HAL Id: halshs-01015794
https://shs.hal.science/halshs-01015794v2
Preprint submitted on 5 Jun 2015
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-
entic research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diusion de documents
scientiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Measuring competition in banking: A critical review of
methods
Florian Leon
To cite this version:
Florian Leon. Measuring competition in banking: A critical review of methods. 2015. �halshs-
01015794v2�
Etudes et Documents n
o
12, CERDI , 2014
The author(s)
Florian L´eon, Clermont Universit´e, Universit´e d’Auvergne, CNRS, UMR 6587,
CERDI, F-63009 Clermont Fd
La s´erie des Etudes et Documents du CERDI est consultable sur le site :
http://www.cerdi.org/ed
Directeur de la publication : Vianney Dequiedt
Directeur de la r´edaction : Catherine Araujo Bonjean
Responsable d’´edition : Annie Cohade
ISSN : 2114 - 7957
Avertissement :
Les commentaires et analyses d´evelopp´es n’engagent que leurs auteurs qui restent
seuls responsables des erreurs et insuffisances.
2
This work was supported by the LABEX IDGM+ (ANR
-
10
-
LABX
-
14
-
01)
within the program
“Investissements d’Avenir” operated by the French National Research Agency (ANR)
Etudes et Documents n
o
12, CERDI , 2014
Abstract
Many studies have attempted to investigate the determinants and implications of
competition in the banking industry. The literature on the measurement of com-
petition can be divided between the structural and non-structural approaches. The
structural approach infers the degree of competition from the structure of the market.
The non-structural approach, based on the New Empirical Industrial Organization,
assesses the degree of competition directly by observing behavior of firms in the
market. This paper reviews the most frequently-used structural and non structural
measures of competition in banking. It highlights their strengths and weaknesses,
especially for studies based on a limited number of observations.
Mots cl´es / Key words : Competition, Bank, HHI, Lerner index, conjectural
variation model, Panzar-Rosse model, Boone indicator
Code JEL / JEL classification : D4, G21, L11, L13, O55
Acknowledgement / Remerciements : I am grateful to Samuel Gu´erineau and
Vianney Dequiedt for their helpful comments. All remaining errors are my own.
3
Etudes et Documents n
o
12, CERDI , 2014
1 Introduction
The issue of competition in the banking sector has attracted much interest in recent
years, not least because of the recent financial crisis. Alongside the usual concerns about
competition, the issue has additional significance in banking because of its crucial role on
non-financial activity. Many theoretical papers have attempted to explain the ambiguous
consequences of competition on access to credit, cost and quality of financial services,
innovation, the stability of financial systems, and thus economic development. To empiri-
cally address these important questions, one first needs to come up with reliable measures
of the intensity of bank competition. The more accurate the measure, the more precise
empirical results are likely to be.
The fact that competition is a complex notion, and therefore not directly observable, has
resulted in the development of many methods for its assessment. The assessment of com-
petition in the banking industry has a long tradition. The literature on the measurement
of competition is generally categorized into two major of streams. Based on traditional
Industrial Organization, early research focused on market structure-performance linkages
(the Structure-Conduct-Performance paradigm) which stated that the likelihood of col-
lusion increases with market concentration. Some authors, however, raised doubt about
the reliability of the Structure-Conduct-Performance paradigm and associated structural
measures of competition. In response to deficiencies found in the structural approach,
non-structural measures of competition have been developed. The aim of the New Em-
pirical Industrial Organization (NEIO) measures is to directly assess the competitive
conduct of firms. The first generation of non-structural measures is based on oligopoly
theory and a static model of competition. The Lerner index, the conjectural variation
model and the Panzar-Rosse model can all be attached to this conception of competition.
Subsequently, other non-structural measures, especially the Boone indicator, have been
developed with the objective of capturing the dynamic of the market rather than focusing
on static analysis.
While some researchers may prefer one measure over another, there is no consensus re-
garding the best measure by which to gauge competition. The different indicators of ban-
king market competition do not provide the same inferences about competition (Carb´o-
4
Etudes et Documents n
o
12, CERDI , 2014
Valverde et al., 2009; Liu et al., 2013). Therefore, the choice of a particular indicator in-
fluences conclusions regarding the implications of competition. The choice of techniques
involves tradeoffs. The usefulness of the different approaches hinges on data availabi-
lity, the conceptions of competition assumed, and the questions being addressed. This
paper’s objective is to present the most widely-applied methods in banking, highlighting
their strengths and weaknesses. It complements existing reviews written by Degryse et al.
(2009) and Liu et al. (2013).
The rest of the paper is structured as follows. Section 2 briefly presents the different
conceptions of competition in the history of economic thought. The following sections
review the different methodological approaches that have been employed to investigate
competition in banking. This empirical research can be subdivided into the structural
(Section 3) and non-structural (Section 4) approaches. These sections highlight assump-
tions, advantages and shortcomings of the different measures as well as differences and
similarities between them. The final section provides a synthesis for studies on banking
systems in developing countries.
2 Two conceptions of competition
Although the concept of competition has always been central to economic thinking, it is
one that has taken on a number of interpretations and meanings. Conceptions of compe-
tition originate in The Wealth of Nations (Smith, 1776).
1
In the Smithian analysis, free
competition is an ordering force toward equilibrium. In the long run, free competition
leads to prices being equal to the costs of production. Nonetheless, for Smith, competi-
tion is not a state or situation but a race between competitors to gain market share. It
is rivalry that forces price towards the equilibrium of supply and demand. An essential
condition for free competition is not the number of rivals (while it may help) but rather
individual freedom.
Subsequent works on the conception of competition were inspired by Smith but have
1. Competition was a familiar concept in economic writing when Smith published The Wealth of
Nations. McNulty (1967) points out that the merit of Smith was his description of the competitive
process as a central element of economy.
5
Etudes et Documents n
o
12, CERDI , 2014
developed two major views of competition (McNulty, 1967; Vickers, 1995; Blaug, 2001).
Standard theory refers to the results of competition as a static equilibrium outcome.
According to this theory, competition is a static state in which firms cannot charge over-
price and then earn abnormal profit. On the other hand, other economists, particularly
the Austrian School, have criticized this static view and have retained the central role
played by rivalry to define competition.
2.1 Competition as a static state
Cournot (1838) was the first to relate free competition to the result of competition.
2
Cour-
not defined the ideal of competition, not as the process that in the long run tends toward
a certain equilibrium position, but rather as the equilibrium condition itself. Competi-
tion is a situation where prices equal the costs of production (natural prices). To obtain
a competitive situation, there must be several assumptions (a considerable number of
rivals, possessing common knowledge about market opportunities, free entry and exit),
which was never explicitly mentioned by Smith with the exception of the number of rivals
(Blaug, 2001). This latter assumption plays a central role in Cournot’s analysis. Accor-
ding to him, the excess of the price of cost approaches zero as the number of producers
increases. The analytical refinement was extended by some economists, particularly Ed-
geworth, Jevons, Walras, Marshall, Clark, and received its fullest expression in Knight’s
book Risk, Uncertainty and Profit.
3
Perfect competition is the antithesis of monopoly. In
monopoly, there is no one to compete and a monopolist could extract abnormal profits,
although limited by the elasticity of demand.
By 1883, Bertrand criticized Cournot’s oligopoly theory arguing that relevant strategies
for firms are prices and not quantities. As a result, the linkages between structure and
conduct are less clear than postulates of the Cournot model. A half century later, Cham-
berlin (1933) and Robinson (1933) wrote other important contributions to the oligopoly
theory. They proposed reconciling perfect competition and reality by developing a theory
of workable competition. Much of the business world is a mixture of competition and
2. Stigler (1957) acknowledges that Cairnes, while attached to Classics, expressed a similar idea.
3. Stigler (1957) documents how each of them add and specify the neoclassical definition of competi-
tion.
6
Etudes et Documents n
o
12, CERDI , 2014
monopoly. Monopolistic competition is a type of imperfect competition such that many
producers sell products that are differentiated from one another as goods but are not
perfect substitutes.
4
In monopolistic competition, a firm takes the prices charged by its
rivals as a given and ignores the impact of its own prices on the prices of other firms.
Different critiques have been introduced in the static oligopoly theory that recognize
different possible forms of market structure (Vives, 2001). The oligopoly theory allows
scholars to derive testable hypotheses and therefore measure degree of competition. As a
consequence, the neoclassical conception of competition based on the oligopoly theory is
at the root of both structural and the majority of non-structural measures of competition
(the Lerner index, the conjectural variation model and the Panzar-Rosse model). The
oligopoly theory distinguishes between the different forms of market structure (perfect
competition, imperfect competition and monopoly). This conception of competition is,
however, challenged by another view that focuses on dynamic aspects of competitive
rivalry.
2.2 Competition as a process of rivalry
The Austrian School, led by von Mises, Schumpeter and Hayek, argue that economists
in the neoclassical tradition misuse the term competition by applying it to a state rather
than to a process.
5
Competition is viewed not as a static state but as a complex process of
rivalry between firms. The core of competition is the behavior of firms (and entrepreneurs)
in the market. Firms are engaged in a continuing dynamic competitive process, constantly
creating and adopting new products and processes in order to cope with competition.
The competition process acts as a selection mechanism through the destructive-creation
principle : Less efficient incumbents are removed and replaced by more efficient entrants.
Stigler (1957) defines competition as ”a rivalry between individuals (or groups or nations),
and it arises whenever two or more parties strive for something that all cannot obtain”.
4. In 1929 Hotelling developed the model of horizontal differentiation. He emphasized the softening
effect of differentiation on price competition. According to Vives (2001), Edgeworth considered product
differentiation and imperfect substitution.
5. Other economists have developed some related arguments such as the theorists of evolutionary
economics (Nelson and Winter, 1982, 2002) or heterodox economists (e.g. Perroux).
7
Etudes et Documents n
o
12, CERDI , 2014
Vickers (1995) points out that this rivalry ”encompasses all sorts of forms of rivalry
(market trading, auctions, races, wars of attrition, etc.), instruments of rivalry (prices,
advertising, R&D, takeover bids, effort levels, etc.), objects of rivalry (profits, market
share, corporate control, promotion, prices, survival, etc.), as well as types of rival”.
Box 1 : An attempt of synthesis : The theory of contestability
The Austrian School’s arguments were revived by the Chicago School. Economists in the
Chicago tradition tend to the view that many if not most markets tend to approximate perfect
competition in the long-run. Thus positive profits are considered a transitory phenomenon
since their presence stimulates entry and hence leads to their demise (Posner, 1979).
