The Use of ATMs in Bank Strategy Is There by yre19594


									      The Use of ATMs in Bank Strategy: Is The re a Customer Relationship Effect?


                                      Nadia Massoud*

                                    Anthony Saunders**

                                      Barry Scholnick*

                                    First Draft: May 2003

School of Business, Unive rsity Of Alberta*
Stern School of Business, NYU**

We would like to thank Rob Engle........ for their comments. We also would like to acknowledge
the financial support of SHRC under grant No....., National Research Program on Financial
Services & Public Policy at York university and the Salomon Center for the Study of Financial
Institutions at NYU.
I. Introduction

        Although the number of U.S. banks continues to decline there has been sustained growth

in ATM networks and the number of ATMs. According to Dove (2002) virtually every U.S. bank

is now a member of a shared network (such as Pulse, NYCE and Cirrus). Moreover, the number

of ATMs had grown to over 324,000 as of the end of 2001 (see, Sienkiewicz (2002)). This

proliferation of ATMs has occurred despite apparent complaints by bankers about the fixed and

variable costs associated with the new ATMs added to their networks (Dove, 2002). One

possible reason for the willingness of bankers to keep adding ATMs is that the revenue generated

from these machines, in the form of direct surcharges to non-bank customers (so-called foreign

customers) as well as other fees 1 outweighs the costs of ATM addition. Indeed, since April 1996

banks that are members of shared networks have generally been free to set their own surcharges

for nonbank (foreign) customer use of their ATMs 2 (see, Hannan, Kiser, Prager, McAndrews

(2002)). Thus, one reason underlying ATM proliferation is that foreign (non-bank) customer

surcharges –so called ATM surcharges- have made adding ATMs to a bank’s network profitable,

even in the presence of higher marginal costs. Indeed, revenues from surcharges were estimated

to exceed $2 billion in 2001 alone (Dove 2002). Industry observers and economist have labeled

this the “direct effect” on bank profits resulting from surcharging (see, for example, Dove

(2002), Hannan, et al (2002) and Massoud and Bernhardt (2002a and 2002b)).

        Increasingly, however, bankers and economists are arguing that there is also a second, or

indirect effect, that emanates from ATM proliferation and surcharging. Indeed, one strik ing

  Other fee revenues includes a fee charged to a bank’s own customer who uses another bank’s ATMs (a so -called
foreign fee) as well as interchange fees paid by the customers bank to the ATM owner when the bank’s own
customer uses the owners ATM. The latter fee is usually set by the network and is constant across all banks in a
network. Other fees that may be paid or charged include: own -bank ATM fee (wh ich is rare), POS fee, card fee and
switch fee (see Stavins, 2000 p. 15).
  Until April 1996 major shared networks such as Cirrus and Plus prohibited ATM surcharges on other network
bank customers. Th is surcharge ban was eliminated in April 1996 and surcharges began to proliferate soon after.
Moreover, surcharging was prohibited in a nu mber of states prior to 1996 (Prager 2001).

result of a recent Dove (2002) survey of banks is that over 50% of large financial institutions

recognized that there may be an “indirect effect” or customer relationship effect from ATMs that

can generate additional profits 3 for a bank. This indirect effect has been recognized in the

theoretical papers of Massoud and Bernhardt (2002a and 2002b) and McAndrews (2002).

         While greater ATM proliferation may well attract more customers, due to considerations

of convenience, the indirect impact of surcharges on total bank profits may be less clear. The

argument here is that if consumers are forced to pay higher surcharge fees they face an incentive

to switch to the bank charging the higher fees so as to avoid paying those fees. This is because

only “foreign” customers, who are not account members of that bank, will pay an ATM

surcharge. If this switching behavior occurs, then these customers will presumably purchase a

variety of other bank products, which in turn will increase bank revenue and profits. While a

variety of papers in the literature (Massoud and Bernhardt (2002a), 2002b) McAndrews (2002)

and Hannan et al (2002)) have described or modeled the direct and indirect effects, this paper is

the first to specifically test for the impact of these effects o n bank profitability. We are able to do

this because we have access to a unique data set containing information, among other things, on

bank ATM surcharges, ATM network size, ATM geographic dispersion, monthly total ATM

transactions, the percentage of foreigners using ATMs for each of these banks and other key

bank ATM variables. Specifically, the major contribution of this paper is that it estimates how

strategic variables controlled by bank managers, in this case ATM surcharges and ATM network

size, impact various outcomes that are of importance to bank profitability, through either the

direct or indirect effect described above. Of particular interest is how a bank’s ATM surcharges

  According to Dove (2002) report 50% of large financial institutional recognized that ATM deployment and pricing
could be used to attract customers to other bank products. For examp le, a banker quoted fro m the Dove report p. 110
regarding the advantages underlying extensive ATM networks observed that such networks provided a bank with
“the ability to leverage 18 million transactions per month into cross -sell opportunities for our products and services.”

(and ATM network size) impact its overall profitability – return on equity (ROE) and return on

assets (ROA)– and its direct operational profitability from running its ATM network (ATM

profits). While the impact of surcharges on a bank’s overall profitability (ROA or ROE) may be

higher when surcharges increase, because of either the direct or indirect effects, ATM

profitability will reflect the impact of the direct effect only. In addition, we believe we are the

first to trace whether there is also evidence consistent with the impact of an indirect effect of

ATMs on bank profitability. This is done by examining a two-step process. In the first step, we

examine how bank surcharges (and ATM networks) impact the percentage of ATM users that are

not bank customers (so-called foreign customers). A finding of a high surcharge being associated

with low foreign percent usage would be consistent with (high) surcharges inducing foreign

customers - especially those of small banks with limited networks --to switch their deposit

accounts to larger banks’ charging relatively high surcharges so as to avoid such transaction


         The second step is to analyze how surcharges (and ATM network size) impact the

demand for bank services. A finding of a link between ATM strategic variables and the demand

for bank services would be consistent with an indirect affect – one that appears to reflect a

customer relationship effect. To proxy for bank services we analyze the sensitivity of depositor

growth, total deposits and total loans to ATM surcharges and network size.

         Analyzing the effect of strategic ATM variables on bank profitability and the impact of

the direct and indirect channels is different from much of the prior empirical research in this

literature, which has tended to focus on conditions (e.g., bank size, market concentration etc..)

under which a bank may or may not impose a surcharge and/or whether it is high or low (e.g.,

Hannan et al, (2002), Stavins (2000)), rather than investigating the impact that ATM surcharges

(and network size) have on bank profitability.

