Competition, foreign ownership and their relationship: an empirical
analysis of the banking market in Eastern Europe
Joep Hoveling
Student number 1423649
University of Groningen - Faculty of Economics International Economics and Business Department Supervisor Michael Koetter
Abstract
Introduction
The creation of the European Coal and Steel Community (ECSC) in 1952 by France, West-Germany, Italy, Belgium, Luxembourg and the Netherlands was the first step towards the European Union (EU) as it exists nowadays. The EU, under its current name and format, was established in 1992 with the Maastricht Treaty and currently consists of 27 member countries. The EU created a single market among its member countries. In order to realize one single market, many policies have been introduced. One of these policies is the ‘free movement of capital’. The creation of the First and Second Banking Directive resulted in the realization of this internal policy. One of the effects of the new Banking Directive is the possibility for banks to branch in other countries. Once a bank has the right to operate in one country, it now has a ‘passport’ to operate in all countries within the EU’s single market. This means an improvement in conditions for banks to compete with each other and to conduct (cross-border) takeovers.
In May 2004, Cyprus, the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Slovakia and Slovenia joined the EU. These countries are so-called ‘accessory countries’. The accessory countries substantially enlarged the EU. In the years before these countries joined the EU, each country transformed its economy from a plan-economy towards a market-plan-economy (except for Malta and Cyprus, who already were market-economies). However, each country made different choices with respect to its privatization process.
The future membership of the EU and the transition towards a market-economy made the accessory countries interesting new markets for (mostly) Western European banks. These banks were given opportunities to enter these markets. Each privatization path gave different possibilities concerning the time a foreign bank could enter the market, the easiness of entering the market, the amount of banks that were privatized and the possibilities of foreign banks to buy up those privatized banks. In each country the above mentioned possibilities took place, but in each case to a more or lesser extent.
effects observed in banking markets. The presence of foreign banks increases efficiency and improves resource allocation.1 It is often seen (in Latin America and Eastern Europe)
that foreign banks are believed to keep lending to firms in times of deteriorating economic conditions and are more independent from the local government compared to local banks. On the downside, local banks are forced to invest, which substantially increases costs, in order to be able to compete with international foreign banks. In some cases, foreign banks choose only to lend to stable (international) firms, leaving the local banks with the weaker firms, so-called ‘cherry picking’. These effects could lead to instability, especially when local banks go bankrupt. Empirical results point out that, in general, foreign banks improve efficiency and stability in the local banking market.
Claessens et al. (2001) find that foreign banks in developing countries have higher interest margins and higher profitability compared to domestic banks. High levels of foreign entry cause a reduction in profitability and margins for the local banks. The above mentioned effects are immediately felt by local banks upon entry of foreign banks.
These effects of foreign entry on a local market, the transition process of the accessory countries and their recent membership of the EU form the inspiration of this paper. This paper will analyze the development of foreign entry, the development of competition and their relationship by looking at the banking markets of the Eastern European entrants of 2004 to the EU, namely; the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia and Slovenia. My research question is “For the period 1997-2004,
how has the level of competition developed in the banking market in the accessory EU countries and what role is played by foreign banks in these specific markets?”
The first part of the research question is answered by estimating the level of competition in these markets according to the ‘Panzar and Rosse’-model (P-R model). An important aspect of this research is to look at the development of competition over time and therefore the P-R model is adapted. This makes it possible to see whether the level of competition increased or decreased over time. The result is a yearly statistic on the level of competition instead of one statistic for a certain period of time (like seen in other
1 This can be achieved in several ways; (1) the introduction of superior management practices and
studies). A dataset is created from bank-level data for the period 1997-2004. The dataset includes information on the ultimate owner of a bank (either foreign-owned or domestic) in order to answer the second part of the research question concerning the role of foreign banks. The second important aspect of this paper is to analyze the development of competition over time in combination with the development of foreign presence in the market. And to address questions like: What is the effect of foreign banks on competition? What is the effect of foreign banks on the amount of revenue earned? And is a foreign bank better at generating revenue than a domestic bank?
1. Literature review
General background information
Consolidation in the banking market is seen all over the world to a more or lesser extent. Bikker and Bos (2004) judge the European banking market as a market where “competition increases, resulting in lower profit margins, higher cost efficiency and lower profit efficiency” and they also expect cost reductions. Although a large amount of mergers and acquisitions is taking place within the countries itself, cross-border mergers and acquisitions are occurring increasingly. Especially banks from developed countries take advantage of the changing conditions in developing countries, like in Latin America and Eastern Europe, to internationalize. This implies that the wave of consolidation and the increasing presence of foreign ownership (more on this later on) go hand in hand. This results in an increasing amount of banks that become international players and operate in a wide range of important international markets.