Baumol et al. (1982) give a neoclassical formalization of this idea through the theory of contes-
tability.
a
They argue that regulation is unnecessary when markets are contestable. In contes-
table markets, the threat of entry would not only restrain incumbents’ market power, but also
generally satisfy the requirements for static welfare maximization. A market is contestable if
(i) the entry is free and without limit ; (ii) the entry is absolute ; and, (iii) the entry is perfectly
reversible. Market forces ensure that monopoly power will usually be short lived. The intensity
of competition is then unrelated to market structure but linked to market contestability. Au-
dretsch et al. (2001) shed light on the divergence between the two conceptions of contestability.
In the theory of market contestability, the analysis remains static, divergence from long-run
equilibrium is temporary and focuses on prices. By contrast, the Austrian School argues that
disequilibrium and monopoly power are the normal functioning of competitive markets. The
theory of contestability cannot explain why entrepreneurs innovate or adopt risky strategies
in markets, and thus it cannot explain the evolution of economics.
Subsequent developments in industrial organization have rediscovered the major contributions
of the Austrian School (dynamics analysis, non-price strategy, etc.). The literature has provi-
ded significant advances moving beyond the traditional static models and price competition
(Audretsch et al., 2001). Audretsch et al. (2001) point out that the evolution of industrial
economics from static to dynamic analysis does not capture the central attribute of Austrian
dynamic competition, namely technical change.
a. While Baumol’s thought about competition and entrepreneurship was highly influenced by Schumpeter’s view, Baumol
cannot be considered as a member of the Chicago School.
In the Austrian School’s perspective, a market is competitive when rivals are sufficiently
aggressive to give an incumbent incentive to improve (better quality, lower price, new
services, more innovation, improved management, etc.) in order to maintain its advantage.
Inefficient firms are directly sanctioned by consumers while more efficient and innovative
companies are rewarded.
8
Etudes et Documents n
o
12, CERDI , 2014
The role of monopoly and market power is revisited in the Austrian School perspective.
While firms are unable to raise prices over marginal cost in a perfect competition fra-
mework, for the Austrians the existence of rents is a normal aspect of the competitive
process. In a free competitive market, each firm innovates and develops risky strategies in
order to gain a competitive advantage over its rivals. Firms that do obtain such an edge
temporarily derive static monopoly power during the interval before imitating competi-
tors replicate their innovation, or supersede it with one that is superior. Successful firms
earn temporary monopoly profits as their reward for risky strategies. As a result, a free
competitive market is compatible with market power and ”abnormal” profit rates.
The fact that the neoclassical conception of competition poses some clear testable hypo-
theses explains that a majority of competition measures are rooted in this model. The
structural approach refers to the structure-performance relationship that exists in Cour-
not’s analysis. The first generation of NEIO measures (the Lerner index, the conjectural
variation model and the Panzar-Rosse model) was built on this model. More recently, new
non-structural measures which are in many ways sympathetic to the Austrian conception
of competition have been proposed, most notably the Boone indicator.
3 Structural approach
3.1 From SCP paradigm to concentration measures
The Structure-Conduct-Performance (SCP) paradigm, initially developed by Mason (1939)
and Bain (1956), seeks to explain aspects of the conduct and performance of firms in terms
of the structural characteristics of the markets in which they operate. The structural cha-
racteristics of a market cover the number of firms and their absolute and relative size as
well as the entry and exit conditions and the extent of product differentiation. Market
structure is expected to influence the conduct of firms. Conduct variables include pricing
strategies, collusion and other forms of strategic decisions (such as product quality and
advertising expenditures). Conduct, influenced by structure determines performance. The
SCP paradigm’s most important insight is that the more concentrated an industry is, the
9
Etudes et Documents n
o
12, CERDI , 2014
easier it is for firms to operate in an uncompetitive manner. For instance, tacit collusion
becomes more likely when the number of firms operating in an industry decreases (Tirole,
1988). Remaining firms can exploit their market power in order to charge prices above
marginal costs and thus become more profitable at the expense of social welfare.
The SCP paradigm argues that competitive features of industry are inferred from struc-
tural characteristics. Empirical works focus on the number of firms and their relative size
in order to gauge market concentration. General agreement prevails on the elements that
constitute concentration measures. Concentration indices take both the distribution in
firm size (inequality) and the number of firms into account in a given market. Hall and
Tideman (1967) argue that a good concentration index should satisfy a number of key
criteria :
1. Concentration should be a one dimensional measure ;
2. Concentration in an industry should be independent of the size of that industry ;
3. Concentration should increase if the market share of any firm is increased at the
expense of a smaller firm ;
4. If each firm in a given industry is divided into two firms of equal size then the
concentration index should be reduced by one-half ;
5. When an industry is divided into N equal sized firms, a measure of competition
should be a decreasing function of N ;
6. A concentration measure should have a range of zero to one (while this property is
not strictly necessary, it makes the measure easier to interpret).
Attention is focused on the three widely-used measures of concentration namely the num-
ber of firms, the concentration ratios and the Herfindahl-Hirschman index (HHI).
6
Mea-
sures of concentration differ according to their weighting schemes and structure. Existing
measures of concentration do not respect all the criteria listed above.
The number of firms is the simplest index to compute insofar as the data requirement
is very limited. However, this index does not take into account the distribution of firms.
6. Bikker and Haaf (2002) present other concentration measures such as the Hall-Tideman index, the
Rosenbluth index, the CCI, the Hannah-Kay index, the U-index, the Hause indices, and the Entropy
measure. Insofar as industrial organization literature focuses on concentration ratios and the HHI, this
paper ignores the other concentration measures.
10
Etudes et Documents n
o
12, CERDI , 2014
The level of concentration between two industries may differ greatly if one industry is
dominated by one firm, while another industry assembles firms with same size. As a result,
few papers employ the number of firms as an index of concentration.
Simplicity and limited data requirements make the concentration ratio one of the most
frequently used measures of concentration in the empirical literature. The concentration
ratio requires more information than the number of firms, insofar as researchers need to
obtain the market share of leading firms. The k-firm concentration ratio measures the
market share of the top k firms in the industry :
CR
k
=
K
X
i=1
s
i
, with s
1
≥ · · · ≥ s
K
≥ s
N
, ∀N ≥ K
where s
i
is the market share of the i’th firm, when firms are ranked in descending order
of market share and N is the total number of firms. The index approaches zero for an
infinite number of equally sized firms and equals 1 if the firms included in the calculation
make up the entire industry. While there is no rule for the determination of the value
of K, commonly used values include 3, 5 or 10. By focusing only on the market share
of the top k firms, the concentration ratio takes no account for the size distribution of
remaining firms. For example, a merger between small firms may not be reflected in the
concentration ratio, although the market becomes more concentrated. Some of the criteria
listed above fail to be fulfilled by the concentration ratio (e.g., the third criteria).
The Herfindahl-Hirschman Index (HHI) is the concentration measure most frequently
used by researchers and antitrust agencies (Hirschman, 1964). The HHI is more data
intensive than the number of firms or the CR
k
, insofar as it requires information on the
entire firm size distribution (market share of each firm). It is computed by summing the
squares of the market share of all firms :
HHI =
N
X
i=1
s
2
i
, ∀i = 1, . . . , N
where N is the total number of firms in the market. The HHI index ranges between 1/N
(for equal-sized firms) and 1 for monopolies. According to current screening guidelines
in the USA, the banking industry is regarded to be a competitive market if the HHI is
less than 0.10, a somewhat concentrated market if the HHI lies between 0.10 and 0.18,
11
Etudes et Documents n
o
12, CERDI , 2014
and a very concentrated market if the HHI is more than 0.18 (Cetorelli, 1999). The HHI
stresses the importance of larger firms by assigning them a greater weight than smaller
ones, thus reflecting their relative importance. Contrary to the concentration ratio, the
HHI avoids the arbitrary cut-off by incorporating each firm individually.
3.2 Advantages and shortcomings
The major advantage of concentration measures is the low data requirement. Even for
developing countries, concentration measures can be computed at least at the natio-
nal level. The majority of contributions focusing on industrial organization in low- and
middle-income countries use concentration measures to proxy competition.
Unfortunately, the SCP paradigm and associated concentration measures suffer from
major conceptual and practical limitations. A number of studies have questioned the
theoretical underpinnings of these concentration measures. Under the SCP hypothesis, a
rise in concentration is regarded as increasing collusive opportunities between firms, and
hence would lead to higher prices and profitability. Alternative theories undermine the
linkage between structure and conduct. Even in a duopoly, price competition can be fully
efficient as the Bertrand equilibrium is a possible outcome. The theory of contestability
(Baumol et al., 1982) suggests that a concentrated markets industry can behave compe-
titively if hurdles for entry and exit are low. The threat of entry can exert pressure on
incumbents and keep the sector competitive. Other theories show that collusive actions
can be sustained even in the presence of many firms. For instance, multimarket contacts
raise the incentive for collusion by changing the relative costs and benefits of cooperating
(Bernheim and Whinston, 1990).
Not only is the linkage between structure and conduct uncertain, but the direction of
causality is also problematic. The efficiency structure (ES) hypothesis (Demsetz, 1973;
Peltzman, 1977) entails the notion that the structure of the market may reflect differences
in efficiency rather than a competitive situation. Under the ES hypothesis, firms that
have higher productive efficiency tend to gain market share, which can lead to higher
market concentration. Therefore, concentration indices are not exogenous and may reflect
12
Etudes et Documents n
o
12, CERDI , 2014
differences in terms of efficiency.
7
The core of the problem is the difficulty of knowing
what different levels of concentration exactly reflect. Market structure and concentration
may proxy for a whole range of conduct-determining bank and market characteristics,
including average bank size, bank complexity in terms of product variety and activities,
the ease of information flow within the market and the overall size of the market itself,
for instance (Ergungor, 2004).
Another part of the argument against concentration measures concerns empirical im-
plementation. The major practical problem concerns the appropriate definition of the
market. Defining the relevant geographical market (local, regional, or national) as well as
the product market can be difficult (Shaffer, 2004b). The relevant geographical market
may differ according to banking activities. For example, banks compete locally to provide
credit to SMEs due to informational requirements. Credit to transparent firms, however,
is based on hard information that is easily transferable in impersonal ways. Therefore, a
bank outside the local market may compete to provide funds for well-established firms.
