Switching Costs and Network Externalities in the Production of Payment Services

¡ ¢ £ ¤ ¥ ¦ ¤ § ¨ ¨ ¨
© ! ! ! # $ ! % %

' ( ) 0 1 2 3 1 4 5 6 3 7 0 3 0
' ' ) 6 7 1 5 3 2 0 2 8

SWITCHING COSTS AND NETWORK
EXTERNALITIES IN THE PRODUCTION OF
PAYMENT SERVICES1
Eero Lehto
Labour Institute for Economic Research
18.5.2000
Address: Pitkänsillanranta 3 A, 00530 Helsinki, Finland
Tel. +358-09-2535 7350
e-mail: eero.lehto@labour.fi
Abstract
This study considers the retail banking duopoly in the production of payment services. Switching
costs make the old clients locked-in. Network externalities are exhibited due to high transaction
costs in making a transfer to another bank's customer, and due to low marginal costs in processing
an intrabank transfer. This makes the discounted marginal profits from the old clientele increase in
the number of old clients that favours a larger bank. This can encourage the larger bank to capture
the new clients which are free of switching costs, even if price discrimination between old and young
clients is not allowed. To obtain this result, requires, however, that a maximal price premium
extracted from a payment transfer to another bank’s customer is small enough. Thus the ordinary
result according to which, it is too big sacrifice for a larger bank to lower also the prices of the old
clients in order to capture the new clients, can be reversed.
JEL-code: G21, L11, L12
Key words: retail banking, switching costs, network externality.
1
This paper was presented in ESEM 1999 (Santiago de Compostela) and in EARIE 1999 (in Torino).

2
1. Introduction
The transfer of payments requires debiting and crediting clients’ accounts. To be able to handle the
payments, the modern bank needs software and also physical capital in the form of buildings,
computers and so on. The more electronified the process is, the lower the marginal costs are in
producing a payment service. To be able to serve a client a bank has to have an access to an client's
bank account. The electronic links between the customers together with the procedures which route
payments to right address and do appropriate debiting, crediting and respective accounting, form a
payment network. The interconnection of the networks guarantees the access from each customer's
account to the payment account of the rival’s clients also. When the networks are unconnected there
is no electronic link from one bank to another.
In modern banking the payments can be operated fully electronically. When the payment transfer is
fully electric the client uses the home computer through, for example, the internet connection. The
use of the automatic teller machine in making payments or fund transfers also relies on electronic
connections. The merchant respectively has an electric link directly to the bank. If the electric
networks are not interconnected a client herself or a merchant must by-pass the missing link from a
payer’s account to a recipient’s account. This requires resorting to more or less manual manoeuvre
e.g. desk payment. In any case a client can pay by cash also. As we know a cash payment is still a
popular way to pay.
But the clients do not necessarily break even, although the networks are interconnected. It is
possible that a client incurs additional transaction costs when paying to some other bank’s client or
that the interbank transfer is of poorer quality than the intrabank transfer. The interbank transfer
may, for example, take more time than the intrabank transfer. Secondly, for the banks the
processing of payment transfers which require the interbank settlement may be more costly than
processing of intrabank transfers.
In this study we consider the forces which promote concentration in the payment service industry.
The clients are divided into old and locked-in clients and into new and uncaptive clients. The main
concern is to examine whether the existing market share of old clients induces the bigger bank to
capture the new clients also. The price level is considered as well. We consider both the case in
which the networks are unconnected and the case in which the networks are joined. In focusing on
the competition under unconnection, we consider two separate cases. In the first case price
discrimination between new and old clients is not allowed. Whereas, in the second case price

3
discrimination is allowed. Finally we consider the equilibrium when two banks have become
compatible under the circumstances in which the intrabank transfer is, however, superior to the
interbank transfer in terms of quality.
If the existing market share can function as an impediment to competition and if it favours the larger
bank, it thus affects also the terms of interconnection necessary to handle payments in an efficient
way. The analysis of this study helps us thus to understand also what is the position of the
unconnected small bank or what are the terms of entry. If the small bank suffers from the
cancellation of the interconnection contract more than the bigger bank, it is obvious that this can be
used as a threat in the negotiations concerning the pricing under interconnection and
compensations owing to interconnection.
In the basic framework of this section there are assumed to be two banks: typically one small bank
and one large bank. The assumed cost structure of the model does not lead necessarily to imperfect
competition. We abstract from possible increasing-returns technology. The banks offer identical
payment services. The electronic payment service, which the banks offer, is based on the use of own
PCs (through e.g. internet) so that the number of ATMs or the location of the bank branch has no
influence.2
We abstract thus also from all those factors related to the location or the other firm
specific attributes which through product differentiation stabilize the market and create a nice
equilibrium with many firms.
If the banks are not interconnected, the clients can make a fully electronic payment only to the same
bank’s clients. Paying to the rival-bank’s clients, the customer must resort then to the half-manual
procedure which includes positive transaction costs. The network-externality in the case considered
depends on the size of the clientele. This externality is wholly internalised only, if the banks join their
networks. But also under compatibility we can talk about network externality, if the interbank
transfer is of poor quality.
We assume that it is costly for the bank’s established client to switch to another bank. There are
remarkable transaction costs in closing an account with one bank and opening another with a
competitor. The switching costs - which may arise also due to psychology and which may be
artificially created to impede competition - are assumed to be so great that it is not profitable for the
old clients to switch to another bank. A new customer does not face switching costs. In any other
2
In many studies in bank competition, the location of the service point makes the products different a’ la
Hotelling (see. e.g. Matutes and Padilla, EER 1994).

4
aspect the clients are assumed to be identical. Thus we abstract from the possible cross subsidising
which arises when the customers differ from each other in the size of their purchases (see. Tarkka,
1995).
In literature the implications of switching costs are analysed widely. Many authors have recognised
that switching costs relax competition and increase the price level in a mature market (see Ausubel
(1991), Amstrong, Cowan and Vickers (1994) and Economides (1993)). The banks may, however,
use a bulk of those profits obtained from captive clients for the competition about new and
uncaptive clients. Klemperer (1987) showed that in the two-period model of Bertrand price
competition, even a small positive switching cost give enough shelter to set the second period price
of the captured (”old”) clients at the monopoly level. In the first period the knowledge of this
monopoly pricing makes the competitors to set the prices of new customers below the sort run
marginal costs. In this highly competitive situation the discounted profits from each customer
remain at zero level. Padilla (1992) considers the two-period model with no price discrimination.
He assumes that the clients are differentiated in respect to their awareness of the price offers.
Surprisingly, in an equilibrium ex ante identical firms have asymmetric market shares. According to
Padilla (1992) overall competition is, however, relaxed due to switching costs.
Farrell and Shapiro (1988) and Beggs and Klemperer (1992) analyze the overlapping-generations
duopoly and thus avoid many drawbacks related to two-period models. Farrell and Shapiro (1988),
showed that in the case where the clients can not be price discriminated, the bigger firm (in
duopoly) is satisfied to serve only captured clients. In this model of two overlapping generations the
markets shares would alternate so that the firm with no old clients captures always the new
generation. This fat cat phenomenon is present also in the more general setting with an infinite-
period market, analysed by Beggs and Klemperer (1992). They showed that when the young
generation’s switching costs are zero, the old are locked-in, and price discrimination is not allowed,
the firm who has more captive clients lets the smaller firm capture the young clients. In the duopoly
the market shares would finally converge to the steady-state level which is 50 per cent in the
symmetry. In the model of Beggs and Klemperer (1992) the prices in an equilibrium are, however,
above the competitive level, not least owing to a positive discount factor.
The above results, according to which, the smaller bank is ready for a fierce competition for new
clients, may, however, be reversed, if the bank’s current profits are constrained from below.
Holmström and Tirole (1997), who considered firms’ capital constraints, stress that low profits
restrict the capital investments of uninformed investors because of the moral hazard problem. In the

