Fees and Surcharging in ATM Networks:
The Role of Nonbanks and Depositor Base
Elizabeth W. Croft*
University of Northern BC
Barbara J. Spencer*
University of British Columbia and the NBER
Revised October 2004
This paper develops a spatial model of ATM networks to explore the implications for banks and
nonbanks of interchange fees, foreign fees and surcharges applied to withdrawals at other than own-bank
ATMs. Account fees are endogenous and consumers have the option of using the branch. Surcharging raises
the price (foreign fee plus surcharge) of a withdrawal above the joint profit-maximizing level. Banks with
more depositors earn less revenue from surcharging and would prefer a no-surcharging agreement, but
nonbanks would always surcharge. The effects of depositor lock-in are contrasted with the strategic use of
surcharges to attract depositors.
Key words: automatic teller machine, surcharges, ATM fees, interchange fees, ISO, lock-in of depositors
JEL codes: L1, G2
* This paper has benefited from presentation at the New York Federal Reserve and at the School of Business,
University of Alberta. We would like to thank Dan Bernhardt and Ralph Winter for particularly helpful
comments. We are grateful for financial support from the SSHRC.
Corresponding author: Barbara J. Spencer, Sauder School of Business, University of British Columbia,
Vancouver, BC V6T 1Z2, Canada.
E-mail addresses: firstname.lastname@example.org; email@example.com
FEES AND SURCHARGING IN ATM NETWORKS:
THE ROLE OF NONBANKS AND DEPOSITOR BASE
The widespread introduction of surcharges on ATM (Automatic Teller Machine) transactions in
North America has become a high profile issue, creating concern among consumer advocates and public
policy makers about disclosure and fairness in ATM fees. Surcharges proliferated after April 1996 when, in
response to pressure from State governments and the courts, both US national ATM networks, Cirrus and
Plus, lifted the prohibition they had maintained on surcharging.1 In Canada, the ban on surcharging by the
Interac Network was also lifted due to the 1996 ruling by the Canadian Competition Tribunal.2 At the same
time the networks in both countries agreed to allow the entry of nonbank ATM owners. These nonbanks, also
called Independent Service Operators or ISO’s in the literature, provide only ATM services. Banks typically
own ATMs and also issue cards so that account holders can access ATMs. As we will argue, network
membership by nonbanks has significant implications in raising the fees for ATM transactions.
In the early 1980s in North America, banks developed networks of shared ATMs in which account
holders could make withdrawals at the ATMs of other banks as well as at the ATMs of their own bank. A
main fee set by the network is an “interchange fee” that banks must pay to other member firms for each
“foreign” transaction made by one of their depositors at another member’s ATM. In order to recover at least
part of the cost, banks typically charge their account holders a “foreign” or “on other” fee for these
transactions. In addition to the foreign fee, account holders making “foreign” ATM transactions may have
1Surcharging was first allowed in1989 by Pulse Electronic Funds Transfer Association as a result of a
binding arbitration ruling. Plus and Cirrus subsequently faced the threat of a legal challenge to their
surcharging ban by ATM owners on anti-trust grounds. The networks were also under pressure from new laws
and regulations in 15 states that permitted surcharging. See Ruud and Webre (1998). In 2001, more than 88%
of banks and savings associations in the United States imposed surcharges. Surcharges vary widely, but the
average is about $1.30 (see Hannan, 2002).
2See the draft consent order issued by the Canadian Competition Tribunal (1996). The Interac network,
which was controlled by large Canadian banks, had specifically prohibited membership by nonbanks.
to pay a “surcharge” directly to the owner of the ATM.3 This fee structure has the interesting and unusual
feature that all three fees apply to the same transaction. Even apart from effects on consumer welfare, this
raises the question as to the differing incentives of network members that drive these three fees.