According to Boot (1999) the influence of politics handicaps free competition. The banking industry is related to politics due to the fear of financial instability and national pride. There is protectionism of big national banks, who are not supposed to fall in the hands of foreign owners. As a result, consolidation in the markets with government interference lags behind compared to countries where there is no or little government interference (Bikker and Wesseling, 2003). And in countries where a strong national bank has been established there is concern about competition or rather a lack of competition. At the moment, the result is differences among countries with respect to competition, concentration and profit margins.
influence of governments. By privatizing national banks the governments positively influenced the consolidation process and competition, instead of playing a protectionist role. Many of these big national banks are now owned by foreign banks. The foreign banks were (and are) the driving factor behind the development of the private banking sector.
The transition economies are now characterized by a bank-based system. During the time of reform the banking market transformed, but the capital market did not transform, this explains the present existence of the bank-based system in the transition economies. The result is an underdeveloped capital market, where companies remain dependent on banks for capital loans.2 This is important to keep in mind, because interest revenue by loans
plays an important role in the P-R model.
Staikouras and Koutsomanoli-Fillipaki (2006) provide some statistics which confirm the general developments in the accessory countries. Privatization is under 5% in all countries by 2004, except for Poland (27%), Slovenia (13%) and Hungary (11%). On average, nearly 70% of the banking assets in the market are owned by foreign banks by 2005 and this percentage is likely to increase in the upcoming years due to further implementations of the regulations of the ‘EU-passport’. The total number of banks has decreased by 17% in the period between 1998 and 2002. And the concentration ratios (the market share of the five biggest banks in the market) are on average 72%, however this average ranges between 55% and 99%. These remarkable statistics reflect the highly concentrated markets of Eastern Europe. Although, according to Yildirim and Philippatos (2003), it has not led to anti-competitive behavior so far. One exception to these developments is Slovenia. Slovenia does not have any foreign banks in the market with a substantial market share. Even though the level of internationalization is low, the Slovenian banking market is performing well when compared to the international level. Related P-R research articles
Many researchers have examined the level of competition in different banking markets by making use of the Panzar-Rosse model. An extensive overview of studies which apply the
P-R model to the banking industry is provided in Appendix 1. A few of those articles relevant to this paper will now be discussed.
The articles by Gelos and Roldos (2002), Yildirim and Philippatos (2003), Drakos and Konstantinou (2005) and Staikouras and Koutsomanoli-Fillipaki (2006) analyze all or some of the countries under research in this paper. These authors selected their group of countries with different intensions and appropriate specific timeframe compared to this research. Gelos and Roldos (2002) analyzed eight Latin American and Eastern European countries (including the Czech Republic, Hungary and Poland) for the period 1994-1999. They conclude that although concentration increased, the markets have not become less competitive. Monopolistic competition is found in all countries, except for Argentina and Hungary where perfect competition can not be ruled out as a possibility.
Yildirim and Philippatos (2003) analyzed the Central and Eastern European countries (including all accessory countries) for the period 1993-2000. All markets are characterized by monopolistic competition, except for Macedonia and Slovakia where monopoly power can not be ruled out. The authors underline the importance of foreign banks for the future of the Eastern European countries.
For the period of 1992-2000, Drakos and Konstantinou (2005) find monopolistic competition in all markets of 10 Central and Eastern European countries (including all accessory countries, except Slovenia). The exception is Latvia, where a monopoly can not be ruled out.
The most recent study is done by Staikouras and Koutsomanoli-Fillipaki in 2006. The countries under research are all countries in the EU for the period 1998-2002. The sample is split up in the 15 original EU-countries and the ten ‘accessory countries’. The level of competition is higher in the accessory countries compared to the level of competition in the original EU-countries. However, this is not the case for the fee market where a higher level of competition is seen in the original EU-countries.3 The fee market in the accessory
countries is not yet (fully) developed, explaining the lower level of competition in the fee market.
A second type of studies, like the ones by Claessens and Laeven (2003) and Yeyati and Micco (2003), takes a closer look at foreign influence in relation to the level of competition. In their extensive study, Claessens and Laeven (2003) state that looking at concentration indexes is not enough to measure competition, the possibility of new entrants and other threats to the market also influence (the level of) competition.4 The
research includes 50 countries (including the Czech Republic, Hungary, Latvia and Poland) for the period 1994 to 2001. They find that low barriers to entry, and especially the role and influence of foreign banks, as well as the lack of activity restrictions increase the contestability of the market. Countries with those conditions in place have a competitive market. All markets under research by Claessens and Laeven are characterized by monopolistic competition. This research is similar to Claessens and Laeven in that it acknowledges the relationship between the level of competition and foreign ownership. But at the same time it differs because it looks at the development over time, gives a continuous development of that relationship and does that for a specific group of countries and for a specific time-frame.
