the links between securities settlement systems: an olihopoly theoretic approach
TRANSCRIPT
International Review of Financial Analysis
13 (2004) 585–600
The links between securities settlement
systems: An olihopoly theoretic approach
Karlo Kauko
Bank of Finland, (Suomen Pankki) PO Box 160, 00101 Helsinki, Finland
Abstract
A duopoly model of the securities settlement industry is presented. Because pooling payments
can help in using liquidity efficiently, issuers prefer systems where a large number of securities are
issued. If the central securities depositories (CSDs) establish a mutual link that enables investors to
make transactions with foreign securities, cost savings can be achieved. However, these links may
have unexpected effects on CSDs’ pricing, and the issuers’ share of the fee burden can increase
substantially. It is not advisable to ban additional fees for using the link, as the CSDs might simply
increase the fee for domestic transactions.
D 2004 Published by Elsevier Inc.
JEL classification: L13; G20
Keywords: Securities settlement systems; Oligopolies
1. Introduction
Modern securities settlement systems (SSSs) are often highly centralised. The core of
the system is a central institution that runs the system, for example, the national central
securities depository (CSD). A relatively small number of institutions participate directly
in the process. These participants (members) are financial institutions such as banks and
investment firms (IFs). Government treasuries and central banks may also be members. A
typical investor uses services provided by a member of the settlement system. There are
many nonintegrated national systems, and most of the securities traded are issued by
domestic entities. Although remote access from abroad to trading has become common-
place, remote participation in SSSs is still exceptional. The report of Giovannini et al.
1057-5219/$ - see front matter D 2004 Published by Elsevier Inc.
doi:10.1016/j.irfa.2004.02.014
E-mail address: [email protected] (K. Kauko).
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600586
(2001) discusses the practical problems related to cross-border clearing and settlement in
the EU.
It is increasingly appreciated that the European securities market infrastructure is going
through a period of profound change. Several mergers have taken place between securities
market institutions. Although much of the consolidation has taken place at the national
level, some international mergers have occurred between EU member countries, and more
are expected to occur over time, bringing historically exceptional change to the increas-
ingly integrated securities market infrastructure. Links between CSDs are a relatively new
arrangement, aimed at improving the possibilities to settle cross-border transactions. These
links are established by bilateral agreements between SSSs. Customers of an SSS can hold
securities in another system via the domestic service provider. The domestic service
provider opens an omnibus account in a foreign book-entry system, and the holdings of its
customers are pooled in this account. The domestic CSD keeps individual records on
investors’ holdings.
This paper analyses the competition between two SSSs in a fragmented market,
where the focus is on pricing and interlinking. The systems operate in a platform
industry, as defined by Rochet and Tirole (2001), and must attract both issuers and
investors. Although there are earlier analyses on network effects and oligopolistic
competition between stock exchanges (Di Noia, 1999; Shy & Tarkka, 2001), these
focus on trading activities rather than on SSSs. The basic model is presented in Sections
2 and 3. Competition for issuers is fierce because trades must be settled in the CSD,
chosen by the issuer. The whole fee burden is borne by the investors. In Section 4 it is
assumed that the two CSDs are linked. Significant cost savings can be achieved in the
use of liquidity and in operating costs. Competition for issuers is relaxed, and their
relative fee burden increases. In Section 5, we demonstrate that it is not advisable to ban
additional fees for using the link because this may simply lead to higher prices for
domestic services. The main conclusions and a few suggestions for further research are
presented in the final section.
2. The basic model
A large number of bond issuers, each of whom issues I bonds in the primary market, are
numbered 0,1,2,. . .,M. The bonds are in book-entry form. The issuers maximise the
difference between the value of bonds in the primary market and the cost of registering and
issuing them. The bonds are sold to investors at a price equal to the difference between a
constant and the expected value of secondary market settlement costs paid by purchasers
in the primary market. The issuer pays all fees related to issuance. Investors live in two
countries, denoted 1 and 2. In each country, they are divided into two groups, A and B.
These groups might be households and institutional investors.
There is a profit maximising CSD in each country. Each CSD operates as a platform
industry because it needs two types of customers that interact—issuers and investors. The
CSD is the registrar of the book-entry system, and it runs the SSS. Running the settlement
process generates costs, whereas registering new issues does not. The unit cost—per
transaction, sale, or purchase—of the centre for country j is denoted zj.
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600 587
Bonds are distributed in the primary market so that the proportion of issuer i’s investors
living in Country 1 is (M�i)/M. Each issuer knows beforehand where its investors live. All
the original buyers of any issue belong to the same investor group. Half of the issues are
subscribed by A investors and half by B investors. Issue number is not related to investor
group. In most cases, not all investors live in the same country, but the two groups exist in
both countries. There is also a secondary market for bonds in which trades are settled in
systems operated by the CSDs. Investors must use IFs to settle transactions in the CSDs. In
each country, there are two IFs competing for the A investors and two competing for the B
investors. IFs cannot become members of the foreign SSS. Investors must use a foreign IF
when they trade in foreign securities. Using a foreign IF is more complicated than using a
domestic one and entails an extra cost h. The IFs achieve turnover via pricing of their
services to investors. The price charged by an IF (v) is the same for a sale or a purchase.