At the extreme, the geography of the relevant market for granting loans to multinational
firms is international. Furthermore, all products that are substitutes need to be included
in the product market definition. In the financial sector, a number of substitutable pro-
ducts are supplied by non-banking firms. As a consequence, defining the relevant market
may be a complex issue. In practice, the choice is often constrained by data availability.
Studies on industrialized countries often produce market share and concentration indices
at the local level (such as the Metropolitan Statistical Areas in the U.S.). In studies on
developing economies, concentration indices can only be calculated at the national level
due to the lack of disaggregated data. This might raise concerns about the relevance of
concentration measures used in applied works. For example, studies on credit conditions
for SMEs should employ concentration measures at the local level insofar as SMEs borrow
locally. When disaggregated data are available, researchers should use proxy concentra-
tion at the local level. An illustration of this approach can be found in Chong et al. (2013).
7. Without referring to the ES hypothesis, one should raise concerns about the exogeneity of market
structure. In a two-stage game, such as the Salop model, entry decision (market structure) and price
setting (conduct) are both endogenous and are determined by the equilibrium of the game. A firm will
enter only if it expects to earn profit in the future.
13
Etudes et Documents n
o
12, CERDI , 2014
They consider each Chinese city as a separate banking market and measure concentration
using the number of branches for each bank in the city. Unfortunately, for many econo-
mies such intensive information is rarely provided. Researchers employ national measures
of concentration to infer conclusions about the degree of concentration in local banking
markets. They implicitly assume that national concentration indices are a good proxy for
local ones.
Despite the fact that concentration measures continue to be widely used in the banking
literature, they suffer from some theoretical and empirical limitations. Unfortunately,
especially in studies on developing countries, researchers do not benefit from rich infor-
mation on banks. They do, however, keep in mind that concentration and competition
are two different concepts that are not always related.
3.3 New structural measures : Regulation in banking
Concentration indicators are still used as the main structural indicators mainly because
they are easier to measure than other dimensions of structure (entry and exit conditions
and product differentiation). Nevertheless, several studies have begun to investigate other
elements of market structure, especially entry and exit conditions. The theory of contes-
tability argues that firms behave competitively in the absence of entry and exit barriers
(see : Box 1).
Barriers consist of both formal and informal obstacles to entering and exiting a mar-
ket. Financial regulation is one of the major constraints to free entry in banking. The
financial sector is among the most regulated sectors in many countries. Several contribu-
tions (Claessens, 2009; Demirg
¨
u¸c-Kunt and Peria, 2010) propose considering regulatory
framework to gauge the degree of contestability in banking. Following the initial work
of Barth, Caprio and Levine (Barth et al., 2005, 2013), a rich international database on
bank regulations (and supervision) has been developed and may be used. It takes into
account entry requirements for domestic and foreign banks, capital requirements and the
regulations affecting bank activities. Not only are de jure (in book) regulations included
but the database also considers regulator activities and de facto limits.
14
Etudes et Documents n
o
12, CERDI , 2014
A major limitation with this approach is the omission of non-regulatory barriers and sunk
costs that play a significant role in the banking industry (Dietsch, 1992). The degree of
contestability in banking is influenced by non-legal barriers, such as technical and in-
formational barriers. The existence of scale and scope economies may create barriers to
entry, while empirical contributions on developed economies fail to provide evidence for
the existence of scale and scope economies in banking (Degryse et al., 2009). Another
important element affecting the nature of competition in retail banking is the presence
of networks. Incumbents can choose to share or extend their network to exclude rivals
from the market and limit competition (Matutes and Padilla, 1994).
8
The most impor-
tant economic barrier is certainly the presence of informational rents of incumbent banks.
The informational gap between incumbents and outside banks acts as a barrier in the
market. Private information may limit effective competition from uninformed outside
banks (Sharpe, 1990; Fisher, 1990; Rajan, 1992) due to the adverse selection problem.
The adverse selection problem for potential entrant banks stems from their inability to
distinguish new (good) borrowers from old (bad) borrowers who have been rejected by
their previous bank. The adverse selection problem and incumbent banks’ information
act as a barrier to entry in the banking industry (Dell’Ariccia et al., 1999; Dell’Ariccia,
2001; Marquez, 2002). Unfortunately, no paper gives a simple method to assess the in-
formation gap between incumbents and outside banks.
9
One might expect that this gap
is larger in more opaque markets. The presence of an informational wedge may explain
why regulatory frameworks are weakly related to competition in banking for developing
economies (Delis, 2012).
8. The Stackelberg-Spence-Dixit model (Tirole, 1988) shows that incumbent(s) may endogenously
deter entry or at least try to limit the expansion of entrants through overinvestment in production
capacity. The bank branches network of incumbent banks is often heavy. An incumbent must establish
several branches and bear the high fixed costs to compete with incumbents.
9. Non-structural models have been developed to assess the sunk-cost (Sutton, 1991; Dick, 2007)
or the decision to enter for outside banks (Bresnahan and Reiss, 1991). These models require strong
assumptions and/or strong data requirements. As a result, they are rarely implemented in the banking
industry. For a presentation of these models and their implementation in the banking industry : see
Degryse et al. (2009).
15
Etudes et Documents n
o
12, CERDI , 2014
To summarize, although the structural approach is open to criticism, these proxies have
often been employed in recent contributions on developing countries. Information on
the structure of markets or regulatory framework is often the sole data that researchers
focusing on less mature banking systems can obtain. However, it is important to be aware
of the strong limitations of these different indicators. With the development of micro-data
on banking institutions, new measures of competition based on bank behavior have been
developed and are beginning to be used in developing countries.
4 New Empirical Industrial Organization
Shortcomings in the structural approach have led to a number of attempts to collect
empirical evidence on the nature of competition by observing conduct directly. The New
Empirical Industrial Organization (NEIO) make conclusions about competitive pressure
by directly observing the conduct of firms in the market. The NEIO employs a variety
of alternative methodologies requiring different data and assumptions. Carb´o-Valverde
et al. (2009) point out that the first generation of non-structural measures is based on
the oligopoly theory (neoclassical conception of competition). These models include the
Lerner index, the conjectural variation models (Iwata, 1974; Bresnahan, 1982; Lau, 1982)
and the Panzar and Rosse (1987) model. While sharing a common standard theoretical
framework, results are often divergent (Carb´o-Valverde et al., 2009; Liu et al., 2013).
A second generation of NEIO measures focus on the dynamics of markets and are thus in
line with the Austrian (dynamic) conception of competition. These include the Persistence
of Profits developed by Mueller (1977, 1986) and the Boone indicator recently proposed
by Boone (2008).
This subsection reviews the widely-used non-structural measures in banking, namely the
Lerner index, the conjectural variation model, the Panzar-Rosse model and the Boone
indicator.
10
10. Insofar as applications of Persistence of Profits in banking are scant, this method is not presented in
this paper. Furthermore its implementation may be complex in highly unstable countries, such as develo-
ping economies. The interested reader may refer to Goddard et al. (2011) for more on this methodology.
Other methods have been developed in the literature such as the Hall-Roeger model (Rezitis, 2010) and
the structural demand models (Dick, 2008). These models require strong assumptions and/or have strong
16
Etudes et Documents n
o
12, CERDI , 2014
4.1 Lerner index
4.1.1 From theory to application
The Lerner index (or price-cost margin) is a popular measure of market power in empirical
research. The market power of a firm is identified by the divergence between the firm’s
price and its marginal cost. The price and marginal cost should be equal in perfect
competition, but will diverge in less competitive environments. A bigger wedge between
price and marginal cost signals greater monopoly power.
The theoretical foundation of the Lerner index is rooted in static oligopoly theory. Let us
suppose a quantity-setting oligopoly model (Cournot model). In an industry producing a
single good, let P be the market price of product Q and let q
i
be the quantity produced
by firm i. The profit maximization problem for firm i is written as :
max
q
i
[P (Q)q
i
− C(q
i
, ω
l
)] (1)
where q
i
is the quantity produced by firm i, Q is the total quantity (Q =
P
J
j=1
q
j
), and
P (Q) the price in the market. C(q
i
, ω
l
) is the total cost of firm i, where ω
l
is the vector
of the prices of the factors of production employed by firm i. Lerner (1934) proposes the
following measure of market power, known as the Lerner index :
L
i
=
P (Q) − C
0
q
i
(q
i
, ω
l
)
P (Q)
(2)
where C
0
q
i
(q
i
, ω
l
) is the marginal cost of firm i. The Lerner index ranges from 0 in situation
of perfect competition to the inverse of the price elasticity of demand in situation of
monopoly or collusion.
Although the Lerner has been known by economists since the mid-1930s, its application
to banking is relatively recent due to the difficulty of assessing marginal costs. Marginal
costs have only been econometrically estimated during the last two decades and are
extracted from the estimation of the cost function. Cost function is often assessed using
the intermediation approach (see the Box 2) from a translog equation including a single
data requirements. As a result, they are rarely implemented in the banking industry. Implementation
of the persistence of the profit model remain scarce in banking. For a presentation of these models and
their implementation in the banking industry, the interested reader may refer to Degryse et al. (2009)
and Liu et al. (2013).
17
Etudes et Documents n
o
12, CERDI , 2014
output (total assets) and three inputs (labor, deposit, and physical capital). The translog
function is generally as follows :
ln(C
i
) = β
0
+ β
1
ln(q
i
) +
1
2
β
2
[ln(q
i
)]
2
+
3
X
l=1
b
l
[ln(ω
l,i
)] +
1
2
3
X
l=1
b
3+l
[ln(ω
l,i
)]
2
+
3
X
l=1
β
2+l
[ln(q
i
)][ln(ω
l,i
)] +
X
l6=l
0
b
6+l
[ln(ω
l,i
)][ln(ω
l
0
,i
)] +
K
X
k=1
κ
k
Z
k,i
+ ε
i
(3)
where C
i
(= C(q
i
, ω
l
)) represents total bank costs of bank i, q represents a proxy of bank
output (total assets), ω
l
the price of l
th
input, and Z a set of control variables. From Eq.
3, the marginal cost is merely obtained by taking the first derivative and multiplying by
the average cost :
C
0
q
i
=
∂C
i
∂q
i
=
β
1
+ β
2
ln(y
i
) +
3
X
l=1
β
2+l
[ln(ω
l,i
)]
!
C
i
q
i
(4)
The price of output (P ) is computed as the average revenue.