5
banking business especially this link is strong. Low profitability encourages the insiders for deceptive
behaviour, that makes the uninformed investors require that the insiders should increase their
capital investments. Losses may also be a sign of the ongoing abuse use of capital. Therefore the
losses related to aggressive market conquering lead easily to the restraint of capital that makes the
bank give up the aggressive strategy and shift back to the conduct which emphasises profitability.
The solvency ratios set by the central bank makes this shift in the bank’s conduct obligatory in the
case in which the bank itself is not yet ready for it. For these reasons, it could reasonable to assume
that bank’s profits are constrained to be non-negative. This constraint would restrict the smaller
bank’s possibilities to capture new generations, which in turn would increase the bigger bank’s
profits.
The recent literature has analysed the impacts of switching costs in regular and potentially
competitive industries. Farrel and Shapiro (1988) enlarge the discussion to concern also the effects
of scale-economies. According to them, the switching costs in combination with large enough fixed
costs could lead to a situation where the incumbent firm excludes the entrant. This result is not very
surprising, whereas, it is still unclear to what extent the network externality – in the absence of scale
economies – could strengthen the bigger firm’s incentives to capture the rest of the market.
We focus on the interplay between the network externalities and the switching costs. This extension
is natural, because the switching costs are typical in network industries especially. More specifically,
we consider the payment service industry, although our results could be generalised to concern
some other network industries as well. The results of this study show that network externality in
combination with switching costs leads easily to a highly concentrated market structure. The
originally larger bank may extend its market share both in the presence and in the absence of price
discrimination. These results may apply also when the networks are joined. Then the larger bank is
favoured by the poor quality of interbank transfers and by the low marginal costs associated with
processing intrabank transfers.
2. The model
There are only two banks: bank A and bank B. The banks are not necessarily interconnected. The
clients of bank A pay price tA for intrabank transfer. This transfer is fully electronic. If bank A’s
client makes a payment transfer to bank B’s client, she must pay price hB for it when the networks

6
are not connected. This interbank transfer is set by bank B and it is by nature half-manual. When
the banks are interconnected A’s client pays hA for the interbank transfer. The freely determined part
of this tariff is set by bank A. The interbank payment leads also to transaction costs for a client,
governed by a constant m. Under incompatibility m describes poor quality. Thus a fully electric
intrabank transfer is preferred to an interbank transfer. When a client pays to another client in the
same bank, the fully electric transfer is available. But in paying to another bank’s customer only the
payment transfer of poorer quality is available.
The clients are either young or old. We assume that there are T generations of old customers and
one generation of new customers. All generations are of equal size. A single client is only a negligible
small fraction of the whole generation. Let k denote the number of bank A’s old generations and let
T-k be the number of B’s old generations in current period. A young client turns to old after one
period. Like Beggs and Klemeperer (1992) we assume that any old client is expected to leave the
clientele with probability 1/(T+1), independently of their past history. In every period the number
of new clients emerging the clientele is exactly one unit. A young client chooses the bank without
any switching costs. The once-and-for-all switching costs for old agents are so large that the old
clients are practically locked-in that makes it possible to set their prices at the monopoly level.
We assume that in their expectations concerning the size of different banks' clientele, the new clients
take only into account the observed number of old clients in the beginning of each period. If,
however, a new client in her decision making is also assumed to take into account the behaviour of
other new clients, the outcome of the game would no longer be predictable. Even if the market
shares were the same in the beginning of the game, it could be strictly profitable, for example, for
bank A to capture the new clients, if all the new clients share the belief that bank A, instead of bank
B, is chosen by new clients. Thus, if we require that expectations should be fulfilled3
, there could
easily be two different equilibrium.
We abstract from these difficulties by assuming that the new clients evaluate the probability that A’s
client will pay to another customer in the same bank being k/T. The same probability concerning
bank B is (T-k)/T. Theses probabilities are updated at the beginning of each period. This
assumption about the clients’ expectations stresses the fact that the single new client in her decision
making is not ware of and does not take into account the intentions of other new clients. The banks
regard the clients’ expectations as given. In their own calculations concerning the capture, the banks
3
The concept used by Katz and Shapiro (1985) for expectations which are ex post fulfilled. Katz and Shapiro
(1985) have pointed out also that network externality is contingent on the expected number of network users.

7
as strategic players, however, take into accounts all the clients: the new and the old. So if bank A will
capture the new generation, bank A calculates that the probability that a single client uses A’s fully
electric service is rather (k+1)/(T+1) than k/T.
The clients demand bank's payment services insofar as the prices are at the reservation level or
below it. The highest possible price for bank A’s electronic service is tA
m
and for A’s half-manual
service hA
m
. For bank B the respective price limits are tB
m
and hB
m
. It is assumed that tA
m
= tB
m
and
hA
m
= hB
m
. From now on these price limits are denoted by tm
and hm
. It is assumed also that tm
= hm
+ m.
Following Farrel and Shapiro (1988) we assume that the clients are myopic.4
This means that the
new client chooses the bank whose offer concerning the current period is the best. This simplifying
assumption can be defended in the context where the firms do not commit to long term contracts. It
turns out that the decisions which the new clients make are ex post rational. So, the myopia
assumption does not lead to serious consistency problems. Furthermore, when price discrimination
is allowed, the price for old clients is foreseen, wherefore the myopia is then in no contradiction with
rationality.
2.1. No price discrimination and no interconnection
We assume now that the bank can not discriminate between the old and young clients.
In this setting the price of the interbank transfer paid by A’s clients is set by bank B (and vice versa).
A young client chooses bank A, if
(1)
k
T
t
T k
T
h m
T k
T
t
k
T
h mA B B A+
−
+
−
+ +( ) ( )
where m describes the client’s transaction costs related to the use of interbank transfer. If the left
hand side of (1), denoted by PA is the same as the right hand side of (1), denoted by PB, the clients
4
Padilla (1992) also assumes that the clients are myopic. Beggs and Klemperer (1992) are critical toward this
assumption. The central result of their model is, however, obtained even if the customers were myopic.
According to them, it is the fact that the firms discount the future which is sufficient to imply that the desire to
exploit old customers outweighs the desire to attract the new ones. The result which states that the market
shares of the two symmetric firms converge to equality from any starting is obtained as well, although the
customers were myopic.

8
choose arbitrarily between A and B. Statistically half of the new clients are then expected to choose
A. Later in the proof of proposition 1, we suppose that if PA ≈ PB, the clients choose arbitrarily,
because they can not see small differences between the current values of tA and tB. This supposition
also guarantees that in nearly symmetric situation in which T-k k T-k+1 or k T-k k+1,
neither A nor B can get advantage in relation to its rival. Condition (1) is symmetric, because the
expectations are static in the short run, and thus the derived equilibrium is unique.
Assume that tA = tB and that hA = hB. Bank B’s price exceeds A’s price if
(2) .0)(
)2()2(
+
−
+
−
− mh
T
Tk
t
T
Tk
AA
From the condition (2) we see directly that the customers would choose the larger bank if tA
hA+m. The network externality thus favours the larger bank, which is assumed to be bank A.
With m = 0, the fully electric service would be a perfect substitute for the half-manual service.
Because m 0, the fully electric service is, however, preferred when tA = hA.
Let ct be constant marginal cost of producing a fully electric service. The respective cost for a half-
manual service is ch. It is well known that marginal costs in producing a fully electric service are
lower than in producing a half-electric (desk) service. Therefore we assume that ct ch. We assume
that hm
≥ ch and show that in the special case where hm
- ch = 0, it is the bigger bank which captures
the new clients. Furthermore, we show that when dh (≡ hm
- ch) increases the “fat cat” effect may
become dominant so that the small bank captures the new clients, even if dh tm
- ct. We assume
that in any case tm
- ct, dh ≥ 0.
In this highly symmetric game two banks can differ from each other only in respect to the size of
their old clientele. Denote by VA,t(k) bank A’s net present value in period t when it has k old
generations. Then bank B’s net present value in period t must be VB,t(T-k). It is assumed that k
T-k, so that A is originally larger than B in terms of the size of the clientele. If bank A captures a
new generation in period t but releases during periods t +1, ..., i, A has in period i (k+1)(1 -
1/(T+1))i
generations. But if bank A releases in periods t, ..., i, the number of its old generations in
ith
period is k(1 - 1/(T+1))i
. In the former case B has in ith
period (T-k+1)(1-1/(T+1))i
generations and in the latter case (T- k)(1 - 1/(T+1))i
generations.