The IFs have simple cost functions. For any IF, the cost of processing a transaction is y,
which is an exogenous constant. One trade, consisting of a sale and corresponding
purchase, is now defined as two transactions. The IFs Bertrand compete with identical
services.
Only two of the randomly determined (equally likely) bond issues are traded. The two
bonds are originally held by the investors of each group. The two issuers’ bondholders
exchange securities. Each investor is simultaneously a seller and a buyer. Price formation
and trading are ignored. The price of every security in the secondary market is the same
and does not depend on settlement costs. High settlement costs reduce both supply and
demand, and the total impact on price is consequently unclear (see Barclay, Kandel, &
Marx, 1998 for empirical analysis). Because each investor is simultaneously a seller and a
buyer, his/her incoming and outgoing monetary payments sum to zero, except for the IF
fees.
Monetary payments between the SSSs and IFs are based on netting.1 The IFs prefer
situations in which they intermediate both sales and purchases and need not make gross
payments to the CSDs. If gross payments are made, the IFs may need large and costly
cash balances. An IF might be unable to acquire liquidity to cover its net payment
obligations, which could result in sanctions. Clearing parties with net obligations may
have to post some assets as collateral, forcing the clearing parties to hold larger
portfolios (e.g., of government securities) than they would otherwise prefer. Moreover,
by pledging an important part of its assets, a clearing party weakens its own
creditworthiness in noncollateralised borrowing because collateral assets cannot be freely
liquidated to cover the debts in case of bankruptcy. This probably affects the cost of
external funding.
The expected total cost and disutility of being a clearing party with a gross payment
obligation is a per security purchase to be intermediated. The analysis is restricted to
cases where 2h>a. It will be seen in due course that there is no meaningful equilibrium
1 This is a satisfactory approximation for many gross settlement systems. In a typical gross settlement
system, no settlement counterparty is obliged to make all the outgoing payments during a day before receiving
any incoming payments. Instead, automatic chaining procedures applied in many systems guarantee that
incoming payments can be used efficiently as a source of funding for outgoing payments.
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600588
with two participating CSDs if this condition is not satisfied. Events happen in the
following order.
1. Profit-maximising CSDs set their prices, including those to be paid by issuers ( pi for
the CSD based in the country i), and a unit price per bond to be paid by clearing parties
(bi),
2. Issuers choose their CSDs, and bonds are issued,
3. Profit maximising IFs simultaneously set their fees (vs),
4. The two bond issues to be traded are determined,
5. Investors choose IFs and try to minimise settlement costs, and
6. The trades are settled, net payments are made, and securities are transferred.
All agents are assumed to be risk neutral and to be able to immediately observe every
decision and random outcome. Steps 4, 5, and 6 could describe a day. These three steps are
repeated hundreds of times every year. The CSDs and IFs make no decisions at these
stages. Expected values analysed in the following sections are simply multiplied by the
number of repetitions and have no fundamental impact on their optimal decisions.
3. Solving the model
3.1. Competition between IFs
Because of Bertrand competition with identical services, the fee charged by an IF (vi)
must equal the marginal cost, which equals the sum of the unit fee charged by the CSD
(bj), the marginal processing cost for the IF ( y), and the expected value of the per-
transaction cost of gross payments (Fi).
vi ¼ Fi þ bj þ y ð1Þ
Gross payment costs are the only complicated component of the marginal cost. If two
bond issues, i and j, are traded, the cost of gross payments (a) materialises as a part of
Country 1’s IF cost in two possible cases. (Terms in parentheses are respective
probabilities.) Bond i is registered in the domestic country (C1) and bond j in the foreign
country (C2). The IF represents half of the buyers. Hence, the expected value of the total
gross payment cost is
aC1C2
2þ C2C1
2
� �¼ aC1C2 ð2Þ
A part of the fee corresponds to this expected value. According to Eq. (1), the expected
value of this cost factor must equal the expected value of the corresponding fee revenue.
The expected value of the number of transactions is the sum of two components. First, the
IF collects fees if one of the two securities to be traded is registered in its home country.
Each customer must pay the fee at once to the IF. The probability of this equals 2C1C2.
Second, if the IF charges fees if both issues to be traded are registered domestically. The
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600 589
probability that both bond issues to be settled are registered in Country 1 is C12. In this
case, each customer has to pay the same fee twice because he presents two settlement
orders to the same IF.