The article by Yeyati and Micco (2003) analyzes eight Latin American countries for the period 1993-2002. The authors observe big changes in the banking industry in Latin America and one of the results is a strong influence and presence of foreign banks. Disregarding the different underlying foundations of these changes, the situation in Latin America shows similarities to Eastern Europe. Yeyati and Micco analyze the influence of changes in competition on concentration and on foreign participation. The authors test this by calculating the level of competition, according to the P-R model, and compare that with CR and Herfindahl-Hirschman Index (HHI) for concentration and with FASSETS (a ratio of the foreign assets as part of the total assets in the market) for foreign participation. The main finding of the paper is that concentration did not reduce competition, but higher levels of foreign ownership did lead to lower levels of competition.
A third type of research is the research by Bikker and Haaf (2002a), who analyze changes in the level of competition over time. For that purpose, Bikker and Haaf made their
4 The theory behind this reasoning is the Contestable Market Theory (CMT) developed Baumol (1982), this
model time-dependent. They take a year in the beginning and a year at the end of their timeframe and test whether the level of competition changed. This approach is different from the approach used in this research which estimates the level of competition on a yearly basis. Bikker and Haaf found that in 50% of the time the level of competition did not change, but in the case that the level of competition did change it increased in 80% of these cases.
The above mentioned points of interest form the positioning of my paper in the academic field of research. It has its own selection on a group of countries, namely the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia and Slovenia. These are all countries that joined the EU in 2004 and are all transition economies. And it will have its own timeframe, starting in 1997 till 2004. This timeframe captures the years up until those countries joined the EU, years in which many changes took place, which makes those years interesting to analyze. Unfortunately this is also a limitation of the research, because it does not look at the years following the EU membership, since it is likely that these years are also interesting to research.
Many other studies analyzed the relationship between concentration indices and the level of competition. My paper will not do that because of the inconclusive outcomes as pointed out by Jansen and De Haan (2003). Instead the focus (and limitation) of this research is on the development of competition over time and on the role and influence of foreign banks. The P-R model will be made time-dependent to get a yearly outcome on the level of competition in order to observe the changes over time. And my model will be able to distinct between domestic owned and foreign owned to observe the effects on competition and the changes between these groups.
2. Methodology
Theoretical considerations
Industrial organization is the underlying theory in the banking literature. It deals with strategic behavior of firms, the structure of the market and the interaction between these two aspects. Oligopoly theory is part of the industrial organization theory and deals with imperfect competition. In oligopoly theory there are few players in the market. This high concentration causes the firms in the market to be aware of the actions of other firms, there is interaction in the market, and every action causes a reaction by the other players. High concentration increases the possibility of collusion among the firms in the market and that results in above normal profits earned by these firms. The market can either be characterized by quantity-setting (Cournot competition) or price-setting (Bertrand competition).5
From industrial organization theory follow two structural models most often used in the banking literature: the Structure-Conduct-Performance (SCP) paradigm and the Efficiency Structure Hypothesis (ESH).6 The SCP paradigm was originally developed by
Mason (1939) and Bain (1951). The SCP paradigm forms a direct link between the structure (concentration) and the conduct (behavior) of firms, and between the conduct of firms and the performance (profits) of the market. The reasoning of the SCP paradigm is that a concentrated market results in market power for the firms in the market. The market power will be exploited resulting in above normal profits for these firms.
The other model is the ESH and was first developed by Demsetz (1973) and Peltzman (1977) and can be seen as a criticism on the SCP paradigm. The ESH sees efficiency as the cause of market concentration. An efficient firm has an advantage over other firms and can exploit this by setting lower prices for its products, thereby gaining market share and profits. Only other efficient firms will be able to stay in the market, causing the
5 With respect to the type of competition, Chun and Kim (2004) place a point of criticism. The P-R model is
built on the assumption that banks are price takers (meaning Cournot competition) with respect to their input price, but this is not always the case in reality.
concentration in the market. The remaining firms will keep each other efficient in order to make a profit.
One non-structural model often discussed is the Contestable Market Theory (CMT) developed Baumol (1982).7 CMT explains why firms in a market can still behave
competitive, even when concentration is (very) high. The reasoning is that this highly concentrated market is faced with a high possibility of new entrants in the market. The current players in the market will act as if they were operating under perfect competition, because if they earn above zero economic profit another player will enter the market. The optimal conditions for a player to enter the market are low entry barriers and no exit costs.
The Panzar and Rosse model relates to the above mentioned theories, but has no one-to-one basis with any of them. It is a non-structural model which tests the relationship between competition and the use of market power. In a research article by Bikker and Bos (2004), the authors develop a framework including several structural and non-structural models (including the P-R model). The framework analyzes a profit maximizing bank on the basis of competition and efficiency. They conclude with respect to the P-R model in comparison to the overall framework; “The P-R model may have picked up the effects of internationalization that are less apparent in other measures”. The authors are not conclusive of why the P-R model in the end turns out to give the best results.