9
In the considered context, it is in the banks’ interests to set the price of the half-manual service,
concerning all clients, at the monopoly level which is hm
. Because the switching costs are high
enough so that the old clients are locked-in, and because price discrimination is allowed, all the
prices concerning the old customers can be set at the monopoly level. The banks compete with the
price for fully electronic services, concerning new customers.
New clients choose bank A, if condition (1) is valid. Then bank A calculates that it captures a new
generation and its current profits are
(3) ΠA A t A h
k
T
t c
k T k
T
h c=
+
+
− +
+ −
+
−
( )
( )
( )( )
( ),
1
1
1
1
2
where hA = hm
. If, on the other hand, it is bank B who captures the new generation so that (1) is
invalid, A’s profits are at the most
(4) R
k
T
t c
k T k
T
h cA A
m
t
m
h=
+
− +
− +
+
−
2
1
1
1
( )
( )
( ),
In (4) A sets tA on the monopoly level. In the considered model the bigger banks benefits from the
bigger clientele for two reasons. Firstly, its clients suffer from transaction costs relatively less than
the rival’s clients that lets the bigger bank set the price for fully electronic service at the higher level
than its rival. Secondly, the bigger bank produces fully electronic services relatively more than its
rival, which favours the bigger bank, because ct ch.
In the dynamic framework, also the long terms effect of the decisions should be taken into account.
The banks maximise their performance in terms of net present values. Define XA,t(k) to be a strategy
in which bank A lets bank B to capture a new generation in current period. Bank A’s net present
value is then
(5) XA,t(k) = RA,t(k) + δVA,t+1(k(1 - 1/(T+1))),
where R, which denotes the current period profits from old locked-in clients, is defined in equation
(4). Respectively, banks B’s net present value would then be
(6) WB,t(T-k) = ΠB,t(T-k) + δVB,t+1((T- k+1)(1 - 1/(T+1))),
where ΠB,t(T-k) describes B’s current period profits from old and young. It has an equation

10
(7) ),(
1
)1(
)(
1
)1(
)(
2
, hBtBtB ch
T
kTk
ct
T
kT
kt −
+
+−
+−
+
+−
=−Π
where hB = hm
. Let WA,t(k), on the other hand, denote a strategy in which A captures the new clients
in the current period. This strategy has then an expression
(8) WA,t(k) = ΠA,t(k) + δVA,t+1((k+1)(1 - 1/(T+1))),
where ΠA,t(k) represent current profits defined in an equation (3). The corresponding net present
value for bank B is then
(9) XB,t(T-k) = RB,t(T-k) + δVB,t+1((T- k)(1 - 1/(T+1))).
In (9) the current profits from old clients has an expression
(10) R (T - k)B,t =
−
+
− +
+ −
+
−
( )
( )
( )( )
( ).
T k
T
t c
k T k
T
h cB
m
t
m
h
2
1
1
1
Define tA,t+h
o
to be A’s price offer in period t+h concerning new clients, and respectively, tB,t+h
o
B’s
price offer concerning period t+h as well. The price offer is credible, if the bank can commit to it
without losses. For example, bank A can by its price strategy affect bank B’s prices insofar as it can
stick to its offer in the case A’s offer beats B’s offer and becomes thus effective. So long as bank A
captures new clients, B’s profits do not, necessarily, depend on price offer tA,t
o
. If B foresees that it
will not be profitable to capture new clients in the future, B is content with taking the full advantage
of its old clients by charging for them monopoly price tm
.
It must be noticed that k T-k. We try to find out, does there exists such a credible price offer tA,t
o
which makes the client to choose bank A in period t subject to condition WA,t(k) - XA,t(k) 0 and
WB,t(T-k) - XB,t(T-k) 0. In other words, it is required that under current prices tA,t and tB,t(tA,t),
clients choose bank A, to whom the capture is profitable, while the capture would be unprofitable for
B. The problem in this setting are the prices offers in the future. Before we go to proposition 1 and
its proof, some notation is introduced, in order to express the present net values also as functions of
current and future prices.
Define tA = tA,t+1, tA,t+1, tA,t+2.,... to be the vector of A’s future prices when the number of its old
generations is k(1- 1/(T+1)) in period t+1. Whereas, tA
*
= tA,t+1
*
, tA,t+2
*
tA,t+3
*
,..., denotes then future
prices when the number of A’s old generations is (k+1)(1- 1/(T+1)) in period t+1. Respectively,

11
B’s prices are then tB = tB,t+1, tB,t+2, tB,t+3.,... when the number of B’s old generations is (T-k)(1-
1/(T+1)) in period t+1, and tB
*
= tB,t+1
*
, tB,t+2
*
, tB,t+3
*
,.. , when B has (T-k+1)(1- 1/(T+1)) old
generations in period t+1.
When a bank releases from now on, and sets the future prices of old clients at the maximum level,
the discounted profits from period t+1 onwards are
δDA,t+1((k+1)(1 - 1/(T+1)) ,tA
m
) = ∑
∞
=1i
δi
RA,t+1((k+1)(1 - 1/(T+1))i
, tA
m
)
δDA,t+1(k(1 - 1/T),tA
m
) = ∑
∞
=1i
δi
RA,t+1(k(1 - 1/(T+1))i
, tA
m
)
(11) δDB,t+1(T-k+1)(1- 1/(T+1)), tB
m
) = ∑
∞
=1i
δi
RB,t+1((T-k+1)(1 - 1/(T+1))i
, tB
m
)
δDB,t+1((T-k)(1 - 1/(T+1)),tB
m
) = ∑
∞
=1i
δi
RB,t+1((T-k)(1 - 1/(T+1))i
, tB
m
)
Denote DA,t+1
k+1
≡ DA,t+1((k+1)(1 - 1/(T+1)),tA
m
), DA,t+1
k
≡ DA,t+1(k(1 - 1/T),tA
m
),
DB,t+1
T-k+1
≡ DB,t+1((T-k+1)(1 - 1/(T+1)), tB
m
) and DB,t+1
T-k
≡ DB,t+1((T-k)(1 - 1/(T+1)),tB
m
).
Let tA,t+h
I
denote such tA,t+h under which A is indifferent between the capture and the release in
period t+h when the number of A’s old generations is k(1- 1/(T+1)) in period t+1 and when VA,t+1
is restricted to be DA,t+1. Then tA
I
= tA,t+1
I
, tA,t+2
I
, tA,t+3
I
, .. is the respective price vector. When the
number of A’s old generations is (k+1)(1- 1/(T+1)) in period t+1 we use notation tA
I*
= tA,t+1
I*
,
tA,t+2
I*
, tA,t+3
I*
, ... for the price vector which implies the indifference. The corresponding vectors for
bank B are tB
I
and tB
I*
. By these definitions
VA,t+1((k+1)(1 - 1/(T+1)), tA
I*
) = DA,t+1
k+1
, VA,t+1(k(1 - 1/(T+1)), tA
I
) = DA,t+1
k
,
VB,t+1((T-k)(1- 1/(T+1)),tB
I
) = DB,t+1
T-k
and VB,t+1((T-k+1)(1- 1/(T+1)),tB
I*
) = DB,t+1
T-k+1
.
Proposition 1 states the main result of this study. We assume that hm
- ch, is very small in relation to
price premium tm
- ct,. For simplicity it is assumed that hm
= ch.

12
Proposition 1. If hm
= ch, then it is the bigger bank who captures the new clients, even if price
discrimination is not allowed.
The exact proof is given in Appendix. In the proof it is assumed that that both bank A and bank B
make the price offers concerning future also and that they both have made price offers which make
the rival indifferent between the capture and the release in the future. In other words, tB,t+h(tA,t+h
o
) =
tB,t+h
I
and tA,t+h(tB,t+h
o
) = tA,t+h
I
when h ≥ 1. If under these assumptions A would capture in the
current period, and if it turns out that it will capture also from then on insofar as it is bigger than B,
its price offers concerning the future are credible. In fact, bank A’s capability to capture in the
current period proves that it can also capture in the future and thus set tA,t+h
o
so that tB,t+h(tA,t+h
o
) =
tB,t+h
I
or slightly below it. It is shown also that although bank B has made A indifferent in the future,
bank B does not capture in current period. This refers to the fact that bank B can capture neither in
the future periods. Thus B’s price offer concerning the future is not credible that means that for the
credible offers tA,t+h(tB,t+h
o
) ≥ tA,t+h
I
. When A captures, tA,t+h = tA,t+h
o
so that tB,t+h(tA,t+h) = tB,t+h
I
. A
makes then B indifferent. Denote the vector tA
*
(tB
I+
) - tA
I*
(with elements tA,t+h
*
(tB,t+h
I+
) - tA,t+h
I*
) by
∆tA
*
where tB,t+h
I+
describes B's indifference price-level when the number of its old generations is T-
k-1 in the current period. Then ∆tA≡ tA(tB
I
) - tA
I
(with elements tA,t+h(tB,t+h
I
) - tA,t+h
I
). Insofar as the
changes ∆tA
*
and ∆tA lead to
(12) VA,t+1((k+1)(1 - 1/(T+1)),tA*) - VA,t+1(k(1 - 1/(T+1)),tA)
VA,t+1((k+1)(1 - 1/(T+1)),tA
I
*) - VA,t+1(k(1 - 1/(T+1)),tA
I
),
the original result, according to which, bank A captures in the current period, is only confirmed. The
higher B’s credible price offers in the future are, the more profitable it is for bank A to capture the
current new generation. At the first stage of the proof we show that when tA,t+h(tB,t+h
o
) = tA,t+h
I
and
tA,t+h
*
(tB,t+h
o
) = tA,t+h
I*
(when h ≥ 1) there exists such an price offer tA,t
o
and tB,t+h(tA,t+h
o
) = tB,t+h
I
which make the client to choose bank A in period t subject to condition WA,t(k) - XA,t(k) 0 and
WB,t(T-k) - XB,t(T-k) 0.
At the second stage of the proof we show then that (12) is fulfilled. In the considered setting it is
sufficient to show that ∆tA
*
≥ ∆tA. This guarantees that inequality (12) is valid. At the second stage
of the proof it is shown also that so long as A charges so that tB,t+h(tA,t+h
o
) is slightly below tB,t+h
I
,
bank B cannot expect its net present value VB,t+1((T- k)(1 - 1/(T+1))) of being greater than
DB,t+1
T-k+1
, if B captures the current new clients. This claim is defended by showing that in symmetry