This indicator is a good measure of individual market power. It allows researchers to
simply quantify the pricing market power of individual bank. The Lerner index has the
main advantage to be bank-specific and to vary over time, allowing comparison of market
power among banks and/or over the period.
Many papers have tried to assess banking competition by averaging the individual Lerner
indices (Fern´andez de Guevara et al., 2005, 2007; Maudos and Sol´ıs, 2011; Weill, 2013,
among others). The Lerner index for the market j is obtained as follows :
L
j
=
X
i∈j
φ
ij
L
ij
where L
ij
is the Lerner index of firm i in market or country j and φ
ij
the weighting of
firm i (often the market share of firm i in market j). An unweighted Lerner index implies
that φ
i
= 1/N, where N is the number of firms in market j.
18
Etudes et Documents n
o
12, CERDI , 2014
Box 2 : Models of banking firms and measures of prices and quantity
Non-structural measures of competition often require a model of a banking firm to specify the
output(s) and input(s). In the literature, two main approaches may be identified.
Under the production approach, financial institutions are thought of as primarily producing
services for account holders. Banks offer various financial services such as savings or credit by
mobilizing labor and physical capital. This view argues that deposit products provide a valuable
service for depositors (safe storage of valuable, record keeping, and means of payment).
Many empirical studies use an alternative approach ; the intermediation approach (Klein, 1971;
Monti, 1972; Sealey and Lindley, 1977). A bank is considered as an intermediary between
depositors and borrowers. A bank employs labor and physical capital to attract deposits,
which are used to fund loans. In addition to labor and physical capital, deposits are considered
as an input. Bank output is often defined as total assets, or total loans.
Insofar as intermediation function remains the core of bank activity in developing economies,
this thesis keeps the traditional intermediation approach. Under the intermediation approach,
outputs are computed as total assets. Contrary to total loans, total assets take into account
other earning assets.
a
Three inputs are considered, namely labor, physical capital and deposits.
Insofar as prices are not directly observable, researchers use balance sheets and income state-
ments to infer prices. Using banks’ financial figures to proxy prices may be problematic. Prices
reflect average prices for each bank and mix the price of each product (or factors) and the
mixing of product (or factors). Changes in prices may reflect changes in absolute prices or
changes in the composition of outputs or inputs. By contrast, changes in absolute (outputs or
inputs) prices are not automatically captured by average price if the bank modifies the com-
position of its portfolio or of its production factors. One might keep in mind these limitations
and be cautious when interpreting results employing ”endogenous” prices.
a. Various factors have fostered the growth of non-traditional banking activities, especially in mature markets. Non-
traditional activities include securitization, standby letters of credit, and derivative securities, which have rarely been considered
as outputs. Recent contributions, in particular on developed countries, should more explicitly consider these new products insofar
as their share in bank revenues increases.
4.1.2 Advantages and shortcomings
The main reasons for the popularity of the Lerner index are its simplicity, its straight-
forward interpretation, and the fact that it does not pose stringent data requirements.
Insofar as the Lerner index provides a firm-year specific measure of market power, it offers
the possibility of studying the evolution of bank pricing behavior over time. This indicator
has other interesting advantages. The Lerner index is a flexible indicator and does not
require defining the relevant market. It allows market power to be measured separately
19
Etudes et Documents n
o
12, CERDI , 2014
for the different banking markets (geographic and by products). Furthermore, researchers
may easily disentangle monopoly and monoposony power by excluding financial costs to
total costs and deposit prices (Turk Ariss, 2010). Finally, the Lerner index can be calcu-
lated with a limited number of observations. The latter advantage is far from anecdotal
insofar as competitive concerns occur mainly when the number of firms is limited (in such
cases, the marginal costs may be proxied by average costs).
The Lerner index, however, suffers from major theoretical and practical limitations. In
fact, it is a measure of pricing market power and not a proxy of competition. In other
words, an increase of average market power over time can be consistent with an increase
in the intensity of competition. Contributions show that there are theoretically possible
scenarios in which price-cost margins increase with more intense competition (Stiglitz,
1987, 1989; Bulow and Klemperer, 2002; Amir, 2010, among others).
11
Recent works show
that even if individual Lerner indices decreases with competition, the average degree of
market power may increase, decrease or remain stable due to the reallocation effect from
inefficient to efficient firms (Boone, 2008; Boone et al., 2013). Efficient firms have higher
price-cost margin than their counterparts. Thus, the weighted average Lerner index can
increase if the increase in the market share of more efficient firms overcompensates the
decrease of the respective individual Lerner indices.
12
Other practical concerns have been raised about the Lerner index. Vives (2008) main-
tains that the Lerner index is not able to appropriately capture the degree of product
11. The Austrian School reaches similar conclusions in a completely different framework. In the Aus-
trian perspective, competition gives firms incentives to innovate in order to cope with competition.
Incumbent’s banks will develop new products or expand their network of branches to maintain an ad-
vantage over their rivals. These banks can continue to extract some margins on customers even after the
increase of competition. An increase of Lerner indices over time does not necessarily signal a worsening
of competitive conditions. On the contrary, this might suggest a strengthening of competitive conditions.
Extraction of market power may be a normal feature of competitive behavior. Therefore, the Lerner
index is, at least theoretically, potentially misleading.
12. The reallocation effect can be partly eliminated by using the unweighted Lerner index as measure
of competition. This reduces the problem caused by the reallocation effect to a certain extent but does
not remove it completely : an increase in competition can induce the destruction of inefficient firms (with
low margins) which raises the average Lerner index in the market.
20
Etudes et Documents n
o
12, CERDI , 2014
substitutability. Oliver et al. (2006) emphasize that, when a bank’s risk taking is not
accounted for, the Lerner index could overestimate market power, because banks that in
relative terms spend more of their resources granting credits enjoy higher margins. This
issue is particularly problematic for studies employing the Lerner index to investigate
the competition-stability nexus (Berger et al., 2009; Turk Ariss, 2010; Beck et al., 2013).
Koetter et al. (2012) point out that the conventional approach of computing the Ler-
ner index assumes perfect technical and allocative efficiency. Unfortunately, banks rarely
operate under perfect efficiency. Operating costs and efficiency vary depending on the
economic environment in which banks operate (Chaffai et al., 2001). As a consequence,
differences across countries (or changes over time) in Lerner indices can be justified by
differences (or changes) in non-competitive factors.
4.2 Conjectural-variation model
4.2.1 From theory to application
The Lerner index cannot distinguish between markets that have high margins due to
inelastic demand and markets that have high margins because they are less competitive
or perhaps collusive. To overcome this problem, the conjectural-variation method has
been introduced by Iwata (1974), Bresnahan (1982), and Lau (1982). The aim is to
control the changes of the Lerner index due to demand changes, and therefor isolate
firms’ competitive behavior.
The conjectural variation refers to the beliefs that one firm has about the way its com-
petitor(s) may react if it varies its output or price (Bowley, 1924).
13
Recalling the profit
maximization problem for firm i (Eq. 1), the first order condition for a firm i is given as :
P
0
Q
Q
0
q
i
q
i
+ P = C
0
q
i
(5)
13. The validity of the conjectural variation model is criticized by theorists as a dynamic principle in a
static framework (Tirole, 1988; Vives, 2001). There is no opportunity for rivals to react to firm i’s change
in the context of one-shot simultaneous moves games. Nevertheless, the conjectural variation approach
has proved useful in applied works because it parameterizes the degree of competition in a market (Tirole,
1988; Martin, 2002).
21
Etudes et Documents n
o
12, CERDI , 2014
For the firm i, the conjectural variation parameter (in quantity) is measured as :
λ
i
=
∂
P
j6=i
q
j
∂q
i
= Q
0
q
i
− 1 (6)
λ
i
is the expectation of firm i of the extent to which its own output initiatives will trigger
changes in the outputs of its rivals.
Eq. (5) becomes :
P
0
Q
(1 + λ
i
)q
i
+ P = C
0
q
i
(7)
By multiplying Eq. (7) by Q/P and rearranging the equation, the Lerner index can be
rewritten as :
L
i
=
P − C
0
q
i
P
= −
(1 + λ
i
)s
i
ε
d
(8)
where ε
d
is the elasticity of demand and s
i
the market share of firm i.
Firm i’s expectations about the reactions of its rivals (conjectural variation parameter)
ranges from −1 to N − 1 (N is the number of firms in the market). In a collusive si-
tuation if firm i increases its production by 1, all firms will do the same thing and raise
their production by one. Hence, the full exploitation of market power exercised by firm
i coincides with λ
i
= N − 1 and the total output increases by N unities.
14
Perfect com-
petition (or Bertrand competition) implies that λ = −1 what changes expression (8) in
the well-known P = C
0
q
i
condition. Firm i expects a change in output of firms j which
exactly compensates its own, so as to leave the price unchanged. If λ = 0, then firm i
expects no reaction to its change in output and therefore it is a Cournot situation.
In practise, the conjectural elasticities is employed instead of the conjectural variation
(Degryse et al., 2009). The conjectural elasticities is defined as : θ
i
= Q
0
q
i
q
i
Q
. After some
manipulations, the Eq. 7 can be rewritten as (Angelini and Cetorelli, 2003) :
P = C
0
q
i
− λ
i
= C
0
q
i
−
θ
i
˜ε
i
(9)
where θ
i
is the conjectural elasticities and ˜ε
i
the semi-elasticity of demand (˜ε
i
= Q
0
P
/Q).
The conjectural elasticity ranges from 0 (perfect competition) to 1 (collusive situation).
The Cournot equilibrium or zero-conjectural variation model occurs when θ = 1/n.
15
Eq.
14. It is easy to show that the Lerner index equals the inverse of the elasticity of demand in this case.
15. Other contributions have extended the classical conjectural variation model. Coccorese (2005) de-
velops a model of conjectural variation in a price-setting framework and Suominen (1994) develops a
two-product Cournot model.
22
Etudes et Documents n
o
12, CERDI , 2014
9 can be rearranged to give an expression of the Lerner index : L
i
= λ
i
/P (Angelini and
Cetorelli, 2003).
A critical step in applied works is to convert the theory to data. This has historically
been done in two ways. The first approach is to estimate a parameter that represents the
behavior of firms. The Iwata (1974) model provides a framework for estimating conjec-
tural variation values for individual firms that supply homogenous products. Bresnahan
(1982) and Lau (1982) propose an alternative approach based on industry data. The ad-
vantage of this method is the ability to use industry aggregate data (more available than
firm-level data). More recently, different variants based on panel data have been pro-
posed (Angelini and Cetorelli, 2003; Coccorese, 2005, 2009; Uchida and Tsutsui, 2005).