This paper considers the implications of all three fees (interchange fees, foreign fees and surcharges)
for the incentives facing banks and nonbanks due to asymmetry in numbers of depositors. Account fees are
also determined endogenously.4 We develop a spatial model in which consumers move around the “city” as
represented by a circular network of ATMs. Consumers face a tradeoff between the cost of a transaction if
they travel to an ATM of their own bank and the total price, the sum of the foreign fee and surcharge, that
they must pay if they travel a shorter distance to another ATM. Fitting with our consideration of nonbanks,
the ATMs in the network are not located in bank branches. Consumers choose between use of the circular
ATM network and travel to the branch so as to use a teller. By increasing the elasticity of demand for ATM
transactions, this modelling innovation reduces monopoly power in ATM pricing. For simplicity, depositors
are assumed to pay marginal cost for “on us” transactions at own-bank ATMs and tellers.5
Conflicts arise because asymmetry in numbers of depositors causes asymmetry in the proportion in
which depositors and non-depositors use the ATMs of each member firm. Foreign fees are paid by depositors,
whereas revenues from interchange fees and surcharges are gained from non-depositors. At the extremes, the
network members could be two banks with equal numbers of depositors or a monopoly bank with all the
depositors together with a nonbank ATM owner. If two banks are of equal size then each has access to the
3Hannan (2002) states that in 2001, 78.5% of US banks and savings institutions charged foreign fees
(average $1.17).We abstract from other fees, such as a “switch fee”, ranging from $0.02 to $0.15, which is
paid by network members to cover the cost of routing transactions through the network (McAndrews, 2003).
4Any interest payments to depositors could be viewed as reducing the net level of account fees.
5As shown by Massoud and Bernhardt (2002a), there are efficiency gains to charging depositors marginal
cost for “on us” withdrawals from own-bank ATMs with revenue extracted through account fees. In the US,
“on us” fees are rare. In Australia, after a number of free transactions, large banks charge fees based on cost,
averaging A$0.6 and A$2.50 for an “on us” ATM and teller withdrawal respectively in 2003 (see Reserve
Bank of Australia, 2004). In an earlier version of this paper (Croft and Spencer, 2003), we considered the
effects on ATM pricing of not charging for “on us” withdrawals.
revenues generated by the same number of depositors and non-depositors, whereas a bank with more
depositors gains more revenue from foreign fees, but less revenue from surcharging because there are fewer
non-depositors using its ATMs. For a bank competing with a nonbank, all ATM transaction revenue is due
to foreign fees, whereas the nonbank gains all its revenue from surcharges and interchange fees. Asymmetry
in depositor base also causes asymmetry in the importance of account fees for total profit.
Consumer prices are affected mainly through network-wide decisions as to the interchange fee and
whether or not to allow surcharging. The pricing incentives arising from the removal of a ban on surcharging
and the conditions under which members of the ATM network would make a voluntary agreement not to
surcharge are both examined. Such agreements are not uncommon. An example is the 2001 agreement by a
group of large banks to exempt their customers from surcharges while travelling abroad.6 Reflecting the idea
that network agreements are hard to change, we assume that network-wide decisions are determined by
generalized Nash bargaining prior to any decisions as to other fees. The state of the law as to whether
networks can ban surcharging affects the outside option in this bargaining game.
In order to focus on the effects of asymmetry in numbers of depositors, we assume that banks and
nonbanks deploy the same number of ATMs, which are evenly spaced around the circle and interleaved so
as to maximize consumer convenience.7 There is both theoretical and empirical support for the idea that banks
with greater numbers of ATMs set higher surcharges.8 Nevertheless, if a bank is dominant both in the size
of its ATM network and in its numbers of depositors, high surcharges applied to a small number of non-
depositors would not generate much revenue. Asymmetry in depositor base would continue to affect the
6These were Bank of America, Barclays, Deutsche Bank, Scotiabank and Westpac (see Demers, 2001).
7Since we assume general numbers of ATMs, our analysis includes the effects of network size on the choice
of fees. In Massoud and Bernhardt (2002b), the number of ATMs and their location are determined
endogenously, but are symmetric across banks. Some nonbanks have large ATM networks. In the US,
profitable locations are becoming saturated leading to consolidation of nonbanks. Cardtronics now has a
network of 25,000 ATMs, or about 7% of the ATMs in the US (see Thomson Media, 2004).
8See Massoud and Bernhardt (2002a), McAndrews (2002) and Hannan et al. (2003). Surcharges rise
because banks (and nonbanks) with more ATMs would have greater monopoly power. Also, the strategic use
of surcharges to gain depositors would be more effective for banks with more ATMs.
benefits and costs of surcharging.