13
or nearly symmetry bank B can expect that its net present values is at the highest DB,t+1
T-k
. Owing to
ruinous competition, corresponding to unstable situation in the neighbourhood of symmetry, the net
present values can easily drop even below this level.
The proposition 1 states that when the profit making opportunities in half-manual transfers are non-
existent, the big bank will capture the new clients although price discrimination is not allowed. The
proof of proposition 1 shows that by setting the user charge tA,t consistently so that tB,t(tA,t) = tB,t
I
,
the bank A can be assured that B does not capture new clients. When prices are set non-
cooperatively and simultaneously there is, however, no Nash-equilibrium for prices. Suppose that A
charges along tB,t(tA,t) = tB,t
I
and B sets tB,t = tm
. Then A has an interest to raise its price offer. If A,
however, raises its offer so that tB,t(tA,t) tB,t
I
, it is in B’s interests to lower its own price all the way
below tB,t(tA,t), even if B expects that its net present value of future profits do not exceed the break-
even level which is the discounted maximum profits from old clients.
It turns out also that the price level tends to raise above the competitive level.
Corollary 1. When hm
= ch, the price margin concerning current price for new clients is at the most
]
)
1
1
1(1
)
1
1
1(
)1(
]1)(2)[(
)1(
)(2
)[(
2
2
22
3
,
δ
δ
+
−−
+
−
+−
+−−
−
+−
−
+
−
−=−
T
T
kTk
kTkT
kTk
kT
k
Tk
ctct t
m
ttA
The proof is given in appendix. Corollary 1 shows that tA,t tm
. The higher is discount factor δ, the
thinner is the price margin. This result shows that the larger present value of an increment in
clientele implied by higher δ, makes the banks fight more fiercely for the new clients. Corollary 1
tells also that the bigger is k, the number of the bigger bank’s old clients, the wider is the price
margin. Price margin tA,t - ct is almost always positive. Only if δ is close to one and if k is much
bigger than T-k, the current price can be negative.
The case without network externality
Above, it was the network externality which resulted in increasing returns on old clients and in the
non-existence of a competitive equilibrium. If the network externality is removed, the industry

14
becomes regular and the larger banks starts to behave like a “fat cat” who is satisfied to its old
clients.
Let us consider a model where bank A who has originally k old generations and bank B T-k old
generations so that k T-k. Each bank produces only one service with user charges tA,t and tB,t.
Bank A's monetary earnings from its old generations are at the most k(tm
- ct) in the current period.
The dynamics of the model is assumed to be the same as previously. Because there are no network
externalities, the new clients choose A, if tA,t tB,t. We show that there cannot exist an equilibrium in
which it pays for A to capture new generations and in which it is not profitable for B to capture new
generations.
Propositions 2. In the absence of network externality, it does not pay for A to capture all the new
generations in the future.
The proof is given in Appendix. The proposition excludes the possibility that the larger bank
captures and thus confirms the previous results (see Farrel and Shapiro, 1988, and Beggs and
Klemperer, 1992) who state that when price discrimination is not allowed it is the smaller firm who
captures the new clients. It turns out that this result can be obtained even in the presence of
network externality, if the price premium hm
- ch, from producing half manual services, is not too
low in relation to premium tm
- ct.
The case: hm
- ch 0
Denote dh ≡ hm
- ch. We will show that when the profitability of half-manual transfers increases
(described by dh) the terms for capture may switch so that it is no more profitable for the larger
bank to capture new generations, although all the time dh tm
- ct.
Above it has been assumed dh is zero or at least very small in relation to margin tm
- ct. When dh
increases in relation to tm
- ct much enough proposition 1 is possibly no longer valid. In fact,
assuming that tm
- ct dh 0, two additional terms should be added to condition (a5) (in
appendix) which tells which one of two banks captures new clients. Those new sum terms are
].)1)(()1([
)1()1)((
2 22
22
+−++−
+−+−
−
− kkTkTk
kTkkT
Tk
dh
and

15
].
)
1
1
1(1
)
1
1
1(
)1()1)((
)1))(1)()(2(
)
1
1
1(1
)
1
1
1(
)1()1)((
)12)(3)(2(
[
22
2
2
22
δ
δ
δ
δ
+
−−
+
−
+−+−
+−−−
−
+
−−
+
−
+−+−
++−−
−
T
T
kTkkT
TkTKTK
T
T
kTkkT
TkTkTk
dh
It is noteworthy that above the absolute value of dh‘s negative coefficient is greater than the
coefficient of (tm
- ct ) in formula (a5). This would suggest that, given (tm
- ct ), it does not pay any
more for A to capture the new clients so long as it is bigger than B.
Corollary 2. In the presence of network externality, it does not pay for A to capture all the new
generations in the future, if dh is close to tm
- ct when 0 dh ( tm
- ct ).
Proof. Proof follows the proof of proposition 2. In the case considered A would also make losses, if
it captures new clients, given that the both banks cannot earn in the future more than the discounted
profits (11). If A, however, would decide to capture new clients, VA,t+1
k+1
- VA,t+1
k
would become
smaller than DA,t+1
k+1
- DA,t+1
k
, because A should lower the offers below tA,t
I
and because tA,t(tB,t
I
) -
tA,t
I
decreases in k. This only confirms the fact that it is not profitable for A to capture the new
clients.
Thus, it is shown that the roles of two banks may switch so that the smaller bank will actually
capture new generations when the profits from half-manual service relatively increase.
2.2. Price discrimination and no interconnection
In the banking and in the provision of payment services, the price discrimination could take various
forms. It is not necessarily reasonable to lower the prices of payment services for that part of
clientele which is assumed to be noncaptive. The youngsters do not use payment services so heavily
as captive clients do. Technically the price discounts concerning the youngsters only would be rather
difficult to carry out. It may thus be more convenient to grant to the young loans on favourable
terms or to attract them with lucrative deposit opportunities. Of course the difficulties to identify the
new client from the old, and the possible arbitrage which take advantage from the offered price
discounts, set limits to price discrimination. We, however, assume next that it is possible to restrict

16
the price discounts to concern the young and uncaptive clients only. The old and captive clients have
then to pay monopoly price for the payment services.
An inequality (1) describes still those terms under which the client chooses bank A, whereas when
bank A captures the new clients its current profits are
(13) ΠA A t A
m
t
m
h
k
T
t c
k k
T
t c
k T k
T
h c=
+
+
− +
+
+
− +
+ −
+
−
( )
( )
( )
( )
( )( )
( ).
1
1
1
1
1
1
Bank B’s profits, in the case it captures the new clients, are
(14) ΠB t B t B
m
t
m
ht k
T k
T
t c
T k T k
T
t c
k T k
T
h c, ( )
( )
( )
( )( )
( )
( )
( ).− =
− +
+
− +
− − +
+
− +
− +
+
−
1
1
1
1
1
1
Equations (4) and (10) still describe the profits from the old clients. Also the dynamics can be
presented in the same way as in the case without price discrimination. It turns out that
Proposition 3. If hm
= ch, then it is the bigger bank who captures the new clients, when price
discrimination is allowed.
The proof is in Appendix. The right hand side of inequality (c1) which describes tB,t
I
, is
).(
)1(
1
)1(
1)(2
)1(
)(
1,
1
1,
kT
tB
kT
tBt
m
B DD
kT
T
c
kT
kT
t
kT
kT −
+
+−
+ −
+−
+
−
+−
+−
+
+−
−
− δ
In the strategy in which A captures the new clients, A sets tA,t so that tB,t(tA,t) is slightly below tB,t
I
.
The above expression shows that tB,t
I
is even negative. Despite this aggressive pricing, the
upperbound for the average current price which is (1/(k+1))tB,t
I
+ (k/(k+1))tm
is even higher than
in the case in which price discrimination is not allowed. This result can be derived using the above
formula and the right hand sides of inequalities (a3) and (c1) which express tB,t
I
in each case.
3.3. Interconnection and no price discrimination
We consider next the situation where the banks are interconnected.5
If bank A’s client pays to
another client in the same bank, the price for the transfer is tA. If A’s client pays instead to bank B’s
5
The behaviour of banks under compatibility is analysed also in Kauko (1998).