All methods confront demand and supply equations in a system in order to disentangle
changes in elasticity demand and modifications of banks’ behavior. The demand equation
estimation gives the value of demand elasticity (ε) and the supply equation (based on Eq
9) provides the λ parameter. By confronting both parameters, the behavioral parameter
θ and the Lerner index can be easily derived.
4.2.2 Advantages and shortcomings
The conjectural parameter methods have been widely applied in banking following (Shaf-
fer, 1989, 1993). The structural model in this approach consists of a market demand
function and an associated supply function for either an individual firm or an average
firm. The advantage of this method is the direct analysis of firms’ conduct based on static
industrial organization theory. The parameter estimated can be treated as a continuous
variable under unrestricted conditions and test statistic maps into all oligopoly solution
concepts : perfect competition, Cournot competition, Bertrand competition, and collu-
sion (Shaffer, 2004b,a). The conjectural variation model builds a bridge between different
conceptions of oligopoly (Cournot and Bertrand). Furthermore, Shaffer (2004b) points
out that even if firms follow a strategy other than profit maximization or if regulations
constrain banks’ behavior, the test remains a valid technique. However, Xu et al. (2013)
shed light on the impossibility for this model to consider interest rate regulations. Finally,
this test can cope with monopsony power in a deposit market without revising the model
23
Etudes et Documents n
o
12, CERDI , 2014
(Shaffer, 1999).
The conjectural variation model is open to criticism. Insofar as the conjectural variation
approach is merely an elasticity-adjusted Lerner index, all concerns about Lerner apply
(see above). The conjectural variation methods have additional practical drawbacks, espe-
cially for studies on less mature banking markets. The structural model in this approach
requires specifying functional forms for demand and supply equations. One might raise
concerns about the functional forms adopted. Flexible functional forms often imply the
need to assess a large number of parameters. This induces two econometric problems.
Firstly, the risk of multicolinearity increases when researchers add interactions (Perloff
and Shen, 2012) ; secondly, studies based on a limited number of observations face major
small-sample issues (identification issue and instability). The latter issue is particularly
problematic insofar as concerns about competition occur in markets with few suppliers. As
a consequence, results may be highly sensitive to the specification employed, rising doubts
about the conclusions. It is therefore important to verify that the estimated parameters
are consistent with theoretical requirements (e.g. negative price elasticity or convex mar-
ginal costs). Furthermore, Shaffer (2001) points out that the estimated parameter can
exhibit a bias if the sample fails to span the complete market.
16
The latter shortcoming
may cause serious problems in cross-country samples based on a limited sample of banks.
4.3 Panzar and Rosse model
4.3.1 From theory to application
The Panzar-Rosse model (Rosse and Panzar, 1977; Panzar and Rosse, 1982, 1987) is the
most widely applied assessment of competition in the banking literature. This indicator
catches the transmission of input prices on firms’ revenues. Weak transmission is inter-
preted to indicate the exercise of market power in pricing and higher values indicate more
competition.
16. In the financial industry, this issue is far from anecdotal due to the development of non-traditional
banking activities by banks and non-bank financial firms. As a result, defining market borders becomes
more difficult.
24
Etudes et Documents n
o
12, CERDI , 2014
The intuition is straightforward in two opposite cases : collusion and perfect competition.
For a monopolist, marginal cost equals marginal revenue at the equilibrium. After input
prices increase, marginal cost increases. To maintain the equilibrium between marginal
cost and marginal revenue, the monopolist should increase the marginal revenue by redu-
cing the total quantity (insofar as marginal revenue is a decreasing function of quantity).
Rosse and Panzar (1977) show that total revenue is reduced if the price elasticity of de-
mand exceeds one. Intuitively, an increase in marginal cost reduces quantity but increases
output price. If the demand elasticity exceeds one, the gain due to price increase does not
compensate for the loss due to reduction of quantity. By contrast, in a competitive setting
an increase in input prices induces an increase in total revenue. Because cost functions
must be homogenous of a degree of one in the input prices, any increase in input prices
generates an equal percentage increase in costs. A firm’s revenue changes by the same
percentage as its total cost, and so by the same percentage as its inputs prices to ensure
the zero profit condition (total cost equals total revenue). The required adjustments in
the total quantity are achieved by a reduction in the number of firms (long-run equili-
brium). As a consequence, an increase of 1 percent in input prices induces an increase of
1 percent of total revenue in competitive markets.
From this theoretical framework, the identification of competitive conditions is obtained
by calculating the sum of the elasticities of the revenue with respect to all input prices.
The sum of elasticity, often called the H-statistic, ranges from −∞ to +1. The greater the
transmission of cost changes into revenue changes, the more competitive the market is.
Under perfect competition, input prices and total revenue increase by the same percentage
and the H-statistic equals one. Shaffer (1982) proves that the H-statistic value equals one
for a monopoly in a contestable market (free entry). The H-statistic is null or negative for
a monopoly. An increase in input prices induces a reduction of total revenue under certain
assumptions (e.g. demand elasticity higher than one). Vesala (1995) documents that the
H-statistic is non-positive in the monopolistic competition equilibrium without threat of
entry or for a collusive oligopolist. This measure is between 0 and 1 for a monopolistic
competitor (Rosse and Panzar, 1977; Panzar and Rosse, 1987; Vesala, 1995).
Applied economists estimate a reduced-form revenue equation. The test is obtained by
regressing revenue (in logarithm) on input prices (in logarithm) and other control va-
25
Etudes et Documents n
o
12, CERDI , 2014
riables :
17
ln(Rev
i
) = α +
L
X
l=1
β
l
ln(ω
l,i
) +
K
X
k=1
γ
k
Z
k,i
+ ε
i
(10)
where Rev is the total revenue (or interest revenue), ω
l
the price of l
th
input, and Z
k
a
set of control variables. In line with the intermediation approach, input prices consider
prices of labor, physical capital and deposits (see Box 2). The H-statistic is defined as :
H =
L
X
l=1
β
l
(11)
Hence, the H-statistic is the sum of the elasticities of the total interest revenue of the
banks with respect to their factor prices. It is worth stressing that the interpretation of the
H-statistic requires that the sector be in a long-run equilibrium insofar as a monopolistic
situation requires an endogenous number of firms.
4.3.2 Advantages and shortcomings
The success of the PR model can be explained by its simplicity and the fact that it
does not pose stringent data requirements. The test can be derived by running only one
equation requiring a few numbers of variables and banks. As a result, the PR model can
be obtained from a relatively small number of observations, which is crucial for studies
on less mature banking industry. Furthermore, Shaffer (2004a) points out that the PR
model is robust to the extent of the market as no specific market definition appears in
the revenue equation. Only the data from firms included in the sample are required to
estimate revenue equation. This is a huge advantage in cross-country studies (Claessens
and Laeven, 2004).
Nonetheless, these benefits come at the cost of other shortcomings. The major pitfall
concerns the econometric identification and the interpretation of the H-statistic. A sample
of firm-level observations in a long-run competitive equilibrium would exhibit H = 1,
while a sample of observations from a profit-maximizing monopoly yields H ≤ 0. However,
17. This specification assumes that the production function is a Cobb-Douglas. This assumption is
open to criticism and a more flexible production function should be used. However, De Brandt and Davis
(2000) show that loglinear specification reduces simultaneity bias and gives similar results to more flexible
equations (such as translog equations).
26
Etudes et Documents n
o
12, CERDI , 2014
theoretical studies report that the H-statistic can be negative in a competitive market
and positive for a monopoly. A negative H-statistic can occur even in highly competitive
conditions in the short-run with a fixed number of firms (Shaffer, 1983) or in the case of
constant average cost (Bikker et al., 2012). Shaffer and Spierdijk (2013) point out that the
H-statistic can be positive in highly noncompetitive settings. Furthermore, for firms facing
constant elasticity of demand, theoretical studies report the H-statistic as alternatively
an increasing (Shaffer, 1983) or decreasing (Panzar and Rosse, 1987) function of the
Lerner index. Put differently, higher values of the H-statistic do not necessarily imply
lower market power.
Another issue is related to the continuous nature of the H-statistic. This outlook has been
adopted in many empirical works (Bikker et al., 2012, Table 1). According to Panzar
and Rosse (1987), the H-statistic can be interpreted as a continuous monotonic index
of conduct ; not only the sign of the H-statistic matters, but also its magnitude. Shaffer
(2004b), however, casts doubt on the use of the H-statistic as a continuous measure of
competition although Vesala (1995) proves that the H-statistic is a continuum under
certain conditions. In empirical studies, the H-value is often considered as a continuum
value although the question remains unsolved (Bikker et al., 2012).
As a consequence, the interpretation of the value of the H-statistic is more complex than
the canonical PR model predicts. In fact, the interpretation of the H-statistic depends
on the hypotheses assumed regarding the market equilibrium, demand elasticity and
cost function (Shaffer, 1983; Panzar and Rosse, 1987; Vesala, 1995; Bikker et al., 2012).
Unfortunately, with the exception of the market equilibrium, applied economists cannot
test these required assumptions. As shown above, the long-run equilibrium (endogenous
number of firms) is necessary to derive testable hypotheses. An equilibrium test has been
developed to validate the market equilibrium hypothesis (Shaffer, 1982). The profits are
uncorrelated with the input prices in the long-run equilibrium. Hence, the test consists
in switching the total revenue by profit rates in Eq 10 :
ln(π
i
) = α
0
+
L
X
l=1
β
0
l
ln(w
l,i
) +
K
X
k=1
γ
0
k
Z
k,i
+ ε
0
i
(12)
where π is the return on assets. The E-statistic is E =
P
L
l=1
β
0
l
. The market equilibrium
condition is E = 0. Bikker et al. (2012) show that the equilibrium test is essentially a
27
Etudes et Documents n
o
12, CERDI , 2014
joint test for competitive conduct and long-run equilibrium, which substantially narrows
its applicability. In a market operating under imperfect competition or a monopoly, the
E-statistic may differ from 0.