Panzar-Rosse model
Panzar and Rosse (1987) initially came up with a monopoly-based model, which they later altered to make it applicable for monopolistic competition and perfect competition as well. The analysis included price-taking firms in long-run equilibrium with a homogenous cost function and behaving competitive to maximize their profits. The model calculates the degree of competition in the market; this is done by calculating elasticities.
7 “The non-structural models (…) were developed in reaction to the theoretical and empirical deficiencies
The basic model consists of; Equation ( 1 )
ln(P) = α + β1 ln(W1) + β2 ln(W2) + β3 ln(W3)
where;
ln = natural logarithm W1 = unit input price 1
P = total revenue W2 = unit input price 2
W3 = unit input price 3
The elasticities calculated are β1, β2 and β3, which are elasticities of input prices (W1, W2
and W3) with respect to output revenue (P). The interpretation of an elasticity is the
reaction of a firm in the occurrence of a change in an input price. Or more clearly; the percentage change in total revenue caused by one percentage increase in input prices. The outcome of the P-R model is the H-statistic. The H-statistic is calculated by adding up the elasticities β1, β2 and β3 (H = β1 + β2 + β3). The H-statistic states the degree of
competition in the market and thereby also the type of market structure. The H-statistic can have several outcomes (for an overview see Table 1). An outcome of the H-statistic equal to one means that when the input price goes up a firm passes on the entire increase in price to the consumer; this is a characteristic feature of perfect competition, because the firm makes zero economic profit and has no market power.
Shaffer (1982) describes two other interpretations when the H-statistic is also equal to one; the natural monopoly in a perfect contestable market and the sales maximizing firm subject to breakeven constraint.
The market is characterized by a monopoly when the H-statistic has a value of zero or below zero, this is proven by Panzar and Rosse (1987). The reasoning is based on economic intuition that a monopolist will not respond or respond in the opposite direction to a change in input prices, thus an increase in input prices results in the equilibrium output to be reduced and that causes lower total revenues.8
In the case the outcome of the H-statistic is between zero and one, the firms in the market exercise some of their market power, but the fear of new entrants in the market also discourages them from taking full advantage of the change in input prices. How big the fear of new entrants is and to what extend they can exercise their market power is reflected in the value of the H-statistic. Therefore an outcome, which lies between zero and one and is close to one, states that the market is very competitive and approaches perfect competition. On the other hand, an outcome between zero and one and closer to zero, states that the market is not very competitive and approaches a monopoly. In his article, Vesala (1995) proves that a continuous interpretation of the H-statistic is appropriate. Before Vesala proved this point, the interpretation of the H-statistic was considered a weak point of the model, because monopolistic competition was often the outcome and no further conclusions could be drawn from it. Now, this is no longer the case, the continuous interpretation of the H-statistic has become generally accepted and applied throughout the academic field (by Bikker and Haaf (2002a), Gelos and Roldos (2002), Claessens and Laeven (2003), Yildirim and Philippatos (2003) and Staikouras and Koutsomanoli-Fillipaki (2006)).
Table 1:
Source: Yildirim and Philippatos (2003)
Application to the banking market
assumption has to be made concerning the production process of a bank in order to make it comparable to a ‘normal’ (read: production) firm. The literature describes two approaches, the ‘production approach’ and the ‘intermediation approach’.9
The intermediation approach is chosen as the appropriate approach, it views the bank as a single product firm and the inputs of the intermediation approach (funds, labor and capital) resemble the cost function of a normal firm.10 The production of a bank is the
amount of loans (measured in money). In order to produce that loan a bank has costs, these costs consist of a deposit or other funds, more specifically the interest paid by the banks over that deposit or other funds, plus the costs of labor and physical capital. These three items make up the cost function of a bank according to the intermediation approach. The following equation will be estimated to run on a panel data set for banks:
Equation ( 2 )
ln(P it) = α + β1 ln(W1,it) + β2 ln(W21,it) + β3 ln(W3,it) +
+ γ1 ln(Y1,it) + γ2 ln(Y2,it) + γ3 ln(Y3,it) + γ4 ln(Y4,it) + δ1E + δ2 F + ε it
where P it is the total interest revenue. Total interest revenue as the dependent variable is
strongly advised by Bikker, Spierdijk and Finnie (2006). In their article Bikker, Spierdijk and Finnie address some improvements in the calculations of the Panzar-Rosse model. The form of the dependent variable which is most often used in the literature is the ratio of total interest revenue to total assets. Using this form of dependent variable will bias the outcome of the H-statistic towards one. But by using total interest revenue as the dependent variable, the H-statistic will reflect the actual level of competition in the market more appropriately, according to Bikker, Spierdijk and Finnie.
