Some Economics of Horizontal
Integration in the Payments Industry*
April 10, 2007
* Prepared for the conference "Nonbanks in the Payments System", Santa Fe, May 2-4, 2007 sponsored by the
Federal Reserve Bank of Kansas City. I benefited from the useful comments of Fumiko Hayashi, Stuart Weiner
and Zhu Wang.
** IDEI, Toulouse University.
The structure of the US retail payments industry has recently experienced dramatic
changes: increased concentration, entry of new firms (including nonbanks such as
telecommunication and utility companies or supermarket chains), consolidation and
reorganisation of IT infrastructures. This is largely due to the move to electronic payments
(which now represent more than half of noncash payments in the US), but it is also clearly
related to exogenous shocks of large magnitude such as financial deregulation. The on-going
integration of national payments system in the Euro area, encouraged by the SEPA initiative
of the European Commission, represents a shock of similar magnitude, also likely to generate
a dramatic change in the retail payments landscape of continental Europe. Other regions of the
world are also experiencing similar changes, as part of the consolidation and globalization of
their financial services industries.1 This article is a first pass at the economic analysis of the
consequences of horizontal integration in the payments industry. We first recall the general
principles of horizontal integration in other industries (Section 2). Then we study the
specificities of horizontal integration within the payments industry (Section 3). Section 4
draws on the burgeoning literature on two-sided networks to try and derive some principles of
horizontal integration in two-sided industries. Finally, Section 5 concludes by suggesting
some policy implications.
Principles of Horizontal Integration
A large fraction of mergers occur in waves, as documented by Mitchell and Mulherin
(1996) and Andrade et al. (2001). This is illustrated by Figure 1, taken from Andrade et al.
Figure 1: Aggregate Merger Activity
Source: Andrade et al. (2001).
1 Claessens et al. (2003) show that networks play an increasing part in the financial services industries of many
countries and argue that competition policy should be adapted accordingly.
A classical explanation of this phenomenon is that these waves of mergers are triggered by
exogenous shocks such as deregulation or development of new technologies. This is
confirmed by the observation that, within a wave, mergers seem to cluster by industry, as
illustrated by the following table, also taken from Andrade et al. (2001).
Table 1: Top 5 Industries based on Average Annual Merger Activity
1970s 1980s 1990s
Oil & Gas
Media & Telecom.
Oil & Gas
Source: Andrade et al. (2001).
Empirical evidence also suggests that mergers tend to discipline managers: corporate
performance typically improves after a merger (Healy et al., 1992). However, Andrade et al.
show that hostile takeovers have become less frequent in the recent years. In the US for
example, 14.3% of takeovers were hostile in the 1980s, but the figure has fallen to 4.0% in the
Stock market reactions
Surprisingly an horizontal merger does not seem to generate (on average) any
significant change in the stock price of the acquiring firm but entails typically a significant
increase for the target (Jensen and Ruback, 1983), (Andrade et al., 2001). Thus mergers
increase shareholder value but this increase is mostly appropriated by the target shareholders.
Another interesting empirical regularity is that the financing mode of the merger or
acquisition has an impact on value creation (Loughran and Vijh, 1997). Indeed, mergers and
acquisitions financed by cash seem to have no impact on acquirer shares, but lead to a large
increase (18%) on target shares. By contrast, stock swaps (which amount to the combination
of a cash takeover and an equity issue) seem to have a negative impact on acquirer shares, and
lead to a smaller increase (11%) on target shares. A possible explanation for this is the market
timing hypothesis (Shleifer and Vishny 2003, Rhodes-Kropf and Viswanathan 2004)
according to which mergers and acquisitions may be triggered by overvaluation of acquirers'
shares: the acquirer uses its overvalued stock to purchase the target. The overvaluation of the
acquirer's shares is then corrected by the market. This may explain why stock swap mergers
and acquisitions underperform cash financed ones.