Alongside these difficulties, applications suffer from additional limitations. First, the PR
model is sensitive to monopsony power. Inputs should be homogenous and their prices
exogenously fixed. The price of deposits is not always exogenous. A bank may behave as a
monopsony when alternative savings products are not available. Monopsony power would
tend to yield higher values of the H-statistic and thereby mask any market power present
on the output side (Shaffer, 2004a). The other non-structural measures of competition
are less subject to the monopsony issue.
18
The monopsony issue is rarely discussed in the
literature.
Finally, Bikker et al. (2012) acknowledge that some applications employ a scaled revenue
equation (introducing size as control) or estimate a price function by using the ratio of
revenue to total assets as dependent variable. However, the properties of the price and
revenue equations are identical in the case of long-run competitive equilibrium but criti-
cally different in the case of monopoly or oligopoly. This disqualifies a large number of
studies. The proper revenue equation must exclude the scale variable and use the total
revenue as dependent variable.
To summarize, the first generation of non-structural measures of bank competition is
based on standard oligopoly theory. The Lerner index assesses the average pricing mar-
ket power. The conjectural-variation model observes how rivals react to changes in a
bank’s behavior. The Panzar-Rosse model investigates how changes in input prices are
transmitted in revenue. A major advantage of these approaches is that they allow dis-
crimination between the different possible situation : collusion, perfect competition and
imperfect competition (oligopoly). Nonetheless, these measures neglect dynamics in the
market and non-pricing strategies. The second generation of NEIO measures of com-
petition takes into account these factors. In the following section, I present the Boone
indicator, which has begun to be used in the banking literature.
18. Other methods employ a cost function. As expressed above, the econometrician may run a cost
function excluding financial costs and deposits to test the sensitivity of results to monopsony power.
28
Etudes et Documents n
o
12, CERDI , 2014
4.4 The Boone indicator
4.4.1 From theory to application
Recently, Boone (2008) has extended the existing set of competition measures by sugges-
ting a new measure based on the idea that efficient firms are more highly rewarded in
more competitive markets. The indicator starts from the notion that in a more compe-
titive market firms are punished more harshly in terms of profits for being inefficient.
19
Under the efficiency hypothesis (Demsetz, 1973), more efficient firms achieve superior
performance in terms of higher profits at the expense of their less efficient rivals and also
attract greater market share. The Boone indicator exploits this reallocation effect from
inefficient to efficient firms. In the most extreme case, the reallocation effect is combi-
ned with a selection effect insofar as the least efficient firms leave the market. Boone
(2008) shows that the reallocation effect increases monotonically with the degree of com-
petition.
20
While an intensification of competition can decrease the output of firms, this
decrease will be smaller for more efficient firms. As a result, the market share and profits
for more efficient firms increase while those the less efficient firms shrink. Phrased diffe-
rently, the relative profit difference is sensitive to the degree of competition (Boone, 2008,
Theorem 1).
Boone (2008) demonstrates how relative profit difference can measure the level of compe-
tition in practice. Although the procedure is theoretically applicable, it is computationally
intensive. It requires the ranking of firms by efficiency levels. In practice, researchers often
gauge the strength of the relationship between efficiency and performance. The intensity
of competition is estimated from the following simple profitability equation (Boone et al.,
19. It should be noted that Hay and Liu (1997) develop a similar concept of competition by investigating
the relationship between market share and firm costs to assess the level of competition.
20. Boone (2008) considers two different forces as causing an increase in competition. The first is a fall
in entry barriers. The lower the entry barriers, the more firms should enter and the more competitive
the industry should be. This intuition underlines the use of the structural approach. Boone, however,
considers another source of competition : the aggressiveness of rivals’ conduct in the market.
29
Etudes et Documents n
o
12, CERDI , 2014
2007) :
21
lnπ
i
= α + βlnc
i
+ ε
i
(13)
where π
i
stands for profit and c
i
a measure of costs (proxied efficiency). The coefficient β
gives the profit elasticity (PE), that is, the percentage drop in profits of bank i as a result
of a percentage increase in bank i’s costs. This indicator is in theory negative, reflecting
the fact that higher marginal costs are associated with lower profits. In addition, its
value should be lower the more competitive market conditions are. Boone et al. (2007)
conducted simulations for the PE indicator and found that changes in competition are
correctly identified with this measure.
Recent contributions introduce two modifications of Eq. 13. First, Van Leuvensteijn et al.
(2011) directly estimate the marginal cost c
i
. When data allow for computing marginal
costs, studies employ this measure of efficiency (Delis, 2012; Tabak et al., 2012). Otherwise
researchers use imperfect proxies such as average costs (Schaeck and Cih´ak, 2013). Second,
efficient banks may choose to translate lower costs either into higher profits or into lower
output prices in order to gain market share. As a consequence, the bank’s market share
rather than its profits is often used as the dependent variable to take into account the
second possibility (Van Leuvensteijn et al., 2011, 2013; Tabak et al., 2012) :
lns
i
= α + β
0
lnc
i
+ ε
0
i
(14)
where s
i
is the market share of bank i.
Equations 13 and 14 are often run by introducing fixed-effects in order to take into
account unobserved heterogeneity. Recent works also control for endogeneity by running
an instrumental approach. Indeed, changes in performance and costs may be induced by
a third factor unrelated to competition.
21. Instead of in a log/log specification, the costs and profits can be expressed in levels. Employing
levels allows for introducing non positive profit values. In doing so, the estimation may introduce a bias
in the sample towards profitable banks. Boone et al. (2013) shows that results are weakly sensitive to
this exclusion. The main advantage of a log/log specification is to facilitate the interpretation of the β
coefficient as an elasticity.
30
Etudes et Documents n
o
12, CERDI , 2014
4.4.2 Advantages and shortcomings
The fact that the relationship between costs and profits is both continuous and monotonic
is the main advantage of the Boone approach. In almost all cases, higher competition
implies that the value of β is larger in absolute terms (more negative), and therefore,
β serves as a continuous indicator of competition. As expressed by Boone in different
works (Griffith et al., 2005; Boone et al., 2007; Boone, 2008; Boone et al., 2013), his
indicator avoids some major theoretical drawbacks of price-cost margin measures (see
Section 3.4.1.2). Moreover, while the identification of a situation from the PR model is
a challenge, the Boone indicators are monotonically related to competition. Furthermore
Xu et al. (2013) show that the existence of binding interest rate regulations renders the
Boone measure more robust than other NEIO measures.
In practice, the Boone approach has additional advantages for studies on developing coun-
tries. This method requires only information on profits (or market shares) and costs. In
particular, if costs are assessed by average costs, the computation of the Boone indicator
does not require information on prices. Furthermore, the Boone estimation is obtained by
a simple linear econometric specification (only one equation with one exogenous variable).
Like other models, the Boone indicator is a simplification of reality and suffers from
some limitations. The Boone indicator approach focuses on one important relationship
affected by competition, thereby disregarding other aspects. Efficient gains may not be
translated into lower prices or higher profits in the short-term. For instance, a bank may
invest these gains (developing new products, building brick-and-mortar branches, etc.) in
order to cope with competition in the future. Van Leuvensteijn et al. (2011) acknowledge
that these distortions are more likely when the Boone indicator is assessed year by year
rather than estimations covering the full sample period. Indeed, banking industry may be
defined as producing homogenous goods in the long term. Differences in terms of quality,
design and innovation are more or less equivalent among firms over the long term.
Schiersch and Schmidt-Ehmcke (2010) show that the Boone indicator makes critical as-
sumptions relative to the definition of the extent of the market. The more precisely we can
capture a market, the less other factors or markets influence the outcome and the better
31
Etudes et Documents n
o
12, CERDI , 2014
the subsequent competition estimates should be. This rule is common to all non-structural
measures of competition.
22
Finally, the β parameter is expected to be negative but can be positive in empirical imple-
mentation. The Boone indicator model assumes that efficiency should be one dimensional
and observable. Using costs is the simplest way to capture difference in efficiency. Howe-
ver, in a market where suppliers offer heterogenous goods, changes in costs may merely
reflect changes in strategies. In response to competitive pressure, banks may adopt strate-
gies to cope with competition (offering well-designed products, targeting new customers,
etc.) instead of reducing prices. In such cases, banks offering the most highly demanded
products may not only yield more profit but also spend more. The relationship between
marginal costs and performance thus turns to positive. Put differently, the β coefficients
may be positive when firms compete in quality (Tabak et al., 2012). In such cases, iden-
tification becomes impossible. One should note that this is not only a limit of the Boone
approach but also a limit of all non-structural measures. All non-structural measures
assume that banks offer homogenous goods and services.
The main advantage of the Boone indicator is that it can both capture market dynamics
and be easily implemented for a limited number of observations (by employing average
costs as measure of efficiency). Nonetheless, one should note that the Boone indicator is
a young tool and therefore has not yet been thoroughly scrutinized by the literature.
22. A second problem underlined by Schiersch and Schmidt-Ehmcke (2010) is related to firm size.
Under the model assumption, the most efficient firm must become the biggest firm in terms of market
share and must make the greatest profit. They note that in reality big firms are not necessarily the most
efficient ones and thus, it is possible to find a nonnegative β parameter. However, when panel data are
used, this issue can be avoided by including fixed-effects. As such, the β parameter will capture change
in terms of profit (or market share) and not the level differences among firms. Another possibility is to
normalize profits by size (Schiersch and Schmidt-Ehmcke, 2010). For example, Schaeck and Cih´ak (2013)
use the return on assets instead of profits as the dependent variable.
32
Etudes et Documents n
o
12, CERDI , 2014
5 Synthesis for studies on less mature banking sys-
tems
The important role of banking in the economy renders competition among banks a crucial
and timely policy issue. Many studies have attempted to investigate the determinants and
implications of competition in the banking industry. A critical step is to determine the
degree and/or evolution of competition. The literature on the measurement of competi-
tion can be divided between the structural and non-structural approaches. The structural
approach infers the degree of competition from the structure of the market. The likeli-
hood of collusion is higher in markets with fewer firms and high barriers to entry and
exit. The non-structural approach, based on the New Empirical Industrial Organization,
assesses the degree of competition directly by observing behavior of firms in the mar-
ket. The non-structural measures of competition can be sub-divided into two strands.
The first generation of NEIO measures originates in the neoclassical model of oligopoly
(Vives, 2001). This includes the Lerner index, the conjectural variation model, and the
Panzar-Rosse model. Another strand of empirical research considers a dynamic approach
of competition more in line with the Austrian School’s notion of competition. In ban-
king, the Boone indicator can be attached to this second generation of works.