17
client, the interbank links must be used. The price for an interbank transfer can be divided into two
parts which are hA‘ and IA . Then hA‘ stands for A’s share and IA for B’s share of that price for a
compensation to process the transfer. It will turn out that both for bank A’s and bank B’s clients the
expected volume of interbank transfers is (k+1)(T-k)/(T+1) and thus the same. We consider an
institutional setting in which A and B have agreed on IA and IB, and due to reciprocity have set IA =
IB. In so far as competition authorities had any influence on interbank transfers, the above equality
would apparently be a minimum requirement. For the simplicity, we assume that IA = cI when cI is
the receiving bank’s marginal costs to complete the processing of an interbank transfer. Bank A’s
profits are then
)]()'[(
1
))(1(
)(
1
)1( 2
IBIhAtAA cIcch
T
kTk
ct
T
k
−++−
+
−+
+−
+
+
=Π
when ch describe marginal costs of the whole interbank transfer. Regarding IA as given, we can write
hA = hA‘ + IA, and because IA= IB, hA = hA‘ + IB also. Thus the above equation reduces to equation
(3). We can still consider tA and hA as A’s competition instruments, although it is actually hA‘ which
is A’s decision parameter.
We consider a situation in which the transfer which uses interbank networks is of quality inferior to
the intrabank transfer. Poor quality means delays in transfers, and worse reliability and safety than
in intrabank transfers. The income from the float is assumed to be part of hA. All the expenses from
the unit interbank payment for the client are hA + m when m describes the influence of poorer
quality. We still assume that tA does not exceed the reservation price tA
m
and hA the reservation price
hm
. Also the relation tm
= hA
m
+ m, is assumed to be valid.
When the banks are interconnected inequality (1) is replaced by inequality
(15) ),()( mh
T
k
t
T
kT
mh
T
kT
t
T
k
BBAA ++
−
+
−
+
where hA = hA‘ + IA and hB = hB‘ + IB. When (15) is in force, the client chooses bank A. Equations
(3) and (7) describe still A’s and B’s current profits when they capture the new clients. It is no
longer clear that the banks would compete with prices tA and tB. Assume that bank A keeps the left-
hand side of (15) constant. Then ∂tA/∂hA = -(T-k)/k and ∂hA/∂tA = -k/(T-k). Keeping the user cost
(which is the left hand side of (15)) constant, it follows then from equation (3) that

18
∂
∂
ΠA
At
k k
T
=
+ +
+
( )( )1 2 1
1
and .
)1(
))(1(
+
−+
=
Π
Tk
kTk
hA
A
∂
∂
This shows that ∂ΠA/∂tA ∂ΠA/∂hA, so that it pays for A to set tA = tA
m
and compete with hA given IA
and IB. Similarly it is obtained for bank B the conditions
∂
∂
ΠB
Bt
T k
T
T k k T k
k T
=
− +
+
−
− − +
+
( ) ( ) ( )
( )
1
1
1
1
2
and
∂
∂
ΠB
Bh
k T k
T
k T k
T k T
=
− +
+
−
− +
− +
( ) ( )
( )( )
,
1
1
1
1
2
which describe how much profits change while keeping the price condition concerning the clients
constant. It follows from the equations above that ∂ΠB/∂tB ∂ΠB/∂hB. Thus it is also in bank B’s
interests to set tB = tB
m
and compete with hB. We will show that also under characterized
interconnection the smaller bank is easily loosing its market share, if the quality of interbank
transfers is inferior.
Proposition 4. If tm
= hm
+ m, hm
= ch and tm
ct, then it is the bigger bank who captures the new
clients, even if the banks are interconnected and if price discrimination is not allowed.
The proof of proposition 4 follows closely the lines of the proof of proposition 1 (see Appendix).
The proposition shows that the larger bank’s aggressive nature does not necessarily vanish when the
banks become compatible. But it is clear that the implications of corollary 2 also are valid now. If
under compatibility, the interbank transfer does not differ from the client’s view point much from
the intrabank transfer, and if there are no remarkable differences in marginal costs, it is possible that
compatibility would make the larger bank a fat cat. The concentration of the industry would no
longer be such a problem which in the long run depresses competition.
Conclusions
We consider bank competition in payment services in the two different situations. In the first case
the networks are not joined. Incompatible networks would, however, complement each other. A
payment transfer to the same bank’s customer is fully electronic. But a transfer to another bank’s
client is by nature half-manual. When the customer must resort to half-manual service her costs
consists of, in addition to the regular user price, transaction costs which arise when one makes
payments in the half-manual way. In fully electric transfers the transaction costs are zero. In the

19
second case the networks are joined. Then an interbank transfer is of quality inferior to an intrabank
transfer, or at least, the marginal costs to process an interbank transfer are higher than the
respective costs in processing an intrabank transfer.
The banks are assumed to compete with prices. The banks differ from each other only in the
numbers of their old clients. The switching costs which the customers face are large enough to deter
any movement from one bank to another. Only the new clients who enter the market are free from
switching costs.
Let us consider first the results in the first case (when the banks are not interconnected). The
probability that the client must pay to another bank’s client and thus resort to half-manual
procedure is then greater for the smaller bank’s customer than for the bigger bank’s customer which
benefits the clients of the bigger bank. The marginal costs of processing fully electronic transfers are
assumed to be lower than the respective costs in processing half-manual transfers. Also this favours
the bigger banks on the expense of the smaller rivals.
When the price discrimination is not possible, it could be too big sacrifice for a larger bank to lower
also the prices of the old clients below the monopoly level in order to capture the new clients. This
”fat-cat” effect is noticed in related literature (see Farrell and Shapiro (1988), and Beggs and
Klemperer (1992)). In the case considered the network externality may improve, however, the
bigger bank’s clients’ position to an extent that in order to capture new clients the bigger bank needs
lower the prices only moderately. Network externality makes the discounted marginal profits from
the old clientele increasing in the number of old clients. These effects dominate the “fat cat“ effect,
if a price premium associated with the supply of half-manual services is zero or small enough. If this
premium increases close to a (maximum) premium associated with the supply of fully electronic
transfers, the “fat cat” effect starts to dominate. This refers that there is a threshold-value for the
price margin in question. Below this threshold concentration is observed, whereas above this
threshold the market shares become smooth. Typically, in the situation in which the larger bank
captures all the new clients, the overall price level remains above competitive level.
It pays for the larger bank to capture the new clients also when price discrimination is allowed, if the
premium from half-manual services is small. In that case the current price of the fully electronic
service for the new clients is even negative. Despite this, the average price of these services for the
whole clientele would be even higher than in the case in which price discrimination is not allowed.

20
One can state that when competitive pricing is restricted to concern new clients only, the larger
bank is even better-off than in the absence of price discrimination.
Finally we consider the case in which the payment networks are joined. Because the quality of
interbank transfer is inferior to the quality intrabank transfer, and the marginal costs in processing
intrabank transfers are lower than in transfers which use interbank links, the larger bank obtains a
competition advantage. If the price premium in interbank transfers is small enough, it is profitable
for the larger bank to capture the new clients, whereas, if there this premium is almost of the equal
size as the premium associated with intrabank transfers, it pays for the larger bank to let the rival
capture the new clients.
For two reasons the interconnection of payment networks, however, could favour competition.
Firstly, when the networks are joined, the quality difference between an intrabank transfer and an
interbank transfer may be smaller than the respective difference between a fully electronic transfer
and a half manual transfer in unconnected networks. Secondly, also the differences between the
marginal costs in processing an intrabank transfer and an interbank transfer may be smaller than the
respective cost difference associated with the processing of a fully electronic and a half-manual
transfer. Thus it is possible that concentration occurs when networks are not linked; and that the
market become more fragmentary when networks are connected.
In this study we abstract from all factors which generate product differentiation and as such explain
stationarity in the market shares. We focus purely on the interplay between networks externality and
switching costs. The world of the model considered does not necessarily correspond to that world in
which the banks operate currently. But it more or less pictures the world toward which retail-
banking is also developing. The geography loses its former importance and the clients rely
increasingly to the PC-access in making payments and fund transfers. Our study shows that the
stable equilibrium in which each bank keeps its former market share or in which small banks
become bigger, may disturb.