        Section 2 of this paper briefly provides an overview of ATM growth and pricing. Section

3 reviews the previous literature. Section 4 presents a model that shows how key ATM strategic

variables (the surcharge and network size) affect: (i) bank profitability, (ii) the degree of foreign

usage of a bank’s ATM networks and (iii) the demand for key bank products. Section 5 discusses

the hypotheses and empirical methodology to be tested. Section 6 discusses the empirical results

and finally Section 7 is a summary and conclusion. An Appendix to the paper describes in detail

the data employed in this study from Dove Consulting (1999) and (2002)) – henceforth Dove.

2. ATMs

        The number of ATMs has grown significantly since being introduced in the late 1960s.

For example, the number of ATMs stood at 324,000 in 2001 versus 83,000 in 1991. There have

been at least 3 phases of growth identified (see Dove (2002)). The first phase was pre-1996, i.e.,

pre-independent surcharging, when there was relatively modest growth in ATMs. The second

phase was 1996 to approximately 1998 when there was rapid ATM growth following the

relaxation of restrictions on individual bank surcharges in 1996. The most recent period (i.e.,

post 1998) has reflected slower growth again. 4

        When a consumer uses ATMs’ of banks other than his or her own (a so-called foreign

consumer) he or she is charged at least two separate fees: (i) a surcharge fee by the bank which

owns the ATM and (ii) a foreign fee by his or her own bank for using ATMs of other banks. 5

  According to Dove (2002) the evolution of ATMs has followed the familiar “S” shape common to many
  Stavins (2002) discusses other ATM related fees such as the interchange fee paid by banks.

         Prior to 1996 banks’ were generally restricted by ATM shared networks from leveling

surcharge fees on foreign customers who used ATMs in the shared network other than those of

their own bank. Since that time, however, the number of banks charging such fees to foreign

users has increased rapidly. By 1998 (only two years later) 78% of US banks were imposing

surcharge fees (Stavins (2000)). In the Dove (2002) Survey more than 90% of the banks

surveyed in 2001 imposed surcharges. Consequently, the surcharge fee has formed a key element

in much of the prior ATM literature – discussed below in Section 3 -- and forms a central

element of this paper.

3. Previous Literature on ATMs

         Following the dramatic increase in the number of banks applying a surcharge to their

foreign customers, there has also been an increase in research on this issue. Massoud and

Bernhardt (2000a and 2000b) and McAndrews (2002) have developed theoretical models which

introduce and analyze the idea of the indirect effect of ATM surcharging on bank profitability.

Other research, e.g., by Hannan et al (2002), Prager (1999), Stavins (2000) Prager (2001) have

examined various empirical elements of ATM pricing.

         Specifically, in some of these papers (e.g., Hannan et al (2002) and Stavins (2000)) the

empirical tests aim to identify factors that determine either the size of a bank’s ATM surcharge

and/or whether a bank sets a surcharge or not 6

  Hannan et al (2002) discuss in some detail the direct effect (“d irect revenue generation”) and the indirect effect
(“strategic motive of attracting customers who wish to avoid paying surcharges”). However, given the nature of their
data they are unable to directly test these hypotheses. They use a logit regression to examine which factors and
market characteristics will impact the choice of whether or not to impose a surcharge. They find, for example, that
the probability of surcharging decreases with ATM share in the market and ATM density while increasing with the
importance of minorit ies in the market population and if the state liberalized early on its regulations on surcharging.
They also find the rate of in-migration to the local banking market has significantly positive effects on surcharges.
This is consistent with an indirect effect being present i.e., surcharges can induce switching by depositors. Stavins
(2000) focuses on the size of ATM networks on surcharges and other fees.

        This paper takes a different approach, in that it seeks to examine how a bank’s strategic

choice of ATM surcharge and its size of ATM network affect bank profitability. In other words,

while much of the literature has attempted to explain surcharging levels, our paper examines

whether or not surcharging impacts key outcomes such as bank profitability. In this way we are

better able to directly test the possible total impact, as well as the direct and indirect effects, of

surcharging on bank profitability and service demand. One reason for our ability to examine this

broader issue is that our data (described in detail in the Appendix) derived fro m market surveys

by Dove ((1999), (2002)) provides both time-series and cross sectional information on bank

surcharges and ATM network size. When combined with Call Report data on bank profitability,

bank services and capital adequacy we are able to develop dynamic insights into how bank

strategic variables impact a bank, and in particular, whether or not an indirect effect is present.

Both the Hannan et al (2002) and Stavins (2000) studies are constrained to analyzing a single

cross-section of bank surcharges and network size. 7

        Moreover, the empirical question posed by Hannan et al (2002) -- the factors determining

whether or not a bank imposes a surcharge -- was clearly of importance in the context of their

1997 database, when only about half the banks in their survey imposed surcharges. However, the

very large increase in the proportion of banks using independent surcharges since 1997 leads us

to ask a different question in this paper – what has been the impact that these surcharges on bank

outcomes and, in particular, on bank profitability?

        The empirical research by Prager (2001) examines the issue of whether consumers from

small banks will switch to larger banks in order to avoid paying a surcharge. This paper does not

use bank level data but rather examines state level data comparing the market share of small

  For example, in Hannan et al (2002), information about surcharges was collected by Moebs Services in 1997 (on
behalf of the Federal Reserve Board) by telephone survey. Stavins (2000) uses data from a survey of financial
institutions conducted by Bank Rate Monitor in May 1997.

banks in states (markets) with and without surcharging over the period 1987 to 1995. She

concludes that small banks actually did a better job of retaining deposit market share in the

presence of surcharging than in its absence. Prager uses this evidence to argue against an indirect

effect. Prager does attempt to examine how surcharging may impact small bank profitability, but

once again the analysis in conducted in terms of a comparison across marke ts (states), with and

without surcharging, rather than at the individual bank level. Finally, Prager’s study covers a

period prior to the liberalization of ATM surcharging and the dramatic growth in ATMs that

occurred after 1996. By comparison our study uses data from the 1996-2001 period.

4. The Theoretical Model

       To establish a framework for an empirical analysis and hypotheses regarding the total,

direct and indirect effects of bank ATM strategic choice variables we utilize the theoretical

framework of Massoud and Bernhardt (2002b).

       In Massoud and Bernhardt (2002b) a spatial game is considered between two banks, A

and B. Each bank is associated with a distinctive spatial line of length Q and each bank chooses

the density of its ATM network on a distinctive line where ATM services can be obtained.