W1,it are interest expenses to total funds (unit price of funds), W21,it are personnel expenses
to total assets (unit price of labor)11, W
3,it are other (non-interest) expenses to fixed assets
9 See Chun and Kim (2004)
10 The production approach is considered less appropriate because it does not include interest payments in
the cost function. A second reason is that the production approach does not look at the amount of loans but at the number of loans produced by a bank. Therefore the intermediation approach is chosen over the production approach.
11 An alternative calculation of the unit price of labor that could have been used is the ratio of personnel
(unit price of capital). These costs (unit price of funds, labor and capital) are similar to the inputs (costs) described earlier by the intermediation approach. The subscripts ‘i’ and ‘t’ denote bank i and year t respectively. The variable F represents time dummies.12
The variable E is the most important control variable in light of the research question. The variable E represents a dummy variable which indicates the ownership of the bank, which can either be foreign owned or domestic. E is used to answer an important part of the research question; it tests the relationship between the level of competition and foreign ownership and also the relationship between interest revenue and foreign ownership. Foreign ownership is expected to have a positive relationship with a bank’s interest revenues. This means that higher interest revenues are expected when the presence of foreign ownership increases. The research by Claessens et al. (2001) confirms the predicted positive sign of E, as mentioned earlier: “foreign banks in developing countries have higher interest margins, profitability and tax payments than domestic banks”, and also: “high levels of foreign entry cause a reduction in profitability and margins for the local banks”. To that extent the expected sign of E is positive.13
Several control variables (Y1,it Y2,it Y3,it Y4,it) are included at individual bank level. The
control variables take into account differences in the business mix among banks which influence their costs and revenues, thereby improving the (relevance and explaining power of) model. The choice for these control variables is based on reviewing the literature, the good results of these control variables in other studies and selecting only the control variables appropriate to this paper.
Y1,it are equity to total assets (as used in Nathan and Neave, 1989; Bikker and
Groeneveld, 2000; Bikker and Haaf, 2002b; Yildirim and Philippatos, 2003; Yeyati and is not always available for the whole data sample and therefore not chosen.
12 Equation 2 does not include country dummies. That is because this research focuses on differences
among banks and not among countries. This results in H-statistics which indicate the difference between domestic owned banks and foreign owned banks. Besides, this dataset also does not fulfill the minimum amount of observations per country in some cases. Including a country dummy in this equation results in the country dummy to be excluded (dropped) from the equation.
13 In a specific calculations the control variable E is replaced by Y6 (the ratio of foreign firms with respect
Micco, 2003; Staikouras and Koutsomanoli-Fillipaki, 2006 etc.). Y1,it captures the risk
banks take when lending out money. It is difficult to state a ‘good’ capitalization ratio. Banks that take more risk and take an aggressive approach can also expect higher interest revenues, but being too aggressive comes at the expense of interest revenues. The same goes for banks that are too passive, since they fail to earn interest revenues while they could. And when lending out money the amount of risk associated with that loan needs to be taken into account (the clients and the country characteristics). This means that Y1,it is
an inverted U-shape. Therefore it is not possible to predict the sign of Y1,it.
Y2,it are loans to total assets (as used in De Bandt and Davis, 2000; Bikker and
Groeneveld, 2000; Bikker and Haaf, 2002b; Gelos and Roldos, 2002; Yildirim and Philippatos, 2003; Yeyati and Micco, 2003; Staikouras and Koutsomanoli-Fillipaki, 2006 etc.) For Y2,it a positive sign is expected, because the more loans on a bank’s balance
sheet, the more interest revenue is to be expected.
Y3,it are (inter)bank deposits to deposits and market funding (to check for the deposit mix
of the bank, used in De Bandt and Davis, 2000; Bikker and Haaf, 2002b; Gelos and Roldos, 2002; Yildirim and Philippatos, 2003; Yeyati and Micco, 2003; Staikouras and Koutsomanoli-Fillipaki, 2006 etc.) The (inter)bank deposit rate is an expensive way of financing, because a penalty (in the form of a high interest rate) is paid for making use of these kind of funds. When banks make use of this kind of funding it has a negative effect on the revenues earned, this results in an expected negative sign for Y3,it.
Y4,it are other income to total assets (as used in Bikker and Groeneveld, 2000; Bikker and
Haaf, 2002b; Yeyati and Micco, 2003; Staikouras and Koutsomanoli-Fillipaki, 2006 etc.). The time and effort going in to generating other income comes at the expense of the interest revenue, therefore the expected sign for Y4,it is negative.