Eckbo (1983) documents that the announcement of horizontal mergers often has a (short term)
negative impact on the shares of rival firms. However, Yan (2006) finds that longer term
impacts of horizontal mergers vary according to the "clusteredness" of the merger. He argues
that waves of mergers may have a prisoners' dilemma component: due to imperfect product
market competition, firms may be "forced" to merge in order to exploit technological
synergies, even though this leads to an increase in competition and a decrease in profit. By
contrast, "off-the-wave" mergers, simply driven by fixed costs savings, generate positive
externalities on rivals. More generally, there is still a controversy on the long-run impact of
mergers, in part because long term effects are almost impossible to assess (due to the absence
of a clear benchmark, and the difficulty to estimate long term" normal" returns).
Reasons for mergers
From a conceptual viewpoint, the reasons for mergers and acquisitions can be
classified into two categories. The first include those that increase economic surplus:
increasing technological efficiency, by exploiting scale economies or cost
synergies (rationalisation of some activities),
increasing financial efficiency by obtaining a better access to capital
markets (scale economies and diversification),
improving governance, by removing bad managers, and thus providing
incentives for managerial effort.
The second category includes the mergers and acquisitions that are beneficial to
shareholders or managers2 but are detrimental to society as a whole. The objectives of these
mergers and acquisitions can be:
increasing market power by reducing competition and/or facilitating
allowing managers to build empires and increase their power and their
perks (Jensen, 1986; Gorton et al., 2005).
The challenge for antitrust laws and regulatory agencies is to find a way to prevent the
second type of mergers while allowing the first type. Economic analysis is a useful guide for
Economic Analysis of Mergers and Acquisitions
As already mentioned, horizontal mergers can reduce competition but may also give
rise to efficiency gains (by coordinating production between different production units or by
exploiting synergies). Economists have thus thrived to find simple criteria for helping
competition authorities or regulatory agencies decide whether or not to allow specific mergers
and acquisitions. A certain number of surprising results have emerged. First Stigler (1950)
2 A more recent literature (Moeller et al. 2005, Malmendier and Tate, 2005) relies on behavioral assumptions:
overconfident managers overestimate the future gains from mergers and acquisitions.
noted that firms which do not participate in a merger may benefit more than the firms who do
participate, thus reducing the private incentives for mergers. Salant, Switzer and Reynolds
(1983) (henceforth SSR) went further by suggesting that mergers may in general be
unprofitable to the merging firms, basing their suggestion on the remark that the profit of a
firm in a Cournot oligopoly with n firms is typically less than the total profit of two firms in a
Cournot oligopoly with (n +1) firms. This tends to suggest that the actual mergers that we
observe in practice may correspond to managerial empire building and be in reality
detrimental to the shareholders of merged firms, as well as to consumers. However the
assumptions of the SSR paper have been widely criticized. Deneckere and Davidson (1985)
show for example that the existence of product differentiation can reverse the conclusion of
SSR. This is because reaction functions are typically upward sloping when firms compete in
prices (like in the Bertrand model with differentiated products) while they are downward
sloping when firms compete in quantities (like in a Cournot model). Thus when products are
differentiated, the initial price increase by the merging firms (associated with higher market
power) is reinforced by the reactions of outsiders (who also increase their prices) thus making
the merger profitable.