       There is a measure n of bank consumers. Consumers are distinguished by how much they

value one bank intrinsically. The relative valuation for bank A is uniformly distributed over the

range [-m,m]. In addition to providing bank deposits and other products banks provide ATM

services for members and non- members (so called foreign users). First, consumers establish a

bank account at a local branch. Consumers are then hit with a bank-specific location shock that is

uniformly distributed over the range [0,Q]. Each customer receives incremental utility M from

consuming bank services. The transportation cost of acquiring a service is Td where d is the

distance traveled to the closest ATM and T is an incremental transportation cost. Each Bank

chooses its own ATM network size,  j , bank “service” fees, F j and an ATM service fee to

members and non- members,            P j ( ) where   1 for members and   0 for non- members.

Here bank “service” and associated service fees (F j) are broadly defined to include the spread on

investing in assets backed by relatively low cost deposits. Stavins (2002) among others argues

that the fee banks charge their own customers for using their own ATM machines is invariably


Timing of the game follows 3 stages:
Stage 1, banks maximize profits by choosing the density of their ATM locations and the prices

charged for different services (e.g. ATM surcharge to foreign users).

Stage 2, each consumer chooses a bank at which to establish an account.

Stage 3, each consumer receives a bank-specific location shock and chooses where to obtain

his/her ATM service.

The expected profit function of Bank A is
 A  N A F A  N B y A (0)( p A (0)  C ATM )   AC , 8                                         (1)

Where N j is the number of bank j’s customers, C ATM is the marginal cost of providing ATM

services to non- members, C is the cost of installing each ATM machines, y A (0) is the percent
of foreigners as customers, (i.e. bank B customers in this game using bank A’s ATMs) and
    p A (0) is the ATM surcharge fee bank A charges bank B customers. The first term in equation (1)
is the bank’s profit from members use of bank products such as deposits and loans, the second
term is the profit from non- members (i.e. foreigners) who use bank A’s ATM service and the last
term is the cost of installing the ATM network.

  For simp licity, we consider a reduced form of the profit function where the in -branch service fee and the ATM
service fees for members are set to be equal to their marginal cost. This simp lification is introduced after
recognizing that banks in equilibriu m charge their members two part prices (Massoud and Bernhardt 2002a and

          When a bank chooses its optimal ATM surcharge it takes into consideration how that

surcharge would directly impact its profitability -- which depends on foreign customers demand

elasticity. It also takes into account the indirect effect on its profitability. That is, the effect of

ATM pricing on a bank’s profitability can be decomposed into two effects: a direct and a indirect


Bank Surcharge

          The effects of a marginal change in the ATM surcharges on a bank’s profitability is

shown by the following first order condition:

  A   N        N                                    y A (0)                        
       A A F A  A B y A (0)( p A (0)  C ATM )  N B  A ( p A (0)  C ATM )  y A (0) 
                                                        p (0)                          
p (0) p (0)
                 p (0)                                                                 

Where      N A       N         y A (0)
                  0, A B  0 and A       0
          p (0)
                     p (0)      p (0)

The first two terms, in equation (2), show the indirect effect of ATM surcharges on bank A’s
profitability, where N A F A is the increase in bank-account membership and bank service
                       p (0)

purchases induced by a marginally higher surcharge times the bank service “fee” or profit
extracted, and N B y A (0)( p A (0)  C ATM ) is the loss in surcharge revenues from those foreign
                 p (0)

customers (i.e. bank B customers in this model) who switch bank- membership to bank A, due to
the increase in ATM surcharges.
          The last term in equation (2),        y A (0)                        
                                           N B  A ( p A (0)  C ATM )  y A (0)  ,
                                                p (0)                          
                                                                                       shows the direct effect of
                                                                                

ATM surcharges on bank A’s profitability which is the impact of increasing ATM surcharges on
surcharge profits from (foreign) customers who continue to establish bank accounts at bank B.
The last term in equation (2) or the direct effect can be rewritten in terms of non- members ATM
demand elasticity,  :

                     ( p A (0)  C ATM )      the sign of this term depends on                         ( p A (0)  C ATM )  .   If this term
    y A (0) N B   
                                           
                                          1                                                   sign  
                                                                                                                               
                                                                                                                              1
                            p A (0)                                                                           p A (0)        

is non-negative,                  ( p A (0)  C ATM )                                                   p A (0)                p A (0)
                               
                                                      1  0 ,
                                                                   then it implies that                               and                  1.
                                         p A (0)                                                     p (0)  C ATM
                                                                                                                              p (0)  C ATM

Given that in general a monopoly operates in a price region such that the elasticity exceeds one 9 ,
then this inequality                                 p A (0)      should hold which implies that ATM profits are
                                    1  
                                                p A (0)  C ATM

positively related to ATM surcharges. Note also that,                                     N A
                                                                                                   0 , shows that an increase in the
                                                                                         p A (0)

ATM surcharge increases a bank’s customer base because of switching, that                                               N B           shows that
                                                                                                                       p A ( 0 )

an increase in the ATM surcharge reduces a rival bank’s customer base and                                              y A (0)        shows
                                                                                                                       p A (0)

that an increase in the ATM surcharge reduces a bank’s market share of non-member (foreign)

Bank ATM network
              The effect of size of a bank’s ATM network on a bank’s profitability is shown by the
following first order condition:

 A N A A N B A                            y A (0)
         F       y (0)( p A (0)  C ATM )           N B ( p A (0)  C ATM )  C                                             (3)
 A  A      A                              A

Where          N A     N      y A (0)                          reflects the relative size of bank A’s ATM network. Here
                     0, B  0,           0 and  A
                A      A      A

again we can decompose the effect of the size of a bank’s ATM network on bank profitability
into direct and indirect effects. The direct effect is captured by the third and the fourth terms of
                               y A (0)
equation (3), i.e.                      N B ( p A (0)  C ATM )  C ,   which shows how a marginal increase in a bank’s
                                 A

ATM network would increase bank revenue from non- member (foreign) users of ATMs. The
direct effect would be positive if the marginal increase in ATM revenue is higher than the cost of
adding one more ATM to the network, that is                                y A (0)
                                                                                    N B ( p A (0)  C ATM )  C .
                                                                             A

  Tirole (1988), page 66, where elasticity of demand is lo wer than 1, the monopolist’s revenue -- and his profits --
are decreasing in quantities.

       The indirect effect is captured by the first two terms in equation (3), where N A F A shows
                                                                                              A

how a larger ATM network would increase a bank’s customer base (bank produc t purchases)
because of switching and N B y A (0)( p A (0)  C ATM ) shows the marginal loss in ATM surcharge
                              A

revenues because of foreign customer switching to become bank customers.