23
Indeed,
each non-structural indicator of competition is based on different assumptions, and thus
measures different things. The Lerner index assesses the average degree of pricing power
in the market. The conjectural variation model investigates rivals’ reactions if a bank in
the market increases its output by one percent. The Panzar-Rosse model analyzes the
transmission of changes in input prices to bank revenue and the Boone indicator is based
on the idea that efficient firms are more highly rewarded in more competitive markets. As
a consequence, measures of competition cannot be considered as perfect substitutes. Each
indicator has its practical and theoretical advantages as well as its limitations. Table 1
provides a synthesis of advantages and disadvantages of each measure of competition.
This discussion has provided four main results. First, despite their large use in the empiri-
23. The Persistence of Profit is also attached to this generation but remain rarely applied in banking.
Moreover, its implementation may be complex in highly unstable countries, such as developing economies.
For more information, see : Goddard et al. (2011).
33
Etudes et Documents n
o
12, CERDI , 2014
cal papers, the structural measures suffer from major theoretical and practical limitations.
Second, while the Lerner index (and the associated conjectural elasticity) is a good mea-
sure of individual market power, it is not always the best measure of competition. The
average degree of market power may increase, decrease or remain stable, even if indivi-
dual Lerner indices decreases due to reallocation effect from inefficient to efficient firms.
Third, the Panzar-Rosse H-statistic is a good measure of competition in a static perspec-
tive. However, it requires strong assumptions (long-term equilibrium, demand elasticity
must exceed one, etc.) that cannot always be tested and checked. In addition, many em-
pirical applications suffer from limitations by estimating scaled revenue equation or price
equation. Finally, the Boone indicator catches dynamic aspects of competition but may
sometimes fail to identify the degree of competition in the short-run.
In studies focusing on less mature banking systems, the choice of measures of competi-
tion is often constrained by data. Despite its strong limitations, the structural approach
cannot be avoided when firm-level data are not available. This approach has the major
advantage of being easily computed for many markets. Non-structural measures require
access to substantial firm-level information. When firm-level information is available, the
Lerner index can be computed even with few numbers of firms. However, one should
remember that the Lerner index is not a measure of competition but rather a measure
of firm pricing power. The Lerner index is thus more useful when the objective is to
analyze the consequences of market power on behavior (e.g., risk-taking) or performance
(e.g., efficiency) of banks. The associated conjectural variation model is rarely used in
developing countries due to strong data requirements. By contrast, even with a limited
number of observations (less than one hundred), the Boone indicator can be estimated.
If price information is available, the Panzar-Rosse model can also be estimated.
In a perfect world, different measures of competition should be employed insofar as each
of them catches an aspect of competition. Researchers should be aware of the limitations
of their analysis if multiple measures cannot be considered.
34
Etudes et Documents n
o
12, CERDI , 2014
R´ef´erences
Amir, R. (2010). Market structure, scale economies and industry performance. available
at http ://kelley.iu.edu/BEPP/documents/AmirMarketStructure.pdf, 65.
Angelini, P. and Cetorelli, N. (2003). The effects of regulatory reform on competition in
the banking industry. Journal of Money, Credit and Banking, 35(5) :663–684.
Audretsch, D. B., Baumol, W., and Burke, A. E. (2001). Competition policy in dynamic
markets. International Journal of Industrial Organization, 19 :613–634.
Bain, J. (1956). Barriers to New Competition. Cambridge : Harvard Press.
Barth, J., Caprio, G., and Levine, R. (2013). Bank regulation and supervision in 180
countries from 1999 to 2011. Journal of Financial Economic Policy, 5(2) :111–219.
Barth, J. R., Caprio, G., and Levine, R. (2005). Rethinking bank regulation : Till angels
govern. Cambridge University Press.
Baumol, W. J., Panzar, J. C., and Willig, P. (1982). Contestable Markets and the Theory
of Industry Structure. Harcourt Brace Jovanovich.
Beck, T., De Jonghe, O., and Schepens, G. (2013). Bank competition and stability :
cross-country heterogeneity. Journal of Financial Intermediation, 22 :218–244.
Berger, A. N., Klapper, L. F., and Turk-Ariss, R. (2009). Bank competition and financial
stability. Journal of Financial Services Research, 35(2) :99–118.
Bernheim, B. D. and Whinston, M. D. (1990). Multimarket contact and collusive beha-
vior. The RAND Journal of Economics, pages 1–26.
Bikker, J. and Haaf, K. (2002). Measures of competition and concentration in the banking
industry : a review of the literature. Economic & Financial Modelling, 9(2) :53–98.
Bikker, J. A., Shaffer, S., and Spierdijk, L. (2012). Assessing competition with the Panzar-
Rosse model : The role of scale, costs, and equilibrium. Review of Economics and
Statistics, 94(4) :1025–1044.
35
Etudes et Documents n
o
12, CERDI , 2014
Blaug, M. (2001). Is competition such a good thing ? static efficiency versus dynamic
efficiency. Review of Industrial Organization, 19(1) :37–48.
Boone, J. (2008). A new way to measure competition. The Economic Journal,
118(531) :1245–1261.
Boone, J., van Ours, J. C., and van der Wiel, H. (2007). How (not) to measure competi-
tion.
Boone, J., van Ours, J. C., and van der Wiel, H. (2013). When is the price cost margin
a safe way to measure changes in competition ? De Economist, pages 1–23.
Bowley, A. (1924). The Mathematical Groundwork of Economics. Oxford University
Press.
Bresnahan, T. (1982). The oligopoly solution concept is identified. Economics Letters,
10(1-2) :87–92.
Bresnahan, T. F. and Reiss, P. C. (1991). Entry and competition in concentrated markets.
Journal of Political Economy, pages 977–1009.
Bulow, J. and Klemperer, P. (2002). Prices and the winner’s curse. RAND journal of
Economics, pages 1–21.
Carb´o-Valverde, S., Humphrey, D., Maudos, J., and Molyneux, P. (2009). Cross-country
comparisons of competition and pricing power in European banking. Journal of Inter-
national Money and Finance, 28(1) :115–134.
Cetorelli, N. (1999). Competitive analysis in banking : appraisal of the methodologies.
ECONOMIC PERSPECTIVES-FEDERAL RESERVE BANK OF CHICAGO, 23 :2–
15.
Chaffai, M. E., Dietsch, M., and Lozano-Vivas, A. (2001). Technological and environ-
mental differences in the european banking industries. Journal of Financial Services
Research, 19(2-3) :147–162.
36
Etudes et Documents n
o
12, CERDI , 2014
Chamberlin, E. H. (1933). The theory of monopolistic competition : A re-orientation of
the theory of value, volume 38. Harvard University Press Cambridge, MA.
Chong, T. T.-L., Liping, L., and Ongena, S. (2013). Does banking competition alleviate
or worsen credit constraints faced by small and medium enterprises ? evidence from
China. Journal of Banking & Finance, 37(9) :3412–3424.
Claessens, S. (2009). Competition in the financial sector : overview of competition policies.
World Bank Research Observer, 24(1) :83–118.
Claessens, S. and Laeven, L. (2004). What drives bank competition ? some international
evidence. Journal of Money, Credit and Banking, 36(3) :563–583.
Coccorese, P. (2005). Competition in markets with dominant firms : A note on the
evidence from the Italian banking industry. Journal of Banking & Finance, 29(5) :1083–
1093.
Coccorese, P. (2009). Market power in local banking monopolies. Journal of Banking &
Finance, 33(7) :1196–1210.
Cournot, A.-A. (1838). Recherches sur les principes math´ematiques de la th´eorie des
richesses. Hachette.
De Brandt, O. and Davis, E. (2000). Competition, contestability and market structure in
european banking sectors on the eve of emu. Journal of Banking & Finance, 24 :1045–
1066.
Degryse, H., Kim, M., and Ongena, S. (2009). Microeconometrics of banking : methods,
applications, and results. Oxford University Press, USA.
Delis, M. D. (2012). Bank competition, financial reform, and institutions : The importance
of being developed. Journal of Development Economics, 97(2) :450 – 465.
Dell’Ariccia, G. (2001). Asymmetric information and the structure of the banking indus-
try. European Economic Review, 45(10) :1957–1980.
37
Etudes et Documents n
o
12, CERDI , 2014
Dell’Ariccia, G., Friedman, E., and Marquez, R. (1999). Adverse selection as a barrier to
entry in the banking industry. The RAND Journal of Economics, pages 515–534.
Demirg
¨
u¸c-Kunt, A. and Peria, M. (2010). A framework for analyzing competition in the
banking sector : an application to the case of jordan. World Bank Policy Research
Working Paper, 5499 :1–24.
Demsetz, H. (1973). Industry structure, market rivalry and public policy. Journal of Law
and Economics, 16 :1–10.
Dick, A. A. (2007). Market size, service quality, and competition in banking. Journal of
Money, Credit and Banking, 39(1) :49–81.
Dick, A. A. (2008). Demand estimation and consumer welfare in the banking industry.
Journal of Banking & Finance, 32(8) :1661–1676.
Dietsch, M. (1992). Quel mod`ele de concurrence dans l’industrie bancaire ? Revue ´eco-
nomique, pages 229–260.
Ergungor, O. E. (2004). Conmment on ”bank competition and access to finance : In-
ternational evidence” by thorsten beck, asli demirg
¨
u¸c-kunt, and vojislav maksimovic.
Journal of Money, Credit and Banking, 36(3) :21–23.
Fern´andez de Guevara, J., Maudos, J., and Perez, F. (2005). Market power in European
banking sectors. Journal of Financial Services Research, 27(2) :109–137.
Fern´andez de Guevara, J., Maudos, J., and P´erez, F. (2007). Integration and competi-
tion in the european financial markets. Journal of International Money and Finance,
26(1) :26–45.
Fisher, K. (1990). Hausbankbeziehungen als Instrument des Bindung zwischen Banken une
Unternehman : Eine Theoretische une Empirische Analysis. PhD thesis, Universit
¨
at
Boon.
Goddard, J., Lui, H., Molyneux, P., and Wilson, J. (2011). The persistence of bank profit.
Journal of Banking & Finance, Forthcoming :1–10.
38
Etudes et Documents n
o
12, CERDI , 2014
Griffith, R., Boone, J., and Harrison, R. (2005). Measuring competition. AIM Research.
Hall, M. and Tideman, N. (1967). Measures of concentration. Journal of the American
Statistical Association, 62 :162–168.