Perry and Porter (1985) also show that horizontal mergers can be profitable when
marginal costs are not constant (as supposed by SSR) but increasing. Similarly, Farrell and
Shapiro (1990) criticize the use of the Herfindahl index3 as an indicator of competitiveness for
an industry. Very often, Competition Authorities tend to permit mergers that maintain the
Herfindahl index of the industry below a certain threshold. Farrell and Shapiro show that
output (or price) changes have to be taken into account as well. They provide conditions
under which any merger that does not create synergies raises price. Spector (2003) generalizes
their result and shows that if marginal costs are non decreasing then any (profitable) merger
that does not generate technological synergies4 causes price to raise. This result holds true
even if new firms enter after the merger and if duplication of fixed costs is avoided among
Thus the general message provided by economic analysis is clear: in a traditional
industry, mergers that do not generate cost synergies are detrimental to social welfare. We
show below that this result might no be true anymore in a two-sided industry like the
Both economic theory and empirical evidence thus suggest that mergers may lead to
price increases, unless they generate sizable cost savings. Thus candidates for mergers need to
convince competition authorities (and courts of justice) that such cost synergies will be
3 The Herfindahl index of an industry is defined as the sum of the squares of the market shares of the firms
present in the industry. A low Herfindahl index indicates that concentration is small. If for example the market is
shared equally between n firms, the Herfindahl index is 1/n.
4 A merger allows merging firms to reduce their costs by rationalizing production, i.e. by coordinating output
decisions across production units. Technological synergies correspond to improvements in the production
technology that go beyond this simple rationalization.
present if the merger is allowed. However such efficiency defences are hard to evaluate in
prospect even for experts of the field. This argument is often used to suggest that it is better to
leave the decision (to allow the merger or not) to a regulatory agency (that has the competent
staff to assess such efficiency defences) rather than to a court of justice that would have to
resort to the advice of outside experts and may have trouble interpreting this advice, unless
there is direct evidence that the merger would harm competition (like in the celebrated
Staples-Office Depot case5).
Some economists have suggested to use stock market evidence as an indicator of the
impact of mergers on competition. For example if the stock price of rivals increases upon the
announcement of the merger, this might indicate that the merger will reduce competition.
However, mergers can have two different types of effects on competition: unilateral effects
(less competition) and coordinated effects (more collusion). A merger that helps collusion will
indeed lead to stock price increases for rivals, but the reverse is true if the merger increases
the market power of insiders, to the detriment of outsiders (exclusionary effects). This shows
that competition authorities and courts of justice have to be cautious when using stock market
evidence for assessing the competitive effects of a merger. In any case, any systematic rule
linking stock price changes to merger decisions would be immediately incorporated into the
expectations of investors, which would very likely destroy its significance (this is a variant of
the Goodhart law on monetary policy indicators).
Application to the Payments Industry
General Consolidation of the Financial Services Industry
The financial services industries of many countries have experienced a general
consolidation movement since the 1990s (see G-10 2001).
Table 2: Financial sector mergers and acquisitions with value greater than USD 1 billion
8 10 6 11 14 23 21 49 58 46
bn) 26.5 22.1 12.4 39.7 23.7 113.0
59.0 233.0 431.0 291.0
Source: Thomson Financial, SDC Platinum, cited in G-10 (2001).
Due to deregulation, globalization and technical change, the number of financial
institutions has decreased and concentration has decreased, in particular in the banking sector,
the focus of the present article. This movement of consolidation has taken two forms: mergers
5 The proposed merger of Staples and Office Depot (two of the three leading office supply superstore chains in
the USA) was challenged by the FTC by showing that Staples was able to charge higher prices in regions where
it competed with Office Depot. This convinced the Court that the merger would lead to a price increase in these
regions, and thus harm consumers.
and acquisitions (like in other industries) and cooperative arrangements like alliances, joint
ventures, and outsourcing of payment processing to jointly owned entities. Cooperative
agreements of this sort have a long tradition in banking, but there is a growing tendency for
banks to specialise in the "sales functions" (collecting deposits and providing payment
instruments) while outsourcing the "production function" (processing of payments) to
specialized entities. Similarly, mergers and acquisitions in the banking sector have often been
followed by internal reorganisation and consolidation of IT infrastructures payment functions
and accounting systems. In any case consolidation does not necessarily reduce competition in
a network industry. For example, in ATM networks, consolidation may enhance competition
for retail deposits by allowing small banks and large banks equal access to a large number of
ATM locations. By contrast, competition might be hindered if ATM networks foreclose new
entrants.6 Therefore it is not consolidation per se that matters, but the governance structure
and the criteria of access to the interbank cooperative entities that are necessary for payments
activities to be undertaken efficiently.