Note that N A  0 shows an increase in a bank’s ATM network increases its ability to expand its
            A

customer base, also N B  0 shows that an increase in a bank’s ATM network would decrease
                       A

the customers bank account base of a rival bank (here bank B). Finally,       y A (0)      shows that a
                                                                                A

larger ATM network would increase the market share of non- member (foreign) users of ATM’s.
       One of the main implications of the Massoud and Bernhardt (2002b) model is that banks
tend to over-provide ATM’s because they extract profits more efficiently from bank members
through bank service fees and other income (such as the spread on deposits) rather than from
ATM use by (foreign) members of other banks, and that a more developed ATM network raises
the attraction of establishing account membership with a bank.

5. Hypotheses
       From the theoretical model in Section 4 in which a bank employs its surcharge level and
ATM network as strategic variables to increase profitability either through the direct or indirect
channels we can derive a number of testable hypotheses. Specifically, we test the following
hypotheses with respect to surcharges, ATM network size and bank profitability:
Total Effect
       Hypothesis 1: If foreign users are relatively price inelastic (direct effect) and/or switching

is sufficiently strong to overcome any loss in revenue if foreign users are price elastic (indirect

effect), then overall profitability will be increasing in surcharge levels:

                            ()   ( )
       H1: ROE j  f ( p j (0),  j )

                            ()   ( )
             ROAj  g ( p j (0),  j )

                Note that we can also test an implication of the Massoud Bernhardt (2000b)

       Model – see Section 4 – that banks tend to over provide ATMs such that bank

       profitability (return) is decreasing in the size of a bank’s ATM network. An alternative

       hypothesis – believed by many bankers (see Dove (1999)) -- is that greater network size

       adds to the customer attraction to a bank and as a result the relationship between overall

       profitability (return) and network size is positive. Here we specify, a priori, a negative

       relationship between overall bank profitability and ne twork size in hypothesis 1, but

       recognize that the sign may be non-negative.

Indirect Effect

       Hypothesis 2: While a finding that overall profits are positively linked to surcharges is

consistent with the presence of both direct and indirect effects we wis h to investigate these two

alternative channels more deeply. In particular, a two-step process is required for the indirect

channel. Consequently, we propose to test H2 and H3:

                                   ( )   ( )
       H2:          y j (0)  h( p j (0),  j )

       H2 tests the first step in the indirect channel, namely that higher surcharges ( p j (0) )

result in a lower proportion of foreign ATM users ( y j (0) ) since foreign users have an incentive

to switch to becoming bank customers to avoid surcharges. Thus, the relationship between

foreign usage proportion and bank surcharge levels is expected to be negative. A conventional

view is that greater network scope (  j ) will add to foreign customers use of the ATM network –

a positive network size effect. That is, the relationship between y j (0) and  j will be positive.

However, if Massoud and Berhardt (2002a) are correct and banks over provide ATMs, then this

relationship may be negative. Note a negative relationship between y j (0) and  j would also be

consistent with a switching effect. That is, the reduction in the proportion of foreigners using a

(larger) bank’s ATMs could be the result of those customers switching to banks with larger ATM

networks that offer consumers greater convenience. Thus, we hypothesize, a priori that a

negative relationship exists between foreign usage proportion and network size, but recognize

that the relationship maybe positive.

Hypothesis 3:

                                                 ()   ()
       H3: Depositor growthj  k ( p j (0),  j )
                                     ( )     ( )
                Deposits j  h( p (0),  j )

                               ()      ()
                Loansj  i( p j (0),  j )

       H3 reflects the second step in the indirect channel, namely that ATM users who switch to

become bank account holders consume more bank products. Here we proxy fo r increased

consumption of bank products by the dollar size of a bank’s deposits and loans. Such

consumption would potentially add to a bank’s overall profitability. We also look at the link

between surcharges and network size and a bank’s depositor growth. If the indirect channel is

operational we would expect both surcharge level and ATM network size to have positive effects

on depositor growth over the succeeding period, which is a precondition for them to expand their

purchases of a bank’s products.

Direct Effect

Hypothesis 4:
                                           ()   ( )
                H4: ATM profitsj  m( p j (0),  j )

Hypothesis H4 seeks to test the direct effect. Specifically, if foreign customer demands were

sufficiently price inelastic then raising surcharge levels would add to a bank’s profits from ATM

provision. That is, there should be a positive relationship between ATM profitability and p j (0)

the surcharge level. Similarly if (marginal) additions to a bank’s ATM network adds to its

revenues and they exceed any additional costs, then greater network size should also increase

bank profits. That is, ATM profitability should be positively associated with network size.

However, to the extent that bank’s over provide ATM’s (as in Massoud and Bernhardt, 2002a

and 2002b) then the relationship between ATM profitability (return) and ATM network size may

be negative. Thus, we hypothesize, a priori a negative sign.

       The next section discusses the empirical methodology used to test these hypotheses and

the empirical results.

6. Empirical Methodology and Results

       The Appendix to this paper provides a detailed description of data used in this study. As

discussed there the empirical tests are based on underlying survey data generated by Dove

Consulting of Boston in two reports on ATM deployment – the first in 1999 and the second in

2002. These data provide bank specific details by bank and year regarding surcharge levels,

ATM network size, transactions per ATM, percent use of a bank’s ATM network by foreigners

as well as other pertinent ATM related data. As a result our basic sample size covers 1996 –

2001 (six years). As discussed in the Appendix not every variable was measured each year and

the sample of banks differed over the 1999 and 2002 surveys. Nevertheless, we feel that this data

set is sufficiently unique to allow us to examine the impact of ATM strategic variables (ATM

surcharge and ATM network size) on bank profitability as well as to gain insights into the

indirect and direct channels through which ATMs affect bank profitability.

         To gain insights into the indirect channel and to control for bank size, bank risk and

market share we employ bank Call Report data and Federal Reserve generated bank market share

data. With respect to the dependent variable data on bank depositor growth (percentage change

in the number of depositors) and the dollar value of deposits and loan these were derived from

Call Reports that most closely matched the dates of the Dove surveys. This was also true for the

independent or control variables used: the bank’s capital – assets ratio (a measure of bank risk),

bank assets (a measure of a bank’s size) and bank market share (a measure of local market

competitive conditions). 10

         An additional independent variable that we include is the number of transactions per

ATM. This variable, taken from the Dove data set, includes both foreign as well as customer

transactions at ATMs. This variable captures the “S” shaped growth curve in the number of

ATMs from 1996 to 2002 described in detail by Dove (2002). By including this variable as an

independent variable, we are able to capture the possible impact of the under/oversupply of

ATMs on the dependent variables of interest.