The expected value of the part of the fee revenue that corresponds to gross payment
costs is thus:
F1ð2C2C1 þ 2C21Þ ð3Þ
The expected value of fee revenue related to gross payment costs (Eq. (3)) must equal
the expected cost of making gross payments (Eq. (2))
aCjCi ¼ Fð2CjCi þ 2C21Þ
ZFi ¼aCj
2ð4Þ
This result is quite intuitive. If the market share of domestic CSD 1 is marginal (C1c0),
it is almost certain that if a domestically registered bond is traded, the other bond to be
traded is registered abroad and netting cannot be used. The IF has a 50% chance of
representing a buyer. The larger the domestic CSD’s market share, the more likely it is that
the IFs will benefit from netting. Therefore, Eq. (1) can be rewritten as
vi ¼aCj
2þ bi þ y ð5Þ
3.2. Pricing decisions by the CSDs
The CSDs maximise the expected value of their profits. Their decision variables are
the price paid by issuers ( pi) and the price per transaction paid by settlement counter-
parties (bi). Issuers minimise the sum of fees paid by themselves and the settlement costs
to their bondholders. Investors’ costs are relevant because they affect security prices in
the primary market. If 100j% of the bondholders live in Country 2, the cost of using
CSD 1 is p1+I(2/M){v1+jh}, where 2/M is the probability that the issuer will be one of
the two issuers whose securities are traded and v1 the price charged by all the IFs of
Country 1. The cost of using CSD 2 is p2+I(2/M){v2+(1�j)h}. Because v1=aC2/2+b1+y
and v2=aC1/2+b2+y, the condition for using CSD 1 can be written as
p1 þ2I
M
� �aC1
2þ b2 þ yþ jh
� �< p2 þ
2I
M
� �að1� C1Þ
2þ b2 þ yþ ð1� jÞh
� �
Zj <f2hI � aI � 2Ib1 þ 2Ib2 þ 2aC1I �Mp1 þMp2g
f4hIg ð6Þ
This condition implies a network externality. The market share of CSD 1 (C1) has a
direct impact on its competitiveness. If most companies are registered at CSD 1, it
becomes more likely that IFs operating in Country 1 need not make gross payments in the
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600590
settlement system. This, in turn, has a direct impact on the fee paid by the investors (v1).
This network externality implies that market forces are, at least, weakly inclined to
integrate securities settlement by centralising the process in one centre.
In equilibrium, the relative share of issuers preferring CSD 1 must equal the market
share C1. Because issuers’ bondholder structures are evenly distributed along the
continuum from j=0 to 100, it follows that Eq. (6) implies
C1 ¼ 1=2þ2b2 � 2b1 þ M
I
� �ðp2 � p1Þ
� �½4h� 2a ð7Þ
The CSD has two decision variables, b and p. As determinants of market shares, these
two variables are perfect substitutes; any combination of p and b that satisfies the
condition 2Ib1+Mp1=D (D being any constant) leads to the same market share. The
CSD can decide these prices in the following way. CSD 1 decides first on a suitable value
of D, then, on the value of p1. These decisions determine the value of b1. Using this
notation, the expected profit of CSD 1 can be written as
P1¼C1 Mp1þ4IðD�Mp1Þ
ð2IÞ
� �� z1
¼ C1½2D�Mp1 � z1Z
BP1
Bp1¼ �C1M < 0
It follows that the optimal price paid by issuers is 0, so the entire fee burden is borne by
IFs and the investors [ p1=p2=0, b1=D/(2I)]. When choosing a CSD, an issuer pays no
attention to the fees paid by future bondholders. However, these fees are a valuable source
of revenue for CSDs.
There are two issues of I bonds to be traded. For each of them, the probability of being
registered in CSD 1 equals its market share (C1), implying that the expected value of the
number of processed securities equals 2IC1. Each traded issue registered in CSD 1 yields
two service fees (b1) per processed security, one for selling and another for buying. The
cost per processed security equals 2z1. Therefore, CSD i’s expected profit is
Pi ¼ 2I ½bi � zi½�a þ 2bj � 2bi þ 2h
½2h� a ð8Þ
The first-order optimisation condition of a CSD is
BPi
Bbi¼ �2I
½a þ 4bi � 2ðbj þ hþ zi½2h� a ¼ 0 ð9Þ
This condition is satisfied for both CSDs iff (i= 1,2)
bi ¼ð�3a þ 6hþ 4zi þ 2zjÞ
6ð10Þ
The second order condition is B2P1/Bb12=�8I/(2h�a). This is negative, and the extreme
values are maxima iff a<2h. If this condition were not satisfied, Eq. (7) would imply that
BCi/Bpi>0 and BCi/Bbi>0, which is not a meaningful result. The result in Eq. (10) is easy
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600 591
to understand intuitively. First, being cost inefficient means a high optimal price. A cost-
inefficient competitor improves the possibility to charge high prices. If the cost of making
cross-border transactions (h) is high, the CSD’s pricing decisions are made in a less
competitive environment because the issuer’s decisions are determined by bondholders’
home countries rather than the CSD’s prices.