Alternatives of P it
P it is also tested. The alternative P it is total revenue and it consists of interest revenue
plus other operating income plus other income. The creation of this alternative P it is due
to changing conditions in the market, which asked for an alteration of a model. The P-R model is based on interest revenue, but the role of the interest revenue is diminishing and this affects the interpretation and importance of the outcome of the model. Researching the alternative P it could yield additional insights on the overall degree of competition in
the market and to check the claim that the importance of interest revenue is decreasing.14
Methodology
The data for my research is panel data, since the data is collected for a group of banks (cross-section) over time (time series). Therefore it is possible to observe differences among banks and the effect of time (the differences observed over time in one bank). The P-R model is a model that has been made time dependent and can distinguish between banks. Panel data has a lot of advantages15, like controlling for heterogeneity bias or the
confounding effects of omitted variables that are stable over time. Omitted variable bias is “a problem that arises when there is some unknown variable or variables that cannot be controlled for that affect the dependent variable”.16
The P-R model is calculated using the fixed effects (FE) model.17 By creating a time
dummy (TD) for each year the FE model is able to observe differences between these years. Introducing an interaction term to the regression between the independent variables and the time dummies makes is possible to observe the effects over time. For another calculation a second interaction term is introduced between the independent variables and E in order to observe the difference in the level of competition between foreign banks and domestic banks.18 These interaction terms are very suitable to answer the main research
question. The specific equations will be provided in the result section.
14 To get a complete and consistent view on the alternative P it, it is used in each calculation. 15 See Yildirim and Philippatos (2003) and Chun and Kim (2004).
3. Data
The firm level data on banks from the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia and Slovenia is obtained from BankScope. BankScope provides detailed information of bank’s financial statements from the period 1990 till 2004. An advantage of firm-level data is that individual preferences can be taken into account as well as group preferences (Cournot theory). Not all banks that are active in these markets are present in the dataset of BankScope.19
The type of data used for this research is unconsolidated data. The choice for unconsolidated statements is a result of the model. The model checks for foreign ownership, an unconsolidated statement means that the subsidiaries from (big) international banks report their own statement, even though they are a part of a bigger bank conglomerate (foreign ownership).20
The initial dataset consisted of 1342 observation (on average 170 banks per year) for the period 1997-2004. After cleaning the data, the final dataset, which is used to perform the calculation on, consisted of 964 observations (on average 120 banks per year).21 How the
observations per country are divided can be seen in graph 1.
The type of banks which are represented in the data are classified as commercial bank, cooperative banks or saving banks.22 From the total number of banks, 4% is classified as
a cooperative bank and 1% is classified as a savings banks.23 This means that the type of
group classified as a commercial bank is far out the biggest group. This also means other types of banks are excluded from the sample, like central banks, mortgage banks,
19 This remark needs to be taken into account when interpreting the results of this research. In terms of size
of the dataset and the credibility of BankScope as an often used source for data, we hope that the banks present in the dataset reflect a representative sample of the banks from these countries and will lead to representative results. However, there will always remain some discrepancies and inconsistencies with international data and this can especially be expected from these countries.
20 When a consolidated statement would have been used only independent banks or banks that are
headquartered in that country would have been reported in the sample.
21 Cleaning means that the dataset has undergone several selections, such as; all observations which lacked
a main variable were deleted, as well as an outlier selection on the remaining data. The outlier selection closely resembles the 1st and 99th percentage of the dataset and is based on the outlier selection with upper
and lower bounds as used by Bikker, Spierdijk and Finnie (2006). For more information, see Appendix 3.
22 These kind of banks play a central role in the financial system.
23 Even tough these percentages are small, we see no reason to exclude them, due to their central role in the
investment banks, bank holdings and other types of specialized banks. These banks are excluded because their role is too specialized in one field of the financial system, this makes their financial statements unfit for this type of research.
Graph 1:
Observations per country 11% 4% 9% 18% 9% 27% 12% 10% # obs Czech Rep.
4. Empirical results
Main resultsThis section discusses the results of two main equations in this research. The first equation (3) includes one interaction term (between the time dummies and the independent variables). The second equation (4) includes two interaction terms (the second interaction term is between foreign ownership and the independent variables). Equation 3 is estimated by using the multiple regression methodology, in the form as described earlier (a FE-model with TD and an interaction term). The calculation is performed using two different dependent variables (P, it ): the original dependent variable
(P11, total interest revenue) and the alternative dependent variable (P21, total revenue). Appendix 4 shows the outcome of equation 3 with both P11 and P21.24
Equation ( 3 )
ln(P, it) = α + β1 ln(W1,it) + β2 ln(W21,it) + β3 ln(W3,it) + γ1 ln(Y1,it) + γ2 ln(Y2,it) +
+ γ3 ln(Y3,it) + γ4 ln(Y4,it) + δ1 E + δ2 F + β4 [F ln(W*,it)] + ε it
The independent variables in both regressions are all significant at the 1% level. The unit price of funds (W1) has a strong positive relationship with the dependent variable and this means that when the costs of interest go up, (interest) revenue also goes up. The unit price of labor (W21) has a strong negative relationship with the dependent variable; this was also seen in research by Molyneux et al. (1994) as well as De Bandt and Davis (2000). The negative relationship of unit price of labor with the dependent variable could suggest that when more personnel is hired it does not automatically raise (interest) revenues or it could suggest that smaller banks are better able at generating (interest) revenues. The last independent variable (W3, the unit price of capital) shows a positive
24 Equation 3 is tested for autocorrelation (by running the ‘xtserial’ command in Stata: xtserial lnP11
$HStatistic $Interaction $Year lnY1 lnY2 lnY3 lnY4 E). The result is: Wooldridge test for autocorrelation in panel data
relationship with the dependent variable. This means that when other (non-interest) expenses go up, (interest) revenue also goes up.