         We also included in the tests a measure of geographic dispersion of a bank’s ATM

network. Dove consulting divides the U.S. up into 7 regions and identifies whether a bank has

ATMs in each region and outside U.S. (internationally). -- making 8 possible region in all. Since

a bank has an endogenous choice to make regarding the number of regions it wishes to span with

  The market share variable used in these tests was the percent of the banks deposits relative to total deposits in the
State in wh ich the bank’s HQ is located.

its ATM network, it can use greater geographic or dispersion scope to attract new customers to

the bank. 11 That is, geographic dispersion could be considered as an additional ATM strategic

variable along with the level of bank surcharge and the size of a bank’s ATM network. The

geographic dispersion variable takes a value between one and eight, where the value of the

variable reflects the sum of the regions over which a bank locates its ATMs. The Appendix

discusses in more detail the different regions identified in the Dove surveys.

        In testing hypotheses H1 to H4 we employ both fixed and random effect tests of the

impact of the ATM strategic choice variables (ATM surcharge and number of ATMs) on the

various outcomes of interest (profits, demand for other bank products etc...). We report both the

fixed effect and random effect results, as well as reporting the Hausman test statistic to determine

if the random effects model is appropriate for a given model.

        In all of our tests we report both contemporaneous results (i.e. where the dependent

variable is measured in the same year as the independent variables) - which are labeled

“Current” - and results when the dependent variable is one year ahead of the independent

variables – which are labeled “Lead”. There are two reasons why we believe that the lead results

offer a better test of the link between ATM strategic variables and bank profitability. First, the

lead results capture causality in the relationship by focusing on how a change in the independent

bank strategy variables (e.g. ATM surcharge) impacts the dependent bank outcome variables

(e.g. return) one year later. Second, the lead equations also capture any likely institutional

frictions, in that it may take a period of time for consumers to p rocess the fact that a bank has

changed its ATM, pricing strategy and a further period of time to decide to switch to a new bank.

  For example, by expanding a network out of reg ion a bank provides convenience for those customers who travel
extensively and may persuade more foreigners to switch their accounts to the bank.

However, it is also possible that this switching decision may be made relatively quickly, which is

why we also report the “current” equations.

Results of the Effects on Total Profitability (H1)

        Tables 1a, 1b, 2a and 2b show, respectively, the effects of surcharge change and ATM

network size on bank ROA and ROE. ATM network “size” is measured by the number of ATMs

owned by the bank. As can be seen a striking result in all four tables is the positive and

statistically significant impact of bank surcharge levels on bank profitability. As noted earlier,

this positive total effect may be due to either a direct effect or an indirect effect -- reflecting

customer switching. The impact of the number of ATMs is generally insignificant. This suggests

little systematic linkage between the size of a bank’s ATM network and its ATM overall

profitability. With respect to the other (control) variables no particular pattern is striking. Thus

H1 is supported for bank surcharge levels, but rejects the hypothesis with respect to a negative

linkage between ATM network size and overall bank profitability.

Results for the Indirect Channel (H2 and H3)

        Tables 3a and 3b show the sensitivity of foreign (non-bank) ATM users to the level of the

bank’s ATM surcharge. As can be seen in both Tables 3a and 3b, for both random and fixed

effect models, surcharge is significantly negative. That is a higher surchar ge results in a lower

foreign percentage usage, which is consistent with switching and supports H2 (it is also the first

step in the indirect channel). With respect to ATM network size, the number of ATM’s is

insignificant in both Tables 3a and 3b. Again, there is little overall pattern in the other control

variables. Overall, there is some evidence that is consistent with the first step in the indirect

channel and thus with switching.

       Tables 4, 5a, 5b, 6a, and 6b seek to test the second step in the indirect channel

(hypothesis H3), i.e., once a customer has switched do we see an expansion in customer demand

for bank products?

       Table 4, analyses the percentage growth in bank depositors on a year-by- year basis. As

can be seen for both the fixed and random effect models a high bank ATM surcharge is

associated with a higher depositor growth rate in the succeeding year. Similarly the number of

ATMs variable is significantly positive for both the fixed and random effects models, again this

is consistent with the switching story and the indirect channel. The geographic dispersion

variable is positive and significant in both the fixed and random affect models in Table 4. That

is, banks with more extensive geographic networks engender greater depositor growth over the

succeeding year. Because geographic dispersion of ATMs can be thought of as a bank strategic

variable this is also consistent with the indirect channel and switching.

       Table’s 5a, 5b, 6a and 6b analyze the effects of bank surcharge level and ATM network

size on specific bank products. For the dollar value of deposits and loans (5a and 6a), where a

one year lead variable is used as the dependent variable, the bank ATM surcharge level is

positive and statistically significantly related to deposits and loans. Furthermore, it also appears

that increasing the number of ATMs (the second strategic variable) has a generally positive and

statistically important effect on (next period’s) deposit and loan amounts. Geographic dispersion

also has a significant and positive impact on the (next period) amount of deposits (6a). These

results are consistent with the second step in the indirect channel, suggesting a potential

customer relationship or switching effect linking ATM strategic variables and bank profitability.

        As described above, we have a preference for the lead model tests (5a and 6a) over the

current model tests (5b and 6b) due to of both econometric issues (the clarity of the causality

relationship between strategy changes this year and outcome changes next year) as well as issues

relating to institutional frictions (the lag in consumers becoming aware of bank strategic changes

and the additional lag in acting on these changes to switch banks). It might be noted that the

results in Tables 5b and 6b for the current models are less strong than the lead results (5a and

6a). Specificantly, general while surcharge has a positive sign it is statistically insignificant in the

current regressions.

The Direct Effect (H4)

        To test the direct effect of the strategic variables on bank profitability (hypothesis H4) we

utilized the Dove consulting data to calculate the profitability of operations of “within branch”

ATMs. This test was unable to include out-of-branch ATMs (and their profitability) since Dove

did not collect these data. Moreover, only one year of data is available on ATM profitability and

even then for just 33 banks. The results of the regression of bank ATM profits on our strategic

and control variable is shown in Table 7. While both surcharge and the size of ATM network

variables have positive signs, however, these variables are statistically insignificant.

Nevertheless, while the evidence in Table 7 tends to reject the direct effect (and H4), the results

may well reflect the limited sample size with which the tests were undertaken.

7. Summary and Conclusion

        This paper has modeled and tested the relationship between a bank’s use of strategic

ATM variables – the foreign-user surcharge and its own network size – and a bank’s overall

profitability. An important aspect of the study was to differentiate the direct and indirect (or

customer relationship) channels linking surcharge levels and network size to bank profitability.