The most interesting finding might be the impact of netting possibilities on price
competition. If almost nothing can be gained by netting (ac0), price competition is
moderate. If the potential cost savings are substantial aH0, price competition is fierce. By
attracting an issuer with aggressive pricing, the CSD can easily attract another issuer
because its bondholders value the possibility to benefit from netting should they exchange
bond issues with the bondholders of the first issuer. This outcome is somewhat analogous
with that of Katz and Shapiro (1986), but it differs in terms of the equilibrium price
externality from the analysis of Economides and Siow (1988).
These pricing decisions imply the following market shares:
Ci ¼½�3a þ 6h� 2zi þ 2zj
½12h� 6a ð11Þ
The profits of the country i CSD can be calculated by combining the results Eqs. (10)
and (11).
Pi ¼½�3a þ 6h� 2zi þ 2zj2
½9ð2h� aÞ ð12Þ
Because the cost of cross-border transactions is the CSD’s only source of market power,
it is not surprising that high costs increase the likelihood that both CSDs will earn positive
profits. In contrast to this, a high cost of gross payments (a) intensifies price competition,
making it increasingly difficult, especially for the smaller CSD, to charge prices that
exceed costs. The empirical prediction of this model is that the issuers are loss leaders, as
defined by Rochet and Tirole (2001), whereas investors are the profit-making segment for
CSDs. It is surprisingly difficult to find suitable data on different sources of CSD revenue,
but casual observation indicates that revenue from secondary market trading has
traditionally exceeded revenue from issuance.
4. Linking the securities centres
Every IF can participate in the SSS of the neighbouring country via the domestic CSD
and the link between CSDs. The domestic CSD has an omnibus account with the foreign
CSD, in which the customers’ securities are held. The operating costs of a transaction are
borne by the CSD of the investor’s home country. From the viewpoint of registration
country processes, there is only one large net transaction on a particular customer account,
whereas the CSD based in the investors’ home country must administer thousands of small
transactions on a large number of accounts. CSD i sets three different prices: first, a price
paid by each issuer ( piz0), second, a basic unit fee paid by IFs for each settlement
transaction involving any security (biz0), and third, an additional fee paid by IFs for each
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600592
transaction with foreign securities (riz0). When trading in foreign securities, the investor
can use either a domestic IF connected to the domestic CSD and the link, or he/she can
participate via a foreign IF. If the investor uses the foreign IF, there is an extra cost, h, as in
the basic model. Events happen in the following order.
1. CSDs set prices for issuers ( ps) and for IFs (bs and rs),
2. IFs set their prices,
3. Issuers choose CSDs, and
4. Investors choose IFs. Trades are agreed, cleared, and settled.
Establishing the link is pointless unless the CSDs know that investors are going to use
it. The following analysis can be restricted to cases where the link is in use. As we will see,
however, investors’ possibilities to use foreign IFs are not completely irrelevant when
CSDs set their prices.
4.1. Competition between IFs
If a cross-border transaction is processed via the link, every investor deals through one
IF. Because no investor has to make a gross payment, the monetary payment between the
IF and the investor is always zero. Consequently, even the monetary payments between the
IF and the CSD can always be netted. Interestingly, even the net payment between the two
CSDs will always be zero because of complete netting at all the levels. If the CSDs offer
no cross-border payment facilities, the IFs in different countries will use an existing
international payment system in which payments can be netted. Hence, the cost component
a vanishes. This reduced need for liquidity argues in favour of consolidation and linking in
the securities settlement industry. Eq. (2) implies that the prices paid by customers in
country i are
vid ¼ bi þ y and vif ¼ bi þ ri þ y ð13Þ
where vid is the price paid by country i investors for domestic transactions and vif is the
price paid for cross-border transactions.
4.2. Market shares
Issuer i, with 100(i/M)% of bondholders in Country 2, prefers to use CSD 1 iff
p1 þ b1I 1� i
M
� �þ ðb2 þ r2ÞI
i
M< p2 þ ðb1 þ r1ÞI 1� i
M
� �þ I
i
Mb2
Zi < MðIr1 � p1 þ p2ÞðIr1 þ Ir2Þ
The CSD’s market share does not enter this expression. The network externality
that prevailed in the previous version of the model no longer exists because all
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600 593
payments can be netted. The formulas for market shares can be derived easily from
the above condition
Ci ¼ðIri � pi þ pjÞ½Iðri þ rjÞ
ð14Þ
Interestingly, the basic transaction fee (bi) is irrelevant for market share. Possibly even
more surprising is that high extra fees charged for cross-border transactions raise the
market share because CSD 1 can make the rival CSD 2 unattractive for Country 1
customers.
