will provide a higher quantity of services than Visa does. Solving

the above problems, we obtain results as shown in column one of Table 1.

30
 From the results in the sector bH
S , we can easily verify that, compared to its counterpart, AmEx

charges a higher merchant pH L I J

S > pS , has a greater demand q > q , and earns a higher profit

π I > π I . The consumer surplus above stands for the total realized surplus in sector bH
S . A part

of it is extracted by networks, and the rest of it is retained by consumers, which is equal to (1 −

H pH −(v L pL +bL −bH )). This is because, the consumer surplus for σ portion of purchases



σ)(vB B B B S S
H pH − (v L pL + bL − bH )) retained



is fully expropriated, while the rest 1 − σ portion has (vB B B B S S

by consumers for each unit purchase of payment. Interestingly, the asymmetrical equilibrium

from business models of networks yields some “left over” surplus to consumers – a better position

compared to the case of symmetrical case, where consumer surplus is fully extracted. The consumer

surplus left in sector bL

S remains zero. The merchant makes zero profit from card service in both

sectors. 10

Under the surcharging regime, both merchants accept credit card payments. The duopoly

platforms compete upstream subject to surcharges imposed by the monopoly merchant in each

sector. Specifically, each platform set the same quality to sell to each sector, since it can not

distinguish between the two sectors. The inverse total demands in each sector are given as

1
pH H
B +s = (B − q I − q J ), pL L I
B +s =1−q −q
J
(48)
B

and surcharges downstream in each sector are given as

1 − bL
S − pB + pS B − bH
S − pB + pS
sL = , sH = (49)
2 2

Plugging the above two equations into profit function, we have platform’s problem as

max π I = B + bH I J
− c q I + 1 + bL I J
− c qI





S − 2B q + q S −2 q +q (50)
qI

max π J = B + bH I J
− c q J + 1 + bL I J
− c qJ





S − 2B q + q S −2 q +q (51)
qJ

The solutions are listed in column two in Table 1.

10
Alternatively, both networks can serve two sectors and set the merchant fee at the same level: pS = pL S =
bL + vB pB and earn a identical profit of π I = π J = 81B
L L 1
(2 + 3B + 5bL 2


S S − 5c) . For brevity, I shall not unfold the
discussion on the symmetrical equilibrium, rather compare the profitability under either case and explain why the
asymmetrical arise. Specifically, it can be easily verified that AmEx obtains a higher profit by focusing on one sector,
given Visa serves two sectors. As mentioned above, Visa also prefers to serve two sectors due to its profitability out
of interest in sector bL
S . Therefore, the above asymmetrical equilibrium sustains.

31
 It can be easily verified that both networks earns higher profits under the NSR clause. The

merchant earns positive profit under surcharging regime compared to zero under otherwise. Con-

sumers in both sectors earns a positive surplus. The consumers in sector bL

S are better off under the

surcharging regime. The welfare implication for consumers in sector bH

S is ambiguous. Likewise,

the comparison of total surplus remains unclear, and they are an empirical questions. However,

with the modeling above, practitioners can easily compare the welfare with those parameter values.

The results are summarized in the table 1.

Table 1: The results of asymmetrical business model

Sector bH
S NSR SR
IH 1
(6B + 8bH L 1
1 + B + bH L


q 6B S − bS − 5c − 1) 6(1+B) S + bS − 2c

q JH 1
(1 − 2bH L 1
1 + B + bH L


3B S + 3bS − c) 6(1+B) S + bS − 2c

pIH
S bH 1 H
S + 12 (1 + 6B + 4bS + 3bS
L
− 7c) NA
pJH L 1 L


S bS + 2 1 + bS − c NA
pH
B
1
6B
(−1 − 4bH
S − 3bL
S + 7c) NA
H
ρ NA B + bH
S − 2BDB

π IH 1
72B
(6B − 1 + 8bH L
S − 3bS − 5c)2 NA
JH
π 1
9B
(−1 + 2bH L
S − 3bS + c)
2
NA
MH

2
π 0 B
9(1+B)2
1 + B + bH L
S + bS − 2c

2
CS H 1
72B 3
(1 + 6B 2 + 4bH L
S + 3bS − 7c)
2 B
18(1+B)2
1 + B + bH L
S + bS − 2c

Sector bL
S NSR SR
IL 1
1 + B + bH L


q 0 6(1+B) S + bS − 2c

q JL 1
− 2bH L 1
1 + B + bH L


3B
(1 S + 3bS − c) 6(1+B) S + bS − 2c

pIL
S NA NA
pJL bL + 12 bL


S S 1+ S −c NA
pL
B c − bL
S NA
ρL NA 1 + bL
S − 2DB

π IL 0 NA
JL

2
π 1
2
1+ bL
S −c NA
ML

2
π 0 B
9(1+B)2
1 + B + bH L
S + bS − 2c

2
CS L 0 1
18(1+B)2
1 + B + bH L
S + bS − 2c

Both sectors NSR SR

2
πI 1
72B
(6B − 1 + 8bH L
S − 3bS − 5c)
2 1
18(1+B)
1 + B + bH L
S + bS − 2c

2
2
πJ 1
9B
(−1 + 2bH L 2 1
S − 3bS + c) + 2 1 + bL
S −c
1
18(1+B)
1 + B + bH L
S + bS − 2c

Calculated and simulated by the authors. “NA” implies either the results are not applicable, or the
expression equations are too long to fix it the table.

32

View publication stats