Using a unique data set provided by Dove Consulting, which contained both time-series as well

as cross-sectional information on bank surcharges and other ATM related variables, it was found

that there is evidence consistent with customer switching and thus the indirect channel.

Specifically, higher surcharge levels were associated with greate r depositor growth, greater

deposit amounts and greater loan amounts. In addition foreign ATM users seemed to be adverse

to high surcharge levels, such that high surcharge levels may have induced foreign users to

switch to becoming account holders of the high foreign surcharge bank. The evidence, in

general, is consistent with presence of an indirect channel and switching. With respect to the

direct channel, the data on ATM profitability that was available was very limited and what was

available failed to support the hypothesis of a significant linkage between bank strategic ATM

variables and ATM profitability. Clearly, this is one area that needs further research as more

extensive databases on ATM profitability become available.

Table 1a: The Relationship Between ATM Strategic Variables and ROA (Lead), 1996-2001
                                               Standard                            Standard
                           Random Effects      error           Fixed Effects       error
Surcharge                   0.173786***        0.0454716        0.1518829**        0.061278

Number of ATMs              7.62E-06           0.0000372       -0.0000322          0.0000912

Geographic Dispersion       0.037714           0.0287631        0.1599676**        0.0780467

Transactions per ATM        5.87E-06           4.34E-06        -4.01E-06           0.0000141

Bank Capital Ratio          0.006841           0.0056544       -0.0008283          0.0185794

Bank Assets                 3.20E-13           6.78E-13         3.58E-13           1.54E-12

Bank Market Share          -0.00017            0.0010936       -0.0015881          0.0012486

Constant                    0.335854***        0.0973816        0.3172375          0.2256431

Sample Size                276                                 276
R 2 Within                 0.09                                0.11
R 2 Between                0.12                                0.04
R 2 Overall                0.09                                0.03

Hausman Test  2           9.54 (0.2161) Random effects is appropriate estimator
*** Indicates p value of 1%
**    Indicates p value of 5%
*     Indicates p value of 10%
(.)   Indicates p value

Source for all Tables: Dove (1999, 2002), Call Report and Source data.

Table 1b: The Relationship Between ATM Strategic Variables and ROA (Current), 1996 -
                                               Standard                             Standard
                            Random Effects     error           Fixed Effects        error
Surcharge                    0.101384**        0.040927         0.096822*           0.051368

Number of ATMs              -6.17E-06          3.52E-05        -2.1E-05             7.72E-05

Geographic Dispersion       -0.006             0.027335        -0.0752              0.065878

Transactions per ATM         5.86E-06          4.18E-06         9.06E-06            1.19E-05

Bank Capital Ratio           0.00092           0.005278        -0.04439***          0.012913

Bank Assets                  3.98E-13          6.33E-13         1.59E-12            1.30E-12

Bank Market Share            0.002601***       0.000955         0.003011***         0.00105

Constant                     0.518622***       0.089926         1.034202***         0.168692

Sample Size                 282                                282
R 2 Within                  0.056                              0.12
R 2 Between                 0.067                              0.00
R 2 Overall                 0.102                              0.00

Hausman Test (  2 )        22.65(0.00) Random effects is inappropriate estimator
*** Indicates p value of 1%
**     Indicates p value of 5%
*      Indicates p value of 10%
(.)    Indicates p value

Table 2a: The Relationship Between ATM Strategic Variables and ROE (Lead)
                                               Standard                            Standard
                           Random Effects      error           Fixed Effects       error
Surcharge                   2.0642340***       0.6466082       2.463823**          0.972484

Number of ATMs             -0.0002266          0.0004920       -0.0013718          0.001447

Geographic Dispersion       0.1826976          0.3781347       1.643325            1.238605

Transactions per ATM        0.0000671          0.0000555       -0.0000657          0.000224

Bank Capital Ratio         -0.2640739***       0.0728048       -0.6417043**        0.294856

Bank Assets                 8.32e-12           9.31E-12        2.25E-11            2.44E-11

Bank Market Share           0.0321767*         0.0169702       0.0141022           0.019815

Constant                   7.5169910***        1.3155360       9.389141***         3.580964

Sample Size                276                                 276
R 2 Within                 0.07                                0.10
R 2 Between                0.18                                0.13
R 2 Overall                0.17                                0.12

Hausman Test  2           9.57 (0.2141) Random effects is appropriate estimator
*** Indicates p value of 1%
**    Indicates p value of 5%
*     Indicates p value of 10%
(.)   Indicates p value

Table 2b: The Relationship Between ATM Strategic Variables and ROE (Current), 1996 -
                                                Standard                            Standard
                            Random Effects      Error           Fixed Effects       Error
Surcharge                    1.487136**         0.610112         1.884333**         0.772716

Number of ATMs              -0.00014            0.000498        -0.00062            0.001162

Geographic Dispersion        0.087227           0.38436         -0.85274            0.991

Transactions per ATM         9.09E-05           5.78E-05         0.000166           0.000179

Bank Capital Ratio          -0.41068***         0.073679        -1.64494***         0.194244

Bank Assets                  2.44E-12           9.11E-12         2.46E-11           1.96E-11

Bank Market Share            0.022404           0.014906         0.029696*          0.015797

Constant                    9.563363***         1.29026         21.63025***         2.537614

Sample Size                 282                                 282
R 2 Within                  0.21                                0.31
R 2 Between                 0.16                                0.13
R 2 Overall                 0.16                                0.13

Hausman Test  2            47.73(0.00) Random effects is inappropriate estimator
*** Indicates p value of 1%
**     Indicates p value of 5%
*      Indicates p value of 10%
(.)    Indicates p value

Table 3a: The Relationship Between ATM Strategic Variables and Foreign Percent ATM
Usage (Lead), 1996 – 2001
                                               Standard                               Standard
                           Random Effects      error           Fixed Effects          error
Surcharge                   -6.06197**          2.893624        -8.36561*              5.04634

Number of ATMs              -0.00413            0.003508        -0.01088               0.011228

Geographic Dispersion       -0.16339            1.864571        -0.45271               5.560722

Transactions per ATM        -0.00066***         0.000267        -0.00092               0.001302

Bank Capital Ratio         26.6787             33.39713        225.3731**             111.3641

Bank Assets                 4.88E-11            9.48E-11         1.33E-10               1.84E-10

Bank Market Share           0.010403            0.202422         1.635177               1.039176

Constant                   44.24957***          6.190163       22.43404                16.61529