BC1
Br1¼ ½Ir2 � p2 þ p1
½Iðr1 þ r2Þ2> 0
4.3. CSDs’ profits
The relative share of securities originally issued in a CSD has no direct impact on the
number of secondary market transactions processed by it. Investors’ home countries
determine where the securities are settled, irrespective of the original country of issuance.
CSD 1 can earn revenue in three different ways. First, when securities are issued, the CSD
collects fees related to securities registration and primary market transactions. This
revenue equals the fee revenue from each issue ( p) times the number of issuers (M)
times the market share of the CSD (C).
p1MC1 ð15Þ
Second, it earns revenue by selling services to buyers. The number of securities to be
bought equals 2I. In expected value terms, half of the buyers live in the home country of
the CSD. A certain percentage of securities is registered abroad, implying that there is an
extra fee revenue worth I(1–C1)r1. Therefore, the total net revenue from buyers equals
Iðb1 � z1Þ þ Ið1� C1Þr1 ð16Þ
Calculating the fee revenue collected from investors that sell securities is more
complicated. The country of registration aligns with the original bondholders’ home
countries. The expected value of net revenue earned by providing services to sellers of
domestically registered security i equals
2
M
� �Iðb1 � z1Þ 1� i
M
� �
Here, (2/M) is the likelihood that the issue is traded, I the number of securities per issue,
(b1�z1) the net revenue per transaction, and (1– i/M) the relative share of bondholders
living in Country 1 and thus using the services of CSD 1.
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600594
As regards securities registered in the foreign country, the net revenue for selling is
determined in a similar way, the difference being the extra fee, r1
2
M
� �Iðb1 þ r1 � z1Þ 1� i
M
� �
The number of domestically registered bond issues is MC1 and the total number of
issuers is M. It follows that the net income from providing services to sellers is
Z MC1
0
2
M
� �Iðb1 � z1Þ 1� i
M
� �diþ
Z M
MC1
2
M
� �Iðb1 þ r1 � z1Þ 1� i
M
� �di ð17Þ
The profit can be calculated as the sum of these three components (Eqs. (15), (16), and
(17)).
Pi ¼ 2biI þ CiMpi þ 2Iri � 3IriCi þ C2i Iri � 2Izi ð18Þ
4.4. Prices and profits
When Eq. (14) is substituted for market shares, optimal prices in the primary market are
determined by the conditions
BP1
Bp1¼ 0
BP2
Bp2¼ 0
There is one combination of ps that satisfies both conditions, namely,
pi ¼ If�8rirj þM 2ðr2j þ 3rirj þ 2r2i Þ þMðr2j þ 5rirj þ 2r2i Þg
½Mð3M � 2Þðri þ rjÞð19Þ
The second-order condition (B2P1/Bp12=�2{(M�1)r1+Mr2}/{I(r1+r2)
2}<0) is satisfied.
Because MH0, both p1 and p2> 0 whenever r1>0 or r2>0, or both.
Differentiation of Eq. (18) yields
BP1
Bb1¼ 2I > 0 ð20Þ
It will always be optimal to charge a higher fee if pricing is not constrained. The upper
limit for this fee is determined by the condition that the total fee paid by cross-border
investors cannot exceed the cost of using foreign IFs. It follows that if a CSD wants to
intermediate settlement orders through the link, the highest feasible price satisfies the
condition
bi þ ri ¼ bj þ hZ bi ¼ bj þ h� ri ð21Þ
Hence, both the fee paid by issuers ( pi) and the basic fee paid by investors (bi)
can be expressed as functions of ri, bj, and h. By substituting Eq. (21) for b1,
substituting the optimal price presented in Eq. (19) for p1, substituting Eq. (14) for
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600 595
market shares, and taking into account the assumption that MH0, differentiation of
Eq. (18) yields
LimM!l
BP1
Br1
¼ Lim
M!lM
f3p2r2 þ 2Ir1ðr1 þ 2r2Þg½9ðr1 þ r2Þ2
ð22Þ
Whenever r1>0, r2>0, or both, this equals +l. It follows that irrespective of rival
prices, the additional fee for cross-border transactions (r1) is the highest fee that still
satisfies Eq. (21), and the optimal combination of b1 and r1 is
r1 ¼ hþ b2; b1 ¼ 0 ð23Þ
Analogously r2=h+b1 and b2=0. It follows that (r1=r2=h). This outcome does not
necessarily depend on the assumption of zero price elasticity of demand.
The pricing decision of the CSD is constrained by the fact that domestic customers
will use foreign IFs whenever the additional fee (r) is higher than the investors’ cost of
using foreign IFs (h). If this condition is binding, it holds irrespective of whether there is
a higher but finite optimal monopoly price. In real life, a CSD’s fees for cross-border
transactions are higher than its fees for domestic transactions (Lannoo & Levin, 2001).