The sign of dummy variable E shows a positive relationship with respect to the dependent variable, this supports the prediction that foreign ownership implies higher revenues. However, the dummy variable E is not statistically significant in both regressions. But E is kept in the regression because of its important role in this research (and because in the remaining calculated equations E will be statistically significant). Both tests reveal a good overall fit of the model (with an R-squared of 55 and 58% respectively).
Looking at the control variables (Y1, Y2, Y3 and Y4), it can be concluded that Y1 and Y2 are statistically significant at the 1% level and Y3 and Y4 are not statistically significant in both regressions. The first control variable (Y1) is concerned with the capitalization ratio of a bank, the expected sign was negative and the outcome, as shown in Appendix 4, confirms this prediction. Molyneux et al. (1994) and Bikker and Groeneveld (2000) also predicted a negative sign for Y1, because ‘less equity implies more leverage and hence more interest income’. This negative relationship is confirmed in other research (like Yildirim and Philippatos, 2003 and Staikouras and Koutsomanoli-Fillipaki, 2006), this means that banks who take more risk earn more revenues.
The second control variable (Y2) deals with the ratio of loans to total assets, the outcome reveals a positive relationship with the dependent variable, this is also in line with the prediction made earlier. The positive relationship between Y2 and total (interest) revenue means that more loans implies more (interest) revenue. The predicted sign is confirmed in other research (like Yildirim and Philippatos, 2003).
The deposit mix of a bank is the third control variable (Y3), besides the fact that the variable is statistically not significant, the signs are also mixed, a positive sign is seen in the first regression and a negative sign is seen in the second regression. This means no statistical valid conclusions can be drawn from Y3.
The time dummy variable F are year dummies25 and “F ln(W*
,it)” represents the (first)
interaction term where each year dummy (F) is multiplied by the three independent variables (W1, W21 and W3).26 A Wald-test has been performed in order to test if the
group of interaction terms (F9lnW1 – F15lnW3) has a statistically significant relationship with the dependent variable. The outcome of the test reveals that the group as a whole has a statistically significant relationship with the dependent variable.
The coefficients β1, β2 and β3 form the H-statistic. In order to make the H-statistic vary
over time, each yearly H-statistic is calculated by summing up the coefficients β1, β2 and
β3 as well as the coefficients of the interaction terms for that year. For example, the year
2000 is calculated by adding up β1 + β2 + β3 + β4 (of F11lnW1) + β4 (of F11lnW21) + β4
(of F11lnW3). The outcome of the H-statistics over the years shows a clear downward trend; in the early years the statistic was positive, but in the more recent years the H-statistic became negative. A negative H-H-statistic is not seen in many other studies, but it is seen in Bikker, Spierdijk and Finnie (2006) for a couple of countries. They argue that the H-statistic is biased towards one (overrated) due to misspecifications (as explained earlier). That is why a negative H-statistic has not been seen often before.27 The trend of
the H-statistic can be viewed in table 2 and graph 2; both regressions (P11 and P21) clearly show the same pattern. Only one conclusion is possible, which is that competition has lowered over the years in the direction of a monopoly. For the years 2001-2004 a monopoly can not be ruled out and the value of the H-statistic even becomes negative in the year 2002 and stays negative in the following years. Although the year 2002 could be considered as an outlier (a dip in the trend of graph 2), the trend is clearly going downwards.
25The results in the appendices show the dummy variable F, each number following the F stands for a year;
F9 = 1998, F10 = 1999, F11 = 2000, F12 = 2001, F13 = 2002, F14 = 2003, F15 = 2004 and the base year is 1997.
26 Example: F9lnW1 = 1998 x lnW1.
Table 2:
P11 FE int. Term P21 FE int. Term
Year H-statistic Year H-statistic
1997 0.36 ¹ 1997 0.26 ¹ 1998 0.37 ¹ 1998 0.29 ¹ 1999 0.35 ¹ 1999 0.27 ¹ 2000 0.25 ¹ 2000 0.17 ¹ 2001 0.18 ² 2001 0.11 ² 2002 -0.17 ² 2002 -0.28 ² 2003 -0.06 ² 2003 -0.16 ² 2004 -0.10 ² 2004 -0.13 ²
¹ H = 0 and H = 1 rejected (level of confidence 95%). ² H = 0 not rejected (level of confidence 95%).