For most of the paper we assume that consumers commit to their bank by opening an account prior
to the determination of foreign fees and surcharges at a Nash non-cooperative equilibrium. Thus consumers
see through to all fees in choosing their bank, but surcharges and foreign fees are set taking the numbers of
depositors of each bank as given and hence are not chosen strategically to attract depositors. This order of
moves has empirical relevance if depositors face a significant cost of switching banks. Apart from the need
to fill out forms and buy new cheques, accounts often have features that enhance lock-in, such as the use of
preauthorized deposits and withdrawals. For example, making sure that a pay packet is deposited to the
correct new account can involve significant effort and frustration.9
To show the generality of our results and to distinguish the specific effects of lock-in, we also
examine the order of moves in which foreign fees and surcharges are set strategically to influence the
subsequent choice of banks by consumers. Similar to Massoud and Bernhardt (2002a), we show that under
this “strategic” order of moves, banks increase surcharges above the revenue maximizing level so as to reduce
the desirability of accounts at rival banks. We also show that banks reduce foreign fees so as to make their
own accounts more attractive. Interestingly, when two equal size banks compete, these two effects from a
switch in the order of moves are exactly offsetting, with the outcome that the price (the sum of the foreign
fee and surcharge) of a foreign ATM transaction is unchanged. While banks may want to use surcharges
strategically, they would also want to protect their own depositor base by introducing features of accounts
that create depositor lock-in. Depending on specific institutional features, both order of moves may be
relevant. Nonbanks do not have a motive to set surcharges strategically.
To help understand the efficiency effects of non-cooperative fee setting, we compare the total
consumer price (the sum of the foreign fee and surcharge) arising at the Nash equilibrium with the total price
9There is some evidence of inertia in switching banks. People switch banks mainly when moving to a new
area and smaller banks do not necessarily lose depositors due to surcharges (see Kisser, 2004, Hannan et al.,
2003, and Prager, 2001). However, Massoud et al. (2003) provide evidence that higher surcharges increase
the share of deposits of banks with larger ATM networks. In the market for credit cards, Stango (2002) shows
that consumer switching costs are significant in predicting the fees charged by banks.
that would be achieved by joint maximization of “direct” profit from foreign ATM transactions. Supposing
depositors are locked in, this comparison yields a useful benchmark. If surcharging is banned and if the
interchange fee equals the marginal cost of an ATM transaction, then the foreign fees set non-cooperatively
by network members serve to maximize joint direct profit. In contrast, since bank ATM cards and ATMs are
complementary products, the use of surcharges in addition to foreign fees raises price above the joint profit-
maximizing level under both order of moves.10 Consequently, if interchange fees are not too far above
marginal cost, laws restricting surcharging would lower ATM fees11. We make no claim, however, as to
overall welfare effects. The extra convenience to consumers from the very substantial increase in ATMs
owned by nonbanks in response to the lifting of bans on surcharging, may more than offset high surcharges.12
Full coverage regulations are assumed to restrict account fees to levels at which all consumers open
accounts. We assume that these regulations are binding on two equal size banks, which implies that the
regulations are binding for larger banks, but not for banks with few depositors. Supposing surcharging is
banned, we show that equal size or larger banks would prefer a zero interchange fee if depositors are locked
in and an interchange fee at marginal cost under the strategic order of moves. In the latter case, the
interchange fee maximizes the sum of consumer and producer surplus13. Smaller banks and nonbanks would
10As first pointed out by Cournot, independent firms ignore the effects of their markups on each other,
leading to a sum of prices under independent production of complementary products that exceeds the joint
profit maximizing level. Relating our market structures to Economides and Salop (1992), if the network sets
the interchange fee at marginal cost and bans surcharging, but members set foreign fees non-cooperatively,
this is an example of ‘one-sided joint price setting’. If foreign fees and surcharges are set non-cooperatively
by banks that issue cards and own ATMs, this an example of ‘parallel vertical integration’.
11Massoud and Bernhardt (2002a) argue that a non-discrimination rule with respect to surcharging would
raise ATM fees above the surcharging level by forcing banks to equalize the surcharge and the fee for an
own-bank ATM transaction. We address the different question as to the effect of setting surcharges at zero.