Interestingly, the link does not promote the integration of securities markets. In the
absence of the link, the cost differential between domestic and cross-border transactions
is an exogenous constant h, at least, if both CSDs are equally cost efficient. Now, in the
presence of the link, the difference still equals h, although it consists of CSDs fees. This
result has a clear policy implication. Even when the link has been established,
governments could try to promote the integration of financial markets by making it
easier and cheaper for investors to use services offered by foreign IFs. This would limit
CSD’s possibilities to exploit the monopoly position. Removing the causes of a high
value of h would enhance integration more efficiently than pressing CSDs to establish
mutual links.
When r1=r2=h and b1=b2=0, Eq. (19) implies that the fee paid by issuers equals
pi ¼Ið2þMÞh
Mð24Þ
Applying Eqs. (14), (18), (23), and (24) yields
Pi ¼I ½hð7þ 2MÞ � 8zi
4ð25Þ
This model predicts that in the presence of the link, CSDs will charge issuers rather
than investors will. In the basic model, issuers can make CSDs compete, whereas investors
cannot. If there is a link between the two CSDs, issuers no longer determine where the
secondary market investors have to operate, which makes it less essential to attract them.
The empirical prediction is clear, but it is surprisingly difficult to find reliable data on
different sources of CSD revenue. Securities issuance is even more cyclical than secondary
market trading, which would make meaningful comparisons difficult even if the data were
readily available. This outcome is analogous to a result of Laffont, Rey, and Tirole (1998),
who concluded that if telecommunications operators can charge additional fees for calls
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600596
between networks, there can be positive externalities between customers of the same
operator that might intensify price competition. Because telecommunications networks are
not two-sided markets, however, the fee burden cannot be shifted to another customer
group.
4.5. Will the link be established?
Profit-maximising CSDs will establish the link if they can earn higher profits with it
than without it. This situation can be analysed by comparing the expressions for profits
under the two alternative arrangements (Eqs. (12) and (15)). CSD i would prefer to
establish the link if
I ½�3a þ 6h� 2zi þ 2zj2
½9ð2h� aÞ <I ½hð7þ 2MÞ � 8zi
4ð26Þ
If z1=z2=z, either both or neither of the CSDs will prefer to establish the link. Thus, Eq.
(26) can be rewritten as
M > 1=2þ 4z� 2ah
ð27Þ
If there are no operating costs (z=0), the CSDs always prefer to establish the link.
Nonlinked CSDs can pass the cost z to investors, whereas linked CSDs cannot.
Whether coincidental or not, ICT has improved SSS’s efficiency, while at the same
time, links between CSDs have become more commonplace. A large number of
issuers (M) makes it more certain that the CSDs will want to establish the link
because issuers become the profit-making sector, and if there is a large number of
them, opening the link is a lucrative strategy. High costs of gross payments (a)encourage CSDs to establish the link; these costs intensify price competition in the
absence of the link.
From the viewpoint of social welfare, the link is particularly useful if using foreign IFs
is expensive for investors (hH0). By opening the link, the two CSDs can transform
investors’ costs of making cross-border transactions into fee revenue. In this sense, their
incentives are ‘‘correct;’’ they want to establish the link when the costs that can be avoided
by using it are high.
5. Regulation of cross-border fees
The EU has issued a regulation saying that fees for cross-border retail payments are not
to differ from those for domestic transactions. This section presents an analysis of the
possible impact of a comparable regulation on SSSs. CSDs are obliged to form a link and
not to charge additional fees for cross-border transactions (r=0). Investors trading in
foreign securities can use domestic IFs and the link, or they can use foreign IFs. As in the
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600 597
previous sections, they incur a fixed cost (h) if they use foreign IFs. If issuers are
indifferent between the two CSDs, they randomise between them. If there is no Nash
equilibrium in the Bertrand game between CSDs, they end up in an undercut-proof
equilibrium, as defined by Shy (2002).
5.1. The undercut-proof equilibrium
The situation presented in this section is highly analogous to the one presented by Shy
(2002). Shy assumes two Bertrand competing firms. Every consumer has an established
relationship with one of the two companies. Buying from the competitor would cause
switching costs. Each consumer buys one unit of good. It turns out that there is no Nash
equilibrium in pure strategies. The optimal price for each firm is that charged by the
competitor plus the switching cost. It is obviously impossible to find any equilibrium in
which both companies charge more than the competitor does. A firm could also undercut
by pricing low enough to get all the customers. The firm whose price is undercut would
find it profitable to charge a fee that exceeds the price of the rival by the transaction cost.
Firms might eventually end up in an undercut-proof equilibrium, both charging the
highest price that the rival cannot profitably undercut, if the alternative is to charge the
price that prevails in the undercut-proof equilibrium. This equilibrium is not a Nash
equilibrium.