21
References
Armstrong, M., S. Cowan and J. Vickers, 1994,”Regulatory reform, economic analysis and British
experience”, The MIT Press, Cambridge.
Ausubel, L., 1991, “The failure of competition in the credit card market”, The American Economic
Review, 81, 51-81.
Beggs, A. and P. Klemperer, 1992, ”Multi-period competition with switching costs”,
Econometrica,60, 651-666.
Economides, N. , 1993, Network economics with applications to finance, Financial Markets,
Institutions and Instruments, 5, 89-97.
Farrel, J. and C. Shapiro, 1988, ”Dynamic competition with switching costs”, RAND Journal of
Economics, 19, 123-137.
Holmström, B. and J. Tirole, 1997, “Financial intermediation, loanable funds, and the real sector”,
The Quartely Journal of Economics, vol CXII, 663-691.
Katz, M. and C. Shapiro, 1985, ”Network externalities, competition and compatibility”, American
Economic Review, 75, 328-335.
Kauko, K, 1998, ”Developing the interbank payment system, Efficiency of public versus private
investments ”, Bank of Finland, Discussion Papers 28/98.
Klemperer, P., 1987, ”Market with consumers switching costs”, The Quartely Journal of
Economics, 102, 375-394.
Matutes, C. and J. Padilla, 1994, ”Shared ATM networks and banking competition”, European
Economic Review, 38, 113-1138.
Padilla, A.J., 1992, ”Mixed pricing oligopoly in markets with consumer switching costs,
International Journal of Industrial Organization, 10, 393-411.
Tarkka, J.,1995, ”Approaches to deposit pricing: a study in the determination of deposit interest and
bank service charges”, Bank of Finland Studies, E:2.

22
Appendix
The proof of proposition 1
Stage one of the proof
Formally we consider does there exist such tA,t and tB,t(tA,t) which make the new generation to choose
bank A so that
(a1) ΠA,t(k,tA,t) - RA,t(k,tA
m
) + δ(DA,t+1
k+1
- DA,t+1
k
) 0
and
ΠB,t(T-k,tB,t(tA,t)) - RB,t(T-k,tB
m
) + δ(DB,t+1
T-k+1
- DB,t+1
T-k
) 0.
Taking into account equations (6) - (13), inequalities (a1) can be expressed in the form
(a2) t
k
k
t
k
k
c
T
k
D DA t A
m
t A t
k
A t
k
, , ,
( ) ( ) ( )
( )
+
+
+
+
−
+
+
−+
+
+
2
2 2 2 1
1
1
1
2 1
1
1
1
δ
and
(a3) ).(
)1(
1
)1(
1)(2
)1(
)(
1,
1
1,222
2
,
kT
tB
kT
tBt
m
BtB DD
kT
T
c
kT
kT
t
kT
kT
t −
+
+−
+ −
+−
+
−
+−
+−
+
+−
−
δ
The right hand side of (a2) is tA,t
I
which defines the level at which A is indifferent between the
capture and the release, given it gets in the future at the most only DA,t+1
k+1
(when A captures in
current period) or DA,t+1
k
(when A releases in current period). In the same way, the right hand side
of (a3) is tB,t
I
which defines a break-even point for B when B's future prospects promise at the most
DB,t+1
T-k
or DB,t+1
T-k+1
. We assume that also in nearly symmetry when T-k k T-k+1 or k T-k
k+1, tA,t
I
- tB,t
I
 η, when η describes a negligible small price difference which the client does
not take into account.
According to condition (1), a young client still chooses bank A, if
tA,t = (T-k)/k tB,t + (2k-T)/k (hA
m
+m) – η,

23
Because η is negligibly small, we can ignore it (when k T-k+1). Thus, subject to condition (1),
(a2) transforms into form
(a4)
t
T k
T k
h m
k
T k k
t
k k
T k k
c
k T
T k k
D D
B t
m
A
m
t
A t
k
A t
k
,
, ,
( )
( )
( )
( )( )
( )
( )( )
( )
( )( )
( ).

−
−
+ +
− +
+
+
− +
−
+
− +
−+
+
+
2
1
2 1
1
1
1
3
2 2
2 1
1
1δ
From (a3) and (a4) it obtained an equilibrium condition under which A captures new clients.
Taking into account that tA
m
= tB
m
≡ tm
, this condition can be presented in the form
( )
( )
( ) [
( )
( ) ( )( )
] [
( )
( )
( )
( )( )
]
( )
( )( )
( )
( )
( ) ., , , ,
2
1 1
2 1
1
2 1
1
1
1
1
1
0
2
2
3
2 2 2
2 1
1
1 2 1
1
1
k T
T k
h m
T k
T k
k
T k k
t
T k
T k
k k
T k k
c
k T
T k k
D D
T
T k
D D
m m
t
A t
k
A t
k
B t
T k
B t
T k
−
−
+ +
−
− +
−
− +
+
− +
− +
−
+
− +
+
+
− +
− −
+
− +
− +
+
+ +
− +
+
−
δ δ
After manipulation and after setting tm
= hm
+ m, the above inequality transforms into the form
(a5)
( )( ( ) )
( )( ) ( )
( )
( )
( )( )
( )
( )
( ) ., , , ,
2 3 2 1
1 1
1
1
1
1
0
2 2
2 1
1
1 2 1
1
1
k T k T k T
T k k T k
t c
k T
T k k
D D
T
T k
D D
m
t
A t
k
A t
k
B t
T k
B t
T k
− − + +
− + − +
−
+
+
− +
− −
+
− +
− +
+
+ +
− +
+
−
δ δ
In (a5) the first row is clearly positive insofar as (2k-T) 0. We then consider the sum terms in the
second row. Using equations (11) it can be shown that
(a6)
])
1
1
1(1)[1(
)
1
1
1)(12(
)(
2
2
1,
1
1,
δ
+
−−+
+
−+
−=− +
+
+
T
T
T
k
ctDD t
mk
tA
k
tA
and that
(a7) ,
])
1
1
1(1)[1(
)
1
1
1)(1)(2(
)(
2
2
1,
1
1,
δ
+
−−+
+
−+−
−=− −
+
+−
+
T
T
T
kT
ctDD t
mkT
tB
kT
tB
from which follows that the second row of (a5) can be presented in the form

24
(a8)
.
)
1
1
1(1
)()
1
1
1(
)]1)(2(
)1(
1
)12(
)1)((
[
2
2
2
2
δ
δ
+
−−
−
+
−
+−
+−
−
+
+−
T
ct
TkT
kT
k
kkT
k
t
m
The above expression is positive, if
(a9) ,0]12)(3)[2( ++−− TkTkTk
which shows that only requirement 2k-T 0 is needed. So we have shown that inequality (a5) also
is valid when 2k-T 0, which completes the first stage of the proof.
The second stage of the proof
The left hand side of expression (a5) tells the difference tB,t
I
- tB,t(tA,t
I
) when the future prices are at
the level tA,t+h
I
and tB,t+h
I
(with h 0). By multiplying this term by (T-k)/k, it is obtained from (a5)
for the difference tA,t(tB,t
I
) - tA,t
I
(≡ ∆tA,t) an expression
(a10)
,0)(
)1(
)1)((
)(
)1(
)1(
)(
)1()1(
)12)(3)(2(
1,
1
1,21,
1
1,2
22
−
+−
+−
−−
+
+
+
−
+−+
++−−
−
+
+−
++
+
+
kT
tB
kT
tB
k
tA
k
tA
t
m
DD
kTk
TkT
DD
k
T
ct
kTkk
TkTkTk
δδ
when DA,t+1
k+1
- DA,t+1
k
is given in (a6) and DB,t+1
T-k+1
- DB,t+1
T-k
in (a7). Taking into account
equations (a6) and (a7), we see from (a10) that ∆tA,t increases in the number of A’s old clients in so
far as k T-k. Therefore ∆tA
*
≥ ∆tA. In the considered setting, where VA,t+1 describes the present
net profits from a strategy in which A captures all the future generations VA,t+1 is increasing in the
number of the clients so that a given rise in tA,t+h increases VA,t+1((k+1)(1- 1/(T+1)), tA
I*
) more
than VA,t+1(k(1- 1/(T+1)),tA
I
). By the definition given above, DA,t+1
k+1
= VA,t+1((k+1)(1-
1/(T+1)), tA
I*
) and DA,t+1
k
= VA,t+1(k(1- 1/(T+1)),tA
I
). Because ∆tA
*
≥ ∆tA, this effect even
strengthens. From this follows that VA,t+1((k+1)(1-1/(T+1)),tA
*
) - VA,t+1(k(1- 1/(T+1)),tA)
VA,t+1((k+1)(1- 1/(T+1)),tA
I*
) - VA,t+1(k(1- 1/(T+1)), tA
I
)) = DA,t+1
k+1
- DA,t+1
k
in the considered
setting. This says that condition (12) is valid and that A’s decision to capture new clients in current