Sample Size                120                                 120

R 2 Within                 0.2667                              0.4117

R 2 Between                0.0887                              0.0114

R 2 Overall                0.1082                              0.0184

Hausman Test  2           14.46 (0.0435) Random effects is inappropriate estimator
*** Indicates p value of 1%
**    Indicates p value of 5%
*     Indicates p value of 10%
(.)   Indicates p value

Table 3b: The Relationship Between ATM Strategic Variables and Foreign Percent ATM
Usage (Current) 1996 – 2001
                                              Standard                            Standard
                           Random Effects     error           Fixed Effects       error
Surcharge                  -6.28685***        2.133036        -6.92507**           2.935974

Number of ATMs             -0.00096           0.00179         -0.00177             0.00523

Geographic Dispersion      -0.07767           1.328582        -7.93143**           4.042646

Transactions per ATM       -0.00059***        0.00024          0.000442            0.000761

Bank Capital Ratio          0.06763           0.256811        -0.62671             1.150436

Bank Assets                -4.40E-12          3.13E-11         8.05E-11            8.70E-11

Bank Market Share          -0.17791           0.169284        -0.02708             0.486601

Constant                   47.73202***        4.691936        62.791.58***        13.50732

Sample Size                182                                182
R 2 Within                 0.1112                             0.1692

R 2 Between                0.1505                             0.0002

R 2 Overall                0.1336                             0.0026

Hausman Test  2           5.98(0.5424) Random effects is appropriate estimator
*** Indicates p value of 1%
**    Indicates p value of 5%
*     Indicates p value of 10%
(.)   Indicates p value

Table 4: The Relationship Between ATM Strategic Variables a time t and the Percentage
Change in the Number of Bank Depositors (t to t + 1), 1996 - 2001
                                               Standard                              Standard
                           Random Effects      error           Fixed Effects         error
Surcharge                  47.81289**          24.43226         50.3209*             26.13786
Number of ATMs              0.053089**          0.027212         0.099056**           0.038599
Geographic Dispersion      47.37285**          21.77648        111.5937***           33.10758
Transactions per ATM        0.000697            0.003593        -0.00114              0.006007
Bank Capital Ratio         -4.29506             4.656989        -4.04919              8.050552
Bank Assets                -1.78E-09***         4.67E-10        -3.07E-09***          6.53E-10
Bank Market Share          -0.87636*            0.499643         -1.33361**           0.528409
Constant                   -35.5029            70.2705         -166.273*             96.73478

Sample Size                268                                 268

R 2 Within                 0.19                                0.21
R 2 Between                0.00                                0.00
R 2 Overall                0.01                                0.00

Hausman Test  2            17.65(0.014) Random effects is inappropriate estimator
*** Indicates p value of 1%
**    Indicates p value of 5%
*     Indicates p value of 10%
(.)   Indicates p value

Table 5a: The Relationship Between ATM Strategic Variables and the Dollar Value of Bank
Loans (Lead), 1996 – 2001
                                              Standard                               Standard
                           Random Effects     error           Fixed Effects          error
Surcharge                  2.36E+09*          1.36E+09        6.73E+09**             2.80E+09
Number of ATMs             3113974***         950782.7        2.35E+07***            4162531
Geographic Dispersion      6.72E+08           6.80E+08        5.05E+09               3.57E+09
Transactions per ATM       40439.83           97145.84        196391.1               642715.6
Bank Capital Ratio         -5.94E+07          1.29E+08        -2.06E+09**            8.46E+08
Bank Assets                0.5669262***       0.022156        -0.0266869             0.070348
Bank Market Share          3.18E+07           4.90E+07        -2.08E+07              5.70E+07
Constant                   -2.68E+09          2.57E+09        2.11E+09               1.03E+10

Sample Size                278                                278
R 2 Within                 0.66                               0.75
R 2 Between                0.99                               0.47
R 2 Overall                0.90                               0.59

Hausman Test  2           283.024(0.00) Random effects is inappropriate estimator
*** Indicates p value of 1%
**    Indicates p value of 5%
*     Indicates p value of 10%
(.)   Indicates p value

Table 5b: The Relationship Between ATM Strategic Variables and the Dollar Value of
Bank Loans (Current), 1996 – 2001
                                              Standard                              Standard
                           Random Effects     error           Fixed Effects         error
Surcharge                  3.86E+08           3.04E+08        3.30E+08              3.28E+08

Number of ATMs             -68695.5           302670.3        -638005.8             493159.9

Geographic Dispersion      -9.24E+08***       2.38E+08        -2.86E+09***          4.21E+08

Transactions per ATM       -538.441           37747.33        81332.05              76027.49

Bank Capital Ratio         4.64E+07           4.65E+07        -3.94E+07             8.24E+07

Bank Assets                0.611503***        0.005274        0.636024***           0.00832

Bank Market Share          5911443            6561399         1.87E+07***           6705126

Constant                   4.36E+08           7.50E+08        4.04E+09***           1.08E+09

Sample Size                282                                278
R 2 Within                 0.99                               0.74
R 2 Between                0.99                               0.46
R 2 Overall                0.99                               0.58

Hausman Test  2            54.45(0.00) Random effects is inappropriate estimator
*** Indicates p value of 1%
**    Indicates p value of 5%
*     Indicates p value of 10%
(.)   Indicates p value

Table 6a: The Relationship Between ATM Strategic Variables and the Dollar Value of Bank
Deposits (Lead), 1996 – 2001
                                              Standard                           Standard
                           Random Effects     error           Fixed Effects      error
Surcharge                   1.62E+09          1.64E+09        4.98E+09*          2.64E+09
Number of ATMs              1.10E+07***       1218081         1.09E+07***        3933569
Geographic Dispersion      -2.38E+09**        9.33E+08        7.42E+09**         3.37E+09
Transactions per ATM       -3002.922          135721.9        91173.39           607362.8
Bank Capital Ratio         -2.37E+08          1.78E+08        -1.65E+09**        8.00E+08
Bank Assets                 0.3284854***      0.023478        0.2584174***       0.066479
Bank Market Share           2.75E+07          4.47E+07        -3.91E+07          5.38E+07
Constant                    3.69E+09          3.28E+09        -2.18E+09          9.71E+09

Sample Size                278                                278
R 2 Within                 0.79                               0.81
R 2 Between                0.89                               0.72
R 2 Overall                0.89                               0.73

Hausman Test  2            21.21(0.0035) Random effects is inappropriate estimator
*** Indicates p value of 1%
**    Indicates p value of 5%
*     Indicates p value of 10%
(.)   Indicates p value