When r1=r2=r, Eq. (19) reduces to pi=I(2+M)r/M. When no additional fees for cross-
border services are allowed (r=0), the optimal fee to be paid by issuers equals zero. The
CSDs end up in Bertrand competition, with identical services and zero marginal costs, and
the price ( p) converges to zero. Therefore, issuers randomise between the two CSDs, and
there will be no correlation between the country of registration of a security and the
investors’ home country. Because issuers pay no fees, and because there are no fees for
cross-border transactions, both CSDs are completely dependent on the basic fees paid by
domestic investors. The expression for profits reduces to
Pi ¼ 2Iðbi � ziÞ ð28Þ
According to this expression, it is always profitable to increase the price (dPi/dbi>0).
However, investors have the option to use foreign IFs. The highest price the CSD i can
charge and still get all the transactions of domestic customers is
bi ¼ bj þ h� e ð29Þ
where e is an arbitrarily small positive constant. The profit would equal 2I(bj+h�zi). All
payments can be netted, there are no costs related to gross payments, and the cost
component a vanishes.
A CSD can also capture all the customers by undercutting. The highest price the CSD
can charge and get all the customers is
bi ¼ bj � h� e ð30Þ
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600598
With this price, CSD i would also get all the issuers because the undercutting CSD
would be a cheaper alternative for all the investors. The expected value of its profit is
4Iðbj � h� e � ziÞ ð31Þ
Undercutting, that is, charging price in Eq. (30) instead of the price in Eq. (29), is
profitable for CSD i iff
4Iðbj � h� e � ziÞ > 2Iðbj þ h� ziÞZbj > 3hþ 2zi þ e ð32Þ
With sufficiently high rival prices (bj), Eq. (32) implies that it becomes profitable to
undercut. When firms decide the basic fees (b’s), the situation is analogous to Shy’s model,
and there is no Nash equilibrium in pure strategies. If a customer living in Country 2 deals
with foreign securities, the cost of a transaction via a domestic IF and domestic CSD 2
equals b2+y. If the customer uses foreign IFs, the cost equals b1+y+h. If CSD 1 undercuts,
the price (b1V) will satisfy the condition b2+y�e=b1V+y+hZ b1V=b2�h�e, where b1V is the
price that CSD 1 uses for undercutting, if it decides to undercut.
The highest price that can be used for profitable undercutting (b1V) satisfies the condition
2Iðb1 � z1Þ þ e ¼ 4Iðb1V� z1Þ and; analogously; 2Iðb2 � z2Þ þ e ¼ 4Iðb2V� z2Þ
Combining these conditions with Eq. (32) and letting e=0 yields
b1 ¼ð6hþ z1 þ 2z2Þ
3b2 ¼
ð6hþ z2 þ 2z1Þ3
b1V¼ð3hþ z1 þ 2z2Þ
3b2V¼
ð3hþ z2 þ 2z1Þ3
If there is no link between the CSDs, the cost of cross-border transactions for investors
living in Country 2 is determined by Eqs. (5), (7), and (10), and the extra cost h:
hþ v1 ¼ aC2=2þ ð�3a þ 6hþ 4z1 þ 2z2Þ=6þ yþ h
¼ �ð1� C2Þa=2þ f6hþ z2 þ 2z1g=3
This is less than the transaction fee charged by CSD 2 (b2) in the undercut-proof
equilibrium. Hence, in the undercut-proof equilibrium, the fee for transactions is higher
than the total cost of cross-border transactions that prevails without a link. Thus, the
regulation does not make it cheaper to trade in foreign securities. Instead, it becomes more
expensive to process transactions with domestic bonds.
In the basic model the two CSDs had to compete for issuers by being attractive
locations for investors. In Section 4, the CSDs competed for issuers even more directly.
Now, the country of registration is no longer relevant to the costs for investors in the
secondary market. Hence, the main source of competitive pressure has been largely
eliminated and it is not surprising that the average price for settlement services increases.
Somewhat similar results have been presented in the earlier literature. Banning price
differentials between domestic and export prices may enhance tacit collusion between
oligopolistic firms (see Martin, 2001, pp. 203–204). If competing telecommunication
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600 599
operators are not allowed to charge different prices for calls between networks, they might
increase the price for all calls (Laffont et al., 1998).
6. Conclusions
This paper has presented an oligopoly model of the securities settlement industry. The
main focus has been on competition between two national CSDs and on the impact of a
securities link on this competition. It was assumed that the need for securities settlement
services is exogenous. Each CSD runs both a book-entry register and a settlement system.
Secondary market transactions must be settled in the system in which the security was
issued. It follows that CSDs compete fiercely for issuers in the primary market because
issuers’ choices determine market shares. Consequently, CSDs collect all the revenue from
IFs and investors, while issuers get services for nothing.