12Since the 1996 Competition Tribunal ruling, over 33% of the ATMs in Canada are “white-label” operated
by nonbanks with high surcharges (see Roseman, 2002). A similar story applies to the U.S.(see U.S. General
Accounting Office, 1998). Based on US data for the period 1994-1999, Knittel and Stango (2004) find that
the net benefits from surcharging tend to be positive in urban areas, but negative in rural areas.
13Our results are consistent with Schmalensee (2002) who argues that, in the context of credit card payment
systems, the privately optimal interchange fee is socially efficient under non extreme assumptions.
prefer a higher interchange fee so as to increase the net amounts transferred from banks with more depositors.
Surcharging neutralizes the interchange fee under both order of moves. Thus an increase in the
interchange fee is exactly offset by an increase in the foreign fee and reduction in the surcharge. Nevertheless,
the introduction of surcharging always benefit nonbanks and tends to benefit smaller banks at the expense
of larger banks.14 In particular, if a monopoly bank and a nonbank compete, all the revenues from surcharging
go to the nonbank. With depositor lock-in, we show that similar size banks would agree not to surcharge, but
a nonbank would never be party to such an agreement.
There is a significant and growing literature concerned with ATM networks (see McAndrews, 2003,
for an interesting survey). The telecommunications industry was a catalyst for the analysis of networks, from
Katz and Shapiro (1985) to more recent work by Laffont, Rey and Tirole (1998). Economides and Salop
(1992) explore the implications of the separate or joint sale of complementary components of a good. Matutes
and Padilla (1994) develop a circular spatial model in which competition for depositors in the presence of
interchange fees and surcharges (called withdrawal fees) gives banks an incentive to make ATM networks
compatible.15 Massoud and Bernhardt (2002a) consider surcharges, account fees and “on us” ATM fees in
a model where two banks, located on a circle, compete for depositors. Asymmetry in bank size is introduced
by assuming that one of the banks has two branches, but interchange fees and foreign fees are not considered.
In Donze and Dubec (2003), banks invest in ATMs so as to profit from the receipt of interchange fees and
foreign fees. By jointly increasing the interchange fee, banks reduce the desirability of attracting depositors.
Profit increases due to an increase in account fees. McAndrews (2002) develops a circular spatial model
involving foreign fees, surcharges and interchange fees so as to examine the effects of differences in numbers
of ATMs and immigration into an area. None of this literature is concerned with nonbanks.
14Massoud et al. (2003) show that in response to the 1996 announcement of surcharge liberalization in the
US, large banks had positive stock market returns, whereas small banks had negative returns. This result may
partly reflect the fact that large banks had greater numbers of ATMs. Also, investors may not have anticipated
the entry of nonbanks, which we argue are the main beneficiaries of surcharging.
15 For compatibility of complementary components between rivals also see Matutes and Regibeau (1992).
The model is developed in Section II. The implications of the maximization of joint network profit
from ATM transactions and comparisons with the Nash equilibrium in foreign fees and surcharges are
developed in Section III. Section IV introduces account fees and explores the conflict between network
members in bargaining over the interchange fee and the conditions under which they would agree not to
surcharge. Section V examines the strategic order of moves and section VI contains concluding remarks.
II. THE MODEL
II(i). Model Overview
There are two firms i for i = B, Z supplying ATM services, where firm B is a bank and firm Z may
be a bank or a nonbank. Consumers each open an account at just one bank. As illustrated in Figure 1, each
firm owns the same general number, denoted M, of ATMs that are spaced symmetrically and interleaved
around a circle, representing a “circular city”. There are N consumers who move around the circle so as to
be uniformly distributed with the same probability of being at any particular location when they make a
transaction, which we assume is a withdrawal. The utility from a transaction at any particular location varies,
but the expected utility around the circle is the same for all account holders of the same bank.
Consumers also have the option of travelling to the branch of their bank to make a withdrawal and,
as well, to conduct other banking business through tellers, such as verifying accounts, exchanging foreign
currency or buying certificates of deposit. The branch or branches are shown as located in the centre of the
circle in Figure 1, but this is not essential for the model. Consumers choose between the ATM network and
the branch based on their expected utility as they move around the circular ATM network, rather than on their
particular position on the circle at any one moment.16 For simplicity, we assume that consumers each make
just one withdrawal and hence the number of consumers serves as a proxy for the number of transactions.