Table 6b: The Relationship Between ATM Strategic Variables and the Dollar Value of
Deposits (Current), 1996 – 2001
                                               Standard                         Standard
                           Random Effects      error           Fixed Effects    error
Surcharge                  -4.02E+08           4.74E+08        -2.36E+08        3.58E+08

Number of ATMs             1438626**           633301          -1646841***      538092.9

Geographic Dispersion      -2.07E+09***        5.22E+08        -7.66E+08*       4.59E+08

Transactions per ATM       -14072.13           89891.54        15841.49         82954.53

Bank Capital Ratio          7.39E+07           1.03E+08         1.79E+07        9.00E+07

Bank Assets                 0.532291***        0.010729         0.580527***     0.009078

Bank Market Share           1.73E+07*          9669386          1.28E+07*       7316046

Constant                    1.79E+09           1.70E+09         1.52E+09        1.18E+09

Sample Size                282                                 282
R 2 Within                 0.99                                0.99
R 2 Between                0.82                                0.78
R 2 Overall                0.90                                0.88

Hausman Test  2           0.00 (1.0) Random effects is appropriate estimator
*** Indicates p value of 1%
**    Indicates p value of 5%
*     Indicates p value of 10%
(.)   Indicates p value

Table 7: The Relationship Between ATM Strategic Variables and In-Branch ATM
Profitability, 2001
                                        Coefficient                    Standard error
Surcharge                                1370.984                             983.3386

Number of ATMs                            -0.39663                            0.627244

Geographic Dispersion                     -105.544                             369.777

Transactions per ATM                      -0.25734                            0.270495

Bank Capital Ratio                        -3.90388                            125.0021

Bank Assets                              1.01E-08                             1.29E-08

Bank Market Share                        36.58582                             43.71877

Constant                                  -816.889                            2349.393

Sample Size                33

R2                         0.22

F-test                     0.99

*** Indicates p value of 1%
**    Indicates p value of 5%
*     Indicates p value of 10%
(.)   Indicates p value



       An important aspect of this paper lies in the uniqueness of the ATM data set employed in

our tests. This data set was purchased from Dove Consulting Inc. and includes bank level data on

a variety of variables that have not been previously used in the empirical ATM literature. In

particular, the data includes a measure of the percentage of ATM users for different banks, who

are foreigners – i.e., those who pay ATM surcharges. This variable – in conjunction with other

data such as surcharge amount and ATM network size, allows us to test the hypotheses discussed

in the paper. The Dove Survey data is used in conjunction with a variety of other publically

available sources of bank level data, including Call Reports (Report of Condition and Income)

taken from the Federal Reserve web site and Market Share data taken from SOURCE.

       The database provided by Dove Consulting is taken from two separate surveys of ATM

providers – one taken in 1998 and the second in 2001(The Dove reports themselves were

published in 1999 and 2002 respectively). In each case data were collected from each bank in the

sample for the preceding three years generating a 6-year sample that spans 1996-2001.

Unfortunately, the two surveys are not identical across the two time periods thus some data are

available for only some of the time periods. For example, while each of the two Dove surveys

asked respondents for information on a variety of variables for each of the proceeding three

years, this was not the case for the foreign percentage variable. The first survey conducted during

1998 did ask for this data for each of the preceding 3 years, however the second survey only

asked the respondents for this data for the final year of that survey i.e., 2001. In other words, for

some of our empirical tests, which require the use of the foreign percentage variable, we use a

data set made up of certain banks for each of 1996, 1997 and 1998 and different banks in 2001.

       A further issue with our data concerns how the banks were asked to report their ATM

surcharges over the preceding three years. In the case of both the 1998 and 2001 surveys, banks

were asked to provide data on their ATM surcharges at the time of the survey. They were also

asked to provide the date of the last change of the surcharge and how much that change was (in

dollars and cents). This information is enough to create a partial historical record of surcharges

charged by each bank. For example, if the date of the pervious surcharge change occurred prior

to the three-year period covered by the survey, then we are able to use the value of the surcharge

in the final year of the survey for all of the previous three years. Similarly, if the most recent

surcharge change occurred during the preceding 3 years we are able to infer surcharges after that

date. However, we would not be able to infer surcharges outside the three- year window of each

data set. In cases where we are not able to infer the surcharge amount from the data, we do not

use the data.

       One important variable employed is the measure of geographic dispersion of a bank’s

ATM network. Dove divides the U.S. into 7 regions and identifies whether each bank (in each

year) has an ATM in one of those regions plus whether ATMs are held internationally (making 8

regions). The 7 U.S. regions identified were New England, Mid-Atlantic, Southeast, South,

Midwest, Mountain and Pacific (see Dove, 1999, p. 27). The geographic dispersion variable

takes a value between one and eight, where the value of the variable reflects the sum of the

regions where a bank locates its ATMs. For example, if geographic dispersion is equal to one, it

indicates an ATM presence in only one region and if it is higher than one it indicates presence in

more than one region.


1. Dove Consulting Group, 1999 ATM Deployer Study, Boston, 1999.

2. Dove Consulting Group, 2002 ATM Deployer Study, Boston, 2002.

3. Hannan, T.H., E.K. Kiser, R. Prager and J.J. McAndrews, “To Surcharge or Not to

   Surcharge: An Empirical Investigation of ATM Pricing,” mimeo 2002, Review of

   Economics and Statistics (forthcoming).

4. Massoud, N. and D. Bernhardt, “Rip-off ATM Surcharges” Rand Journal of Economics,

   Spring 2002, pp. 96-115.

5. Massoud, N. and D. Bernhardt,” Endogenous ATM Location and Pricing,” Working

   Paper, Dept. of Finance, School of Business, University of Alberta.

6. McAndrews, J.J., “A Model of ATM Pricing: Foreign Fees and Surcharges,” mimeo,

   Federal Reserve Bank of New York, 2002.

7. Prager, R., “ATM Network Mergers and the Creation of Market Power,” The Antitrust

   Bulletin, summer, 1999 pp. 349-363.

8. Prager, R., “The Effects of ATM Surcharges on Small Banking Organizations,” Review

   of Industrial Organization, 2001, pp. 161-173

9. Sienkiewicz, S., “The Evolution of EFT Networks from ATMs to New On- Line Debit

   Payment Products,” Discussion Paper: Payment Cards Center, Federal Reserve Bank of

   Philadelphia, April 2002.

10. Stavins, J., “ATM Fees: Does Bank Size Matter?” New England Economic Review,

   January/February 2000, pp. 13-24.


To top