Section 4 analysed the situation where the two CSDs are linked. Investors can settle
transactions with foreign securities via either domestic IFs and the domestic CSD, or they
can use services offered by foreign IFs. Transactions do not have to be settled in the CSD
chosen by the issuer, and it is no longer essential for CSDs to attract as many issuers as
possible. Consequently, issuers become the profit-making segment, whereas investors can
get domestic services free of charge and cross-border services at reasonable prices. In
Section 5, we analysed what might happen if the CSDs of the two countries are obliged to
establish the link and not to charge fees beyond the basic fee for domestic transactions for
cross-border transactions through it. It may be unclear how the CSDs would price
secondary market settlement orders, but they might simply increase the price for all the
transactions to a level that is higher than the costs of cross-border transactions in the
absence of a link.
The paper has one major policy implication. If policy makers want to enhance the
integration of financial markets, they should focus on eliminating the fundamental causes
of obstacles to cross-border settlement. Encouraging the private sector to develop new
ways to bypass barriers that remain basically unchanged may be much less useful. It is
always possible that access to a detour is controlled by an institution with monopoly
power. The institution could easily benefit from its monopoly position and capture all the
benefits from the detour.
In real life, many securities market infrastructure institutions are joint-venture under-
takings of financial institutions. If the two CSDs maximise the sum of their own profits
and the profits of the shareholder IFs, it is possible that basically nothing will change. IFs’
pure profits in their own operations are zero in any case, and it would be in their common
interest to instruct the CSD to maximise its own profits (see Park & Ahn, 1999 for a
detailed analysis on jointly owned upstream suppliers). Future work could profitably
continue along the following lines. First, by taking the price elasticity of demand into
account, we could analyse consumer surplus and issuer profits. Price elasticities could be
highly important determinants of platform industries’ pricing (Rochet & Tirole, 2001).
Fortunately, however, the intuition behind the results is not greatly affected by the absence
of price elasticity. Second, there are many kinds of ownership connections and co-
operative agreements between securities market institutions. The analysis could be
K. Kauko / International Review of Financial Analysis 13 (2004) 585–600600
expanded by taking into account several alternative structures. The settlement system and
the book-entry register are often operated by different institutions. Moreover, the market
place could also be explicitly modelled. In many cases, stock exchanges operate their own
settlement systems. One might also include international SSSs. Third, the impact of
institutional problems that issuers face in issuing securities in a foreign system could also
be modelled. In the case of equities, there are largely unresolved legal problems. Finally,
there are more than two CSDs in the real world. Fortunately, the intuition behind the main
results is not dependent on the assumption of two CSDs. For instance, investors must use
the CSD chosen by issuers, irrespective of the number of CSDs. Establishing the link
would weaken this effect, making it possible and necessary to compete for investors,
irrespective of the number of CSDs.
Acknowledgements
Thanks are due to Juha Tarkka, Tuomas Takalo, Esa Jokivuolle, Kari Korhonen,
Daniela Russo and seminar participants at the Bank of Finland research department for
several insightful comments.
References
Barclay, M. J., Kandel, E., & Marx, L. (1998). The effects of transaction costs on stock prices and trading
volume. Journal of Financial Intermediation, 7, 130–150.
Di Noia, C. (1999). The stock-exchange industry: Network effects, implicit mergers and corporate governance
(Commissione Nazionale per le societa e la Borsa, Studi di Ricerche 33).
Economides, N., & Siow, A. (1988). The division of markets is limited by the extent of liquidity. American
Economic Review, 78, 108–121.
Giovannini, et al. (2001). Cross-border clearing and settlement arrangements in the European union (Brussels
November 2001).
Katz, M. L., & Shapiro, C. (1986). Product compatibility in a market with technological progress. Oxford
Economic Papers, vol. 38 (pp. 146–165) (supplement).
Laffont, J. -J., Rey, P., & Tirole, J. (1998). Network competition: II. Price discrimination. Rand Journal of
Economics, 29, 38–56.
Lannoo, K., & Levin, M. (2001). The securities settlement industry in the EU (Centre for European Policy
Studies Research Report).
Martin, S. (2001). Industrial organization, a European perspective. Oxford University Press.
Park, Y. -S., & Ahn, B. -H. (1999). Joint ownership and interconnection pricing in network industries. Interna-
tional Journal of Finance and Economics, 8, 183–198.
Rochet, J. -C., & Tirole, J. (2001). Platform competition in two-sided markets (Institut d’economie industrielle
working papers, Universite de Toulouse I).
Shy, O. (2002). A quick and easy method for estimating switching costs. International Journal of Industrial
Organization, 20, 71–87.
Shy, O., & Tarkka, J. (2001). Stock exchange alliances, access fees and competition (Bank of Finland Discussion
Paper 22/2001).