However, even with a fixed number of transactions, by modelling the branch as outside the circular network
of ATMs, the extent of use of the ATM network varies endogenously. Monopoly power in ATM pricing is
16By taking this approach, we rule out the kinked demand curve due to an outside good (see Salop, 1979).
reduced due to the increase in the elasticity of demand for foreign ATM transactions.
Figure 1. Circular Market for ATM Services.
In addition to benefits from transactions at ATMs or tellers, we follow Donze and Dubec (2003) in
assuming that consumers benefit through access to general banking services, such as automatic deposits and
bill payments or the ability to buy and sell bank-owned mutual funds through an internet banking site. Since
we assume the demand for general services is differentiated due to differences in taste, these services provide
a mechanism whereby two otherwise identical banks can set strictly positive account fees. Our model also
encompasses the case of a monopoly bank and a nonbank.
Each bank i specifies an account fee, denoted by Fi, as well as the fees, for own-bank ATM and teller
transactions, denoted by c and cO respectively, which are set at marginal cost. Consumers pay a price, denoted
pi, for foreign ATM transactions. For depositors at bank B, pB consists of the foreign fee, fB, charged by bank
B plus the surcharge, FZ, charged by firm Z if surcharging is allowed. Similarly, if firm Z is a bank, account
holders at Z pay pZ = f Z + FB for a transaction at B’s ATMs.
Assuming depositor “lock-in”, there are four stages of decision. Joint decisions as to the interchange
fee are made at stage 0 based on generalized Nash bargaining with network members fully aware of
consequences for the subsequent stages of the game. Agreements not to surcharge are also made at this stage.
In stage 1, each bank sets its account fee, taking into consideration the effects on consumer choice of banks
in stage 2. After consumers become locked in, foreign fees and surcharges (if allowed) are determined at a
Nash non-cooperative equilibrium in stage 3.17 In Section V, we examine the “strategic order of moves” in
which foreign fees and surcharges are chosen in stage 1 and stage 3 is omitted.
II(ii). The spatial ATM network and consumer demand
As shown in Figure 1 above, the distance between adjacent B and Z machines around the unit circle
is R / 1/(2M) where M is the number of ATMs owned by each bank or nonbank. In moving around the circle,
account holders at bank B can make a transaction at the closest of B’s ATMs, which involves a disutility
equal to the sum of the distance travelled, denoted dB, and the fee, c, for the use of an own-bank ATM.18 They
also have the option of travelling the distance, R - dB, to the closest ATM of firm Z, but this convenience
involves paying the price, pB, for a foreign ATM transaction. Letting x denote the utility of a withdrawal, B’s
account holders gain utility, <H = x - dB - c (superscript H for home) from an own-bank ATM transaction and
utility, <F / x - (R - dB) - pB (superscript F for foreign) from a foreign ATM transaction. Travel to the branch
involves a disutility equal to the sum of the distance, dO, to the branch plus the fee, cO, for a teller transaction.
Letting dB = ÷B satisfying <H = <F represent the distance at which an account holder of bank B is
indifferent between the two ATMs, we obtain ÷B = (R + pB - c)/2 for pB 0 [0, R+ c].19 Thus B’s depositors
prefer B’s ATM for dB # ÷B and prefer Z's ATM for dB > ÷B. The price, pB = R + c is prohibitive in the sense
that account holders only use the ATMs of their own bank (÷B = R), making the shared networks no different
from two separate unlinked ATM networks. Using an analogous argument for Z’s account holders (if Z is
a bank) implies that for pi 0 [0, R + c] and i = B, Z,
17Our results would be unchanged if foreign fees and surcharges were determined at stage 2 simultaneously
with consumer choice of banks. However, the use of stage 3 emphasises the role played by lock-in.
18Units are set so that a one unit increase in a fee or in distance travelled have the same effect on utility.
19If the utility from B’s or Z’s ATM is equal, we assume a customer will choose an own-bank ATM.