Hay, D. A. and Liu, G. S. (1997). The efficiency of firms : What difference does competition
make ? The Economic Journal, 107 :597–617.
Hirschman, A. O. (1964). The paternity of an index. The American Economic Review,
54(5) :761–762.
Iwata, G. (1974). Measurement of conjectural variations in oligopoly. Econometrica :
Journal of the Econometric Society, 42(5) :947–966.
Klein, M. A. (1971). A theory of the banking firm. Journal of Money, Credit and Banking,
3 :205–218.
Koetter, M., Kolari, J., and Spierdijk, L. (2012). Enjoying the quiet life under deregu-
lation ? evidence from adjusted lerner indices for us banks. The Review of Economics
and Statistics, 94(2) :462–480.
Lau, L. (1982). On identifying the degree of competitiveness from industry price and
output data. Economics Letters, 10(1-2) :93–99.
Lerner, A. (1934). The concept of monopoly and the measurement of monopoly power.
The Review of Economic Studies, 1(3) :157–175.
Liu, H., Molyneux, P., and Wilson, J. O. (2013). Competition in banking : measure-
ment and interpretation. Handbook of Research Methods and Applications in Empirical
Finance, page 197.
Marquez, R. (2002). Competition, adverse selection, and information dispersion in the
banking industry. Review of Financial Studies, 15(3) :901.
Martin, S. (2002). Advanced Industrial Economics. Blackwell Publishers.
Mason, E. (1939). Price and production policies of large-scale enterprise. The American
Economic Review, 29(1) :61–74.
39
Etudes et Documents n
o
12, CERDI , 2014
Matutes, C. and Padilla, A. J. (1994). Shared atm networks and banking competition.
European economic review, 38(5) :1113–1138.
Maudos, J. and Sol´ıs, L. (2011). Deregulation, liberalization and consolidation of the
mexican banking system : Effects on competition. Journal of International Money and
Finance, 30(2) :337–353.
McNulty, P. J. (1967). A note on the history of perfect competition. The Journal of
Political Economy, 75(4) :395–399.
Monti, M. (1972). Deposit, credit, and interest rate determinantion under alternative bank
objectives. In Szego, G. P. and Shell, K., editors, Mathematical Methods in Investment
and Finance. Amsterdan : North-Holland.
Mueller, D. C. (1977). The persistence of profits above the norm. Economica,
44(176) :369–380.
Mueller, D. C. (1986). Profits in the long run. Cambridge University Press.
Nelson, R. R. and Winter, S. G. (1982). An evolutionary theory of economic change.
Harvard University Press.
Nelson, R. R. and Winter, S. G. (2002). Evolutionary theorizing in economics. The
Journal of Economic Perspectives, 16(2) :23–46.
Oliver, A. M., Fum´as, V. S., and Saurina, J. (2006). Risk premium and market power in
credit markets. Economics Letters, 93(3) :450–456.
Panzar, J. and Rosse, J. (1982). Structure, conduct and comparative statistics. Bell
Labotaries Economic Discussion Paper.
Panzar, J. and Rosse, J. (1987). Testing for” monopoly” equilibrium. The Journal of
Industrial Economics, XXXV(4) :443–456.
Peltzman, S. (1977). The gains and losses from industrial concentration. Journal of Law
and Economics, 20 :229–263.
40
Etudes et Documents n
o
12, CERDI , 2014
Perloff, J. M. and Shen, E. Z. (2012). Collinearity in linear structural models of market
power. Review of Industrial Organization, 40(2) :131–138.
Posner, R. A. (1979). The chicago school of antitrust analysis. University of Pennsylvania
Law Review, 127(4) :925–948.
Rajan, R. (1992). Insiders and outsiders : The choice between informed and arm’s-length
debt. The Journal of Finance, 47(4) :1367–1400.
Rezitis, A. N. (2010). Evaluating the state of competition of the greek banking industry.
Journal of International Financial Markets, Institutions & Money, 20 :68–90.
Robinson, J. (1933). The economics of imperfect competition. Macmillan New York.
Rosse, J. N. and Panzar, J. C. (1977). Chamberlin vs. Robinson : an empirical test for
monopoly rents. Bell Laboratories Economic Discussion Paper.
Schaeck, K. and Cih´ak, M. (2013). Competition, efficiency, and stability in banking.
Financial Management.
Schiersch, A. and Schmidt-Ehmcke, J. (2010). Empiricism meets theory : Is the boone-
indicator applicable ?
Sealey, C. and Lindley, J. (1977). Inputs, outputs, and a theory of production and cost
at depository financial institutions. The Journal of Finance, 32(4) :1251–1266.
Shaffer, S. (1982). A non-structural test for competition in financial markets. In Pro-
ceedings of a Conference on Bank Structure and Competition, Federal Reserve Bank of
Chicago, pages 225–243.
Shaffer, S. (1983). The rosse-panzar statistic and the lerner index in the short run.
Economics Letters, 11 :175–178.
Shaffer, S. (1989). Competition in the us banking industry. Economics Letters, 29(4) :321–
323.
Shaffer, S. (1993). A test of competition in canadian banking. Journal of Money, Credit
and Banking, 25(1) :49–61.
41
Etudes et Documents n
o
12, CERDI , 2014
Shaffer, S. (1999). The competitive impact of disclosure requirements in the credit card
industry. Journal of Regulatory Economics, 15(2) :183–198.
Shaffer, S. (2001). Banking conduct before the european single banking licence : A cross-
country comparison. North American Journal of Economics and Finance, 12 :79–104.
Shaffer, S. (2004a). Comments on ”what drives bank competition ? some international
evidence”. Journal of Money, Credit and Banking, 36(3) :585–592.
Shaffer, S. (2004b). Patterns of competition in banking. Journal of Economics and
Business, 56(4) :287–313.
Shaffer, S. and Spierdijk, L. (2013). Duopoly conduct and the panzar-rosse revenue test.
Working paper.
Sharpe, S. A. (1990). Asymmetric information, bank lending, and implicit contracts : A
stylized model of customer relationships. The Journal of Finance, 45(4) :1069–1087.
Smith, A. (1776). The wealth of nations. New York : Modern Library, 740.
Stigler, G. J. (1957). Perfect competition, historically contemplated. The Journal of
Political Economy, 65(1) :1–17.
Stiglitz, J. E. (1987). Competition and the number of firms in a market : Are duo-
polies more competitive than atomistic markets ? The Journal of Political Economy,
95(5) :1041–1061.
Stiglitz, J. E. (1989). Imperfect information in the product market. Handbook of industrial
organization, 1 :769–847.
Suominen, M. (1994). Measuring competition in banking : A two-product model. The
Scandinavian Journal of Economics, 96(1) :95–110.
Sutton, J. (1991). Sunk Costs and Market Structure : price competition, advertising and
the evolution of concentration. The MIT press.
42
Etudes et Documents n
o
12, CERDI , 2014
Tabak, B. M., Fazio, D. M., and Cajueiro, D. O. (2012). The relationship between banking
market competition and risk-taking : Do size and capitalization matter ? Journal of
Banking & Finance.
Tirole, J. (1988). The Theory of Industrial Organization : Jean Tirole. MIT press.
Turk Ariss, R. (2010). On the implications of market power in banking : Evidence from
developing countries. Journal of Banking & Finance, 34(4) :765–775.
Uchida, H. and Tsutsui, Y. (2005). Has competition in the japanese banking sector
improved ? Journal of Banking & Finance, 29(2) :419–439.
Van Leuvensteijn, M., Bikker, J. A., van Rixtel, A. A., and Sørensen, C. K. (2011). A
new approach to measuring competition in the loan markets of the euro area. Applied
Economics, 43(23) :3155–3167.
Van Leuvensteijn, M., Sørensen, C. K., Bikker, J. A., and Van Rixtel, A. A. (2013).
Impact of bank competition on the interest rate pass-through in the euro area. Applied
Economics, 45(11) :1359–1380.
Vesala, J. (1995). Testing for competition in banking : Behavioral evidence from Finland.
Bank of Finland Studies.
Vickers, J. (1995). Concepts of competition. Oxford Economic Papers, pages 1–23.
Vives, X. (2001). Oligopoly pricing : old ideas and new tools. The MIT press.
Vives, X. (2008). Innovation and competitive pressure. The Journal of Industrial Econo-
mics, 56(3) :419–469.
Weill, L. (2013). Bank competition in the eu : How has it evolved ? Journal of Interna-
tional Financial Markets, Institutions and Money.
Xu, B., van Rixtel, A., and van Leuvensteijn, M. (2013). Measuring bank competition in
china : A comparison of new versus conventional approaches applied to loan markets.
BIS Working Papers, 422 :1–68.
43
Table 1 – Synthesis of the major advantages and shortcomings of measures of competition
Method Advantages Shortcomings
Structural approach
Concentration measures (CR
k
, HHI) Low data requirement (available for almost all countries) Weaknesses in theoretical foundations
Does not require firm-level variables Require to define relevant market (geographical, products)
Banking sector contestability (regula-
tory and supervision framework)
Data available for a large number of countries (BCL database)
Does not require firm-level variables
Entry and exit in banking industry can be influenced by non-legal barriers
(technological and informational barriers), especially in developing countries
New Empirical Industrial Organization approach
Lerner index Bank-year specific measure of market power Weaknesses in theoretical foundations (Market power 6= Competition)
Requires a few number of observations Requires firm-level variables and information on prices
Simplicity and straightforward interpretation Can be fooled by spending in other activities and inefficiency
Flexible indicator
Conjectural variation (Iwata model
and Bresnahan-Lau model)
Test gives direct estimate of firm’s conduct that is continuous
and maps into all oligopoly solution concepts
Requires a large number of observations on firms including prices (system
estimation) Sensitive to market definition
Technique remains valid if firms do not maximize profit Limits to Lerner index apply
Panzar-Rosse model Can be estimated by simple, single-equation, linear model Identification can be an issue and some important assumptions have to be
Robust to the extent of market verified
Requires a limited number of observations Requires information on prices and monopsony power can mask monopoly
power
Boone indicator Strong theoretical foundations Efficiency should be one dimensional and observed
Can be estimated by simple, single-equation, linear model Different forms of competitive situation cannot be distinguished
Requires only a few number of observations Requires firm-level variables and information on prices
Continuous measure of competition Identification could be an issue in some cases (e.g. if firms compete in quality)
Considers (partially) non-price strategies