Whereas deregulation was the main driver of consolidation in the US banking
industry, the creation of Euro is likely to have similar consequences in continental Europe, but
probably with some delay. Anxious to stimulate competition for payment services, the
European Commission has launched the SEPA (Single European Payments Area) initiative.
SEPA aims to create a single market for payments throughout the Euro area. The idea is to
integrate national payment systems, with the objective of generating economies of scale and
making cross-border competition feasible.
Nonbanks in the Payment System
A recent book by Bradford, Davies and Weiner (2003) (henceforth, BDW) shows that,
although nonbank participation in the US payments system has always existed, it has
increased dramatically in the recent years. This goes hand in hand with the development of
electronic payments and the consolidation in the banking industry. Roughly speaking, BDW
classify payment activities in three categories: authorization, processing and instrument
provision. Processing activities tend to be dominated by nonbanks. A good example is First
Data, who controlled 300 million card accounts (in 2003) in the card-issuer processing
business (another big player in TSYS, who controlled 250 million card accounts in 2003) and
simultaneously controlled almost of the market for processing card transactions on the
merchant side.7 By contrast, card networks are largely controlled by banks or bank-owned
nonbanks (but this could change if proprietary networks or merchant-controlled networks
gained market share).
6 Matutes and Padilla (1994) provide a strategic analysis of cooperation between competing banks within ATM
networks. They show that full cooperation (the socially optimal outcome) is never spontaneously chosen by
7 Since its merger with Concord EFS, First Data also controls a large ATM network.
Figure 2: Ownership of Top 20 Regional ATM Networks
United States, 1985-2005
Source: EFT Network Data Book (various years), cited by Sullivan (2006).
Figure 3: Share of ATM Transaction Volume by Ownership of ATM Network
Notes: The break at 2002-2003 is due to different methods of calculating transaction volume. Prior to 2003,
many ATM transactions were counted by more than one ATM network. As a result, measures of aggregate
market share could be above 100 percent. Much of the double counting was eliminated for 2003 to 2005.
Source: EFT Network Data Book (various years), cited by Sullivan (2006).
In any case the increasing role of nonbanks in the payment industry seems to be essentially
important from the point of view systemic (or "system-wide") risk, which is outside the scope
of this paper.8 Given our focus on competition issues, what matters for us is the governance
8 This aspect is discussed in Sullivan (2006).
structure of an access rules to payment networks, and not so much whether they are owned or
controlled by banks or nonbanks.
Integration in Two-Sided Networks
By the very nature of the payment activity, any means of payment provides a joint
service to two distinct users, the payor and the payee, whom we will call for simplicity the
buyer and the seller. Except when they can perfectly bargain ex-ante on the sharing of
transaction fees (the cardholder fee and the merchant service charge in the case of card
payments) the structure of prices (i.e. the relative contributions of the buyer and the seller to
the total cost of the payment) and not only the total price (i.e. the sum of the fees paid by the
buyer and the seller) matters. Thus the payments industry is two-sided9 (see Rochet and Tirole
2006 for a more formal definition). This section surveys briefly some economics of two-sided
networks and shows how traditional antitrust analysis has to be amended to take care of this
"two-sidedness" of the payments industry.
Horizontal Integration in a Network Industry
Before delving into the specificities of two sided industries, let us contrast the impact
of horizontal integration in a traditional industry and in a network industry. In a traditional
industry, when two or several (identical) firms with constant marginal cost compete in prices
(Bertrand competition), competition drives these prices down to the (common) marginal cost
of the competing firms, and thus leads to a situation that maximizes social welfare, defined as
the sum of consumer surplus and industry profit (see Appendix 1 for a formal analysis). By
contrast, if these firms merge and form a monopoly, the price will increase up to the
(monopoly) price that equalizes the Lerner index (price minus marginal cost divided by price)
with the inverse elasticity of demand. This is the main reason why mergers are often viewed
as detrimental to social welfare.