25
period is confirmed. The original result, according to which, in an equilibrium it is A who captures
the new clients and B who releases, is thus valid.
Finally we consider is it sufficient for bank A to set tA,t
o
so that tB,t(tA,t
o
) = tB,t
I
– η, so long as it is
bigger than B, to guarantee that B will never consider the capture. B’s prospects depend very much
on the behaviour in symmetry where A and B are about of the same size. It is remarkable that in the
symmetry (or near the symmetry) both would compete the prices offers down to indifference levels
tA,t
I
and tB,t
I
, if they would take it as given that each of them could credibly keep the rival indifferent
between the capture and release in further periods. In that case
VA,t+1((k+1)(1 - 1/(T+1)),tA*) - VA,t+1(k(1 - 1/(T+1)),tA) = DA,t+1
k+1
- DA,t+1
k
VB,t+1((T-k+1)(1 - 1/(T+1)),tB*) - VB,t+1((T-k)(1 - 1/(T+1)),tB) = DB,t+1
T-k+1
- DB,t+1
T-k
.
But, if the banks see that by bidding under the rival they can become bigger and seize on possibly
bigger profits, the pricing in symmetry would alter. Given tB,t, bank A has in symmetry an incentive
to bid lower until tB,t is so low that B gets at the most DB,t+1
T-k
, when A does not choose to bid lower
so that B captures the future generations. If both banks make low bids they suffer, however,
remarkable losses. Thus in the symmetry there is no stable equilibrium. The possibility of price war
does not increase the banks’ expected net present values in the symmetry. On the contrary, the
possibility that both banks would bid under indifference levels would make the concerning net
present value even smaller. Thus we regard that in symmetry (when T-k ≈ T/2), B’s net present
values is at the highest DB,t+1
T-k
.
In fact, in the symmetry both banks can make its rival indifferent between a capture and release, and
at least one of them can be made indifferent from now on. The bank, which lets its rival to capture
the new clients in the future, must in the symmetry be equally well-off as its rival. From this follows
logically that the value of neither bank can in the symmetry be above D-function.
Consider then a situation in which k T-k, and in which B captures new generations (during i
periods) until it is of equal size with A. A is assumed to set tA,t
o
so that tB,t(tA,t
o
) = tB,t
I
– η and η is
negligible small. Because B’s expected net present value after i periods is at the highest DB,t+1
G
,
where G = (T-k+1)(1-(1/(T+1))i
+(1-(1/(T+1))i-1
+ …+ (1-(1/(T+1)), B’s cumulated losses
from capturing new generations are, however, then bigger than the profits gained in the form
δi
DB,t+1+i
G
- DB,t+1
T-K
.

26
Proof of corollary 1.
From the proof of proposition 1 appears that when bank A captures new clients, bank A should set
tA,t so that tB,t is slightly below the threshold tB,t
I
expressed on the right hand side of (a3). This
conditions is
).(
)1(
1
)1(
1)(2
)1(
)(
1,
1
1,222
2
,
kT
tB
kT
tBt
m
BtB DD
kT
T
c
kT
kT
t
kT
kT
t −
+
+−
+ −
+−
+
−
+−
+−
+
+−
−
δ
Taking into account that A should set tA,t also subject to condition (1) and make thus sure that new
clients choose bank A in the current period, the above inequality can be written as
).(
)1(
))(1(
)1(
1)(2
)1(
)(
)(
2
1,
1
1,222
3
,
kT
tB
kT
tBt
mm
AtA DD
kTk
kTT
c
kTk
kT
t
kTk
kT
mh
k
Tk
t −
+
+−
+ −
+−
−+
−
+−
+−
+
+−
−
++
−
δ
Above
.
))
1
1
1(1)(1(
)
1
1
1)(1)(2(
)(
2
2
1,
1
1,
δ
+
−−+
+
−+−
−=− −
+
+−
+
T
T
T
kT
ctDD t
mkT
tB
kT
tB
(see, the proof of proposition 1). Inserting this into the above expression of tA,t and setting tm
= hm
+ m, gives the expression in corollary 1.
The proof of proposition 2
We consider a model where bank A who has k old clients can earn at each period at the most k(tm
-
ct). Because there are no network externalities, the new clients choose A, if tA,t tB,t. We show that
there cannot exist an equilibrium in which it pays for A to capture new generations and in which it is
not profitable for B to capture new generations. Suppose A has originally k generations. It is
profitable for A to capture new clients, if

27
(b1) ).(
)1(
1
)(
)1(
1,
1
1,,
k
tA
k
tAtt
m
AtA VV
k
cct
k
k
t +
+
+ −
+
−+−
+
δ
In (b1) VA,t+1
k+1
is A’s net present value in period t+1 when it had (k+1)(1- 1/(T+1)) old
generations in that period. Respectively in VA,t+1
k
A has k(1- 1/(T+1)) old generations. When it
does not pay for B to capture new generations, it is required
(b2) ).(
)1(
1
)(
)1(
)(
1,
1
1,,
kT
tB
kT
tBtt
m
BtB VV
kT
cct
kT
kT
t −
+
+−
+ −
+−
−+−
+−
−
δ
Inequalities (b1) and (b2) imply that it is profitable for A to capture new clients so long as
(b3) .0
)1(
)(
)1(
)(
)1)(1(
))(2( 1,
1
1,1,
1
1,

+
−
+
+−
−
−
+−+
−−
− +
+
+
−
+
+−
+
k
VV
kT
VV
kTk
ctTk
k
tA
k
tA
kT
tB
kT
tBt
m
δδ
The left hand side of (b3) is tB,t
I
- tB,t(tA,t
I
). Suppose that A captures all the new generations from now
on. It pays for A to set prices so that B will be almost indifferent between the capture and release. In
that case we can replace VB,t+1
T-k+1
for VB,t+1((T-k+1)(1 - 1/(T+1)),tB
I
*) = DB,t+1
T-k+1
, and VB,t+1
T-k
for VB,t+1((T-k)(1 - 1/(T+1)),tB
I
*) = DB,t+1
T-k
. Then VB,t+1
T-k+1
- VB,t+1
T-k
in (b3) is the same as
DB,t+1
T-k+1
- DB,t+1
T-k
which has an equation
(b4) .
)
1
1
1(1
)
1
1
1(
δ
+
−−
+
−
T
T
Suppose that also A is originally indifferent between capture and release. Then also VA,t+1
k+1
- VA,t+1
k
can be substituted for expression (b4). Using expression (b4), the left hand of inequality (b3)
becomes
(b5) ].
)
1
1
1(1
)
1
1
1(
1[
)1)(1(
))(2(
δ
+
−−
+
−
+
+−+
−−
−
T
T
kTk
ctTk t
m
Expression (b5) is negative which shows that A can not capture new generations by profit.
Because there are no network externalities tA,t
I
= tB,t(tA,t
I
) and tB,t
I
= tA,t(tB,t
I
). The expression (b5)
describes tB,t
I
– tA,t
I
on the condition that both banks are indifferent between the capture and the