Of course, we have seen that this simple reasoning may not hold anymore in some
cases (differentiated products or decreasing marginal costs) but in essentially all other
situations, mergers tend to be detrimental to social welfare, unless they generate technological
Appendix 2 shows that in a network industry this fundamental result does not hold,
even in the simplest possible set-up: two identical firms with constant marginal costs. This is
because of the network externality: consumers get a higher utility from a bigger network,
therefore a greater scale of operation generates a higher economic surplus. Network
externalities are thus similar to increasing returns to scale.10
9 There are many other examples of two-sided (or multi-sided) industries, like media, software, intermediaries,…
See for example Evans (2002).
10 For an economic analysis of network externalities, see for example Economides (1996) and Katz and Shapiro
(1985). McAndrews (1997) discusses network issues in payment systems.
Measuring Market Power
The usual reference for reviewing mergers between horizontal competitors is the
Horizontal Merger Guidelines issued jointly by the U.S. Department of Justice and the F.T.C.
The main message of these Guidelines is that "mergers should not be permitted to create or
enhance market power… [defined as] the ability profitably to maintain prices above
competitive levels for a significant period of time.11" The associated test is called the SSNIP
test, and aims at determining whether an hypothetical profit-maximizing monopolist would
profitably impose (at least) a small but significant and non-transitory increase in price.
The SSNIP test is relatively straightforward to apply in a traditional (one-sided)
industry by comparing the actual loss that the monopoly would make, due to decreased sales
(consequently for the price increase) to the critical loss that the monopoly could afford, and
equal to the gain on inframarginal sales. If demand elasticity is big enough, the actual loss is
greater than the critical loss and the SSNIP (small but significant nontransitory increase in
price) would not be profitable for the hypothetical monopoly.
However in a two-sided industry like the payment industry there are two distinct sides
(the buyers' and the sellers') and thus two distinct prices, the price p paid by the buyer
(cardholder fee in the case of a payment card network) and the price p paid by the seller
(merchant service charge in the case of a payment card network). Thus it is not clear a priori if
the SSNIP test should be applied to one side (and the corresponding price) or the other.
However, Rochet and Tirole (2003) have studied the optimal pricing for a monopolist in a
two-sided network such as a payment card network and shown that the pricing decision could
be separated into two steps: choosing the price structure ( p , p ) that maximizes the volume
of payments for a given total price p + p ≡ p ; Then finding the total price p that maximizes
the total profit of the monopoly network.
Therefore, as explained in more detail by Emch and Scott Thompson (2006), the
SSNIP test can be extended to the payment industry12 by using the total price cost margin
(sum of price cost margins on the buyer and the seller side) as a measure of profitability (thus
determining the critical loss of a hypothetical monopoly) and the elasticity of the payment
volume (once price structure has been optimised) as a measure of substitutability with other
payment instruments (thus determining the actual loss of a hypothetical monopoly).
According to Emch and Scott-Thompson (2006), this methodology has been used in the
recent merger case U.S. vs First Data Corporation and Concord EFS.13 However, it is far from
11 See US DOJ (1997), Section 0.1.
12 Rochet and Tirole (2006) suggest a possible way to extend this methodology to other two-sided industries by
defining the notion of "per-interaction prices" p and p (roughly speaking, this is done by aggregating fixed
and variable fees on each side and dividing these aggregate prices by the number of "interactions"), "total price"
p = p + p and usage volume V. However there are important measurability issues ("usage volume" is not
easy to measure) and the approach does not generalize well to the case of several firms.
13 In 2003, the US government sued to block the merger of two large debit card networks (NYCE and STAR).
The case was settled before the trial, the parties agreeing to divest the NYCE network.