28
release in the future. This expression is negative which refers to the fact that tB,t
I
tA,t
I
insofar as
both are indifferent in the future. Suppose that A decides however to capture new clients in the
current period. Then it should lower its prices below tA,t
I
to the level tA,t = tB,t
I
. If A also behaves in
this way consistently in future, its future prices tA,t+h
I
are replaced by prices tA,t+h (= tB,t+h
I
). This
means that VA,t+1
k+1
- VA,t+1
k
, which was originally VA,t+1((k+1)(1 - 1/(T+1)),tA
I
*) - VA,t+1(k(1 -
1/(T+1)),tA
I
) (= DA,t+1
k+1
- DA,t+1
k
) becomes VA,t+1((k+1)(1 - 1/(T+1)),tA*) - VA,t+1(k(1 -
1/(T+1)),tA) when tA
*I
= tA,t+1
I*
, tA,t+2
I*
, tA,t+3
I*
, ..., tA = tA,t+1
I
, tA,t+2
I
, tA,t+3
I
, .. , tB
*
= tB,t+1
I*
, tB,t+2
I*
,
tB,t+3
I*
, ..., and tB = tB,t+1
I
, tB,t+2
I
, tB,t+3
I
, … Because tB,t
I
– tA,t
I
in (b3) decreases in k (number of A’s
old generations), it is clear that also tB,t+h
I
– tA,t+h
I
decreases in the number of A’s old generations
when h ≥1. This means that tA
*
- tA
I*
tA - tA
I
. Because each tA,t+h
I*
is lowered more than each tA,t+h
I
(when h ≥1), the difference VA,t+1((k+1)(1 - 1/(T+1)),tA*) - VA,t+1(k(1 - 1/(T+1)),tA) becomes
smaller than VA,t+1((k+1)(1 - 1/(T+1)),tA
I
*) - VA,t+1(k(1 - 1/(T+1)),tA
I
) even when net present
values describe the situation in which A captures all the future generations. This only confirms that
inequality (b3) cannot be true when A captures new clients.
The proof of proposition 3.
The proof of corollary 1 follows the proof of proposition 1. Using equations (13) and (14) instead
of equations (3) and (7), the conditions (a3) and (a4) turn into form
(c1) ).(
)1(
1
)1(
1)(2
)1(
)(
1,
1
1,,
kT
tB
kT
tBt
m
BtB DD
kT
T
c
kT
kT
t
kT
kT
t −
+
+−
+ −
+−
+
−
+−
+−
+
+−
−
− δ
and
(c2)
t
T k
T k
h m
k
T k k
t
k k
T k k
c
k T
T k k
D D
B t
m
A
m
t
A t
k
A t
k
,
, ,
( )
( )
( )
( )( )
( )
( )( )
( )
( )( )
( ).

−
−
+ −
− +
+
+
− +
−
+
− +
−+
+
+
2
1
2 1
1
1
1
2
1
1
1δ
From (c1) and (c2) follows the requirement

29
( )
( )
( ) [
( )( )
( )
( )
] [
( )
( )
( )
( )( )
]
( )
( )( )
( )
( )
( ) ,, , , ,
2
1 1
2 1
1
2 1
1
1
1
1
1
0
2
1
1
1 1
1
1
k T
T k
h m
k
T k k
T k
T k
t
T k
T k
k k
T k k
c
k T
T k k
D D
T
T k
D D
m m
t
A t
k
A t
k
B t
T k
B t
T k
−
−
+ +
− +
−
−
− +
+
− +
− +
−
+
− +
+
+
− +
− −
+
− +
− +
+
+ +
− +
+
−
δ δ
in which tA
m
= tB
m
≡ tm
. Next we insert tm
= hm
+ m into the above inequality and manipulate the
terms after which it transforms into
(c3)
( )( ( ) )
( )( )( )
( )
( )
( )( )
( )
( )
( ) ., , , ,
2 2 2 1
1 1
1
1
1
1
01
1
1 1
1
1
k T k T k T
T k k T k
t c
k T
T k k
D D
T
T k
D D
m
t
A t
k
A t
k
B t
T k
B t
T k
− − + +
− + − +
−
+
+
− +
− −
+
− +
− +
+
+ +
− +
+
−
δ δ
The sum of the terms in the first row of (c3) is positive insofar as 2k-T 0. In the second row
DA,t+1
k+1
- DA,t+1
k
DB,t+1
T-k+1
- DB,t+1
T-k
, from which follows directly that the sum of the terms in
the second row is positive also so long as 2k-T 0. This proves that (c3) is valid when k T-k.
Following the proof of proposition 1, we should also prove that ∆tA
*
≥ ∆tA which guarantees that
condition (12) is valid, in so far as 2k-T 0. Comparing (c3) with (a5), it is clear, on the basis of
the proof of proposition 1, that this requirement is met.
The proof of proposition 4.
We show again that that there exists such an credible price offer hA,t
o
which makes the client to
choose bank A in period t subject to condition WA,t(k) - XA,t(k) 0 and XB,t(T-k) - WB,t(T-k) 0.
The first stage of the proof
Following the proof of proposition 1, we assume that the net present values in period t+1 are
DA,t+1
k+1
, DA,t+1
k
, DB,t+1
T-k+1
and DB,t+1
T-k
. Then it is profitable for bank A to capture new clients, if
(d1) ).(
)1)((
1
)(
)1)((
12
1,
1
1,,
k
tA
k
tAt
m
htA DD
kkT
T
ct
kkT
k
ch +
+
+ −
+−
+
−−
+−
+
− δ

30
Let us assume that bank A sets hA,t only slightly below the level in which the client is indifferent
between A and B. Taking this and condition (15) into account, (d1) can be expressed in terms of hB,t
by
(d2) h
k T
k
m
T k
k
c
k
k k
c
k k T T
k k
t
T
k k
D DB t h t
m
A t
k
A t
k
, , ,
( )
( )
( )
( )
( )
( ). −
−
+
−
+
+
+
+
− − −
+
−
+
+
−+
+
+
2 2 1
1
2 1
1
1
1
1
1
1δ
The condition under which it is unprofitable for B to capture the new clients, is now
(d3) ).(
)1(
)1(
)(
)1(
1)(2
1,
1
1,,
kT
tB
kT
tBt
m
htB DD
kTk
T
ct
kTk
kT
ch −
+
+−
+ −
+−
+
−−
+−
+−
− δ
Taking into account assumptions tm
= hm
+ m and hm
= ch and combining of conditions (d2) and
(d3), yields
(d4)
( )
( )( )
( )
( )
( )
( )
( )
( ) ., , , ,
2
1 1
1
1
1
1
01
1
1 1
1
1
k T
k k T k
t c
T
k k
D D
T
k T k
D D
m
t
A t
k
A t
k
B t
T k
B t
T k
−
+ − +
−
+
+
+
− −
+
− +
− +
+
+ +
− +
+
−
δ δ
The first row in (d4) is positive. Using expressions (a6) and (a7), the second row of (d4) can be
expressed in the form
(d5) ,
)
1
(1
)(
)1)(1(
2
2
δ
δ
+
−
−
++−
−
T
T
ct
kkTk
Tk t
m
which is clearly positive, if 2k-T 0.
The second stage
At the second stage we should prove also that ∆hA
*
≥ ∆hA when ∆hA
*
and ∆hA are similarly defined as
∆∆∆∆tA
*
and ∆∆∆∆tA above. This is sufficient for the condition
VA,t+1((k+1)(1 - 1/(T+1)),hA*) - VA,t+1(k(1 - 1/(T+1)),hA)
VA,t+1((k+1)(1 - 1/(T+1)),hA
I
*) - VA,t+1(k(1 - 1/(T+1)),hA
I
)
to be fulfilled. To show that ∆hA,t (= hA,t(hB,t
I
) - hA,t
I
) is increasing in the number of A’s old clients,
we form ∆hA,t by multiplying expression (d4) by k/(T-k). This yields

31
.0)(
)1)((
1
)(
)1)((
)1(
)(
)1)(1)((
)2(
1,
1
1,1,
1
1, −
+−−
+
−−
+−
+
+
−
+−+−
−
−
+
+−
++
+
+
kT
tB
kT
tB
k
tA
k
tA
t
m
DD
kTkT
T
DD
kkT
T
ct
kTkkT
Tk
δδ
The first row clearly increases in k when expressions DA,t+1
k+1
- DA,t+1
k
and DB,t+1
T-k+1
- DB,t+1
T-k
have equations (a6) and (a7). In the above equation the second row transforms into
,
)
1
(1
)(
]1)
1
1(2[
)1)(1)((
)2(
2
δ
δ
+
−
−
−−
++−−
−
T
T
ct
TkkTkT
Tk t
m
which clearly increases in k. The rest of the proof relies on the second stage of the proof of
proposition 1.

Switching Costs and Network Externalities in the Production of Payment Services

More Related Content

Viewers also liked (14)

Similar to Switching Costs and Network Externalities in the Production of Payment Services (20)

More from Palkansaajien tutkimuslaitos (20)

Recently uploaded (20)

Switching Costs and Network Externalities in the Production of Payment Services