Graph 2: Development H-statistic -0.40 -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 0.40 0.50 1997 1998 1999 2000 2001 2002 2003 2004 Year V al ue H -s ta ti st ic P11 P21
-0.15 and -0.23 respectively).28 Another interesting point to mention is that the control
variable E is statistically significant at the 1% level and shows the expected positive sign.29
Table 3
Period H-statistic Period H-statistic
P11 P21
1997-2000 0.24 ¹ 1997-2000 0.15 ¹
2001 - 2004 -0.15 ² 2001 - 2004 -0.23 ²
¹ H = 0 and H = 1 rejected (level of confidence 95%). ² H = 0 not rejected (level of confidence 95%).
The relationship between foreign ownership and (the level of) competition can be tested in a more direct way.30 The second main equation makes this possible by introducing a
second interaction term (consisting of E and W1, W21 and W3). The second interaction term is added to equation 3. The result is equation 4, which gives an H-statistic that is split up in two lines and still varies over time as well. The two lines represent the difference between domestic ownership and foreign ownership.
Equation ( 4 )
ln(P, it) = α + β1 ln(W1,it) + β2 ln(W21,it) + β3 ln(W3,it) + γ1 ln(Y1,it) + γ2 ln(Y2,it) +
+ γ3 ln(Y3,it) + γ4 ln(Y4,it) + δ1 E + δ2 F + β4 [F ln(W*,it)] + β5 [E ln(W*,it)] + ε it
Tables 4 and 5 and graphs 4 and 5 show the result of equation 4.31 The main calculation
(Appendix 4) and this calculation are very similar to each other. A difference is that the dummy variable E has become negative, which is strange especially when adding up the coefficients of the interaction terms of E (β5 of ElnW1, β5 of ElnW21 and β5 of ElnW3) is
positive (+0.08). The positive summation of these interaction terms means that the effect of W1, W21 and W3 on the dependant variable differs between domestic owned banks
28 See table 3. 29 See Appendix 5.
30 So far the dummy variable E has been tested with respect to the dependent variable; by including E in the
second interaction term it becomes possible to test the effect of E on the level of competition (the H-statistic) in a more direct way.
and foreign owned banks. This could indicate that E has a positive relationship with (the level of) competition. So when the number of foreign banks in the market increases, the level of competition increases and this is in line with the predicted sign of E.
Table 4:
P11 FE E int. Term E=0 P11 FE E int. Term E=1
Year H-statistic Year H-statistic
1997 0.32 ¹ 1997 0.40 ¹ 1998 0.32 ¹ 1998 0.39 ¹ 1999 0.32 ¹ 1999 0.39 ¹ 2000 0.19 ¹ 2000 0.27 ¹ 2001 0.10 ² 2001 0.17 ² 2002 -0.23 ² 2002 -0.15 ² 2003 -0.11 ² 2003 -0.03 ² 2004 -0.16 ² 2004 -0.09 ²
¹ H = 0 and H = 1 rejected (level of confidence 95%). ² H = 0 not rejected (level of confidence 95%).
Table 5:
P21 FE E int. Term E=0 P21 FE E int. Term E=1
Year H-statistic Year H-statistic
1997 0.22 ¹ 1997 0.30 ¹ 1998 0.23 ¹ 1998 0.31 ¹ 1999 0.24 ¹ 1999 0.31 ¹ 2000 0.11 ² 2000 0.19 ¹ 2001 0.03 ² 2001 0.10 ² 2002 -0.33 ² 2002 -0.26 ² 2003 -0.21 ² 2003 -0.13 ² 2004 -0.20 ² 2004 -0.12 ²
Graph 4: Development H-statistic -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 0.40 0.50 1997 1998 1999 2000 2001 2002 2003 2004 Year V al ue H -s ta ti si ti c P11 E=0 P11 E=1 Graph 5: Development H-statistic -0.40 -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 0.40 1997 1998 1999 2000 2001 2002 2003 2004 Year V al ue H -s ta ti st ic P21 E=0 P21 E=1
The equilibrium test
The P-R model assumes that the firms in the market are operating in a long-run equilibrium. This is a requirement in order to be able to draw statistical valid conclusions.32 The equilibrium-test tests if this assumption holds. The economic
32 “If the sample is not in long-run equilibrium Shaffer (1985) states that negative values of the H-statistic
reasoning behind the equilibrium test is that in a situation of long run equilibrium the capital market will adjust for any risk among banks because the market is competitive, ‘thereby the ROA should not be correlated statistically with the input prices’.33 To
increase the robustness of this check, the equilibrium test is performed using two forms of return of assets (ROA); a return on assets ‘pre tax’ and a return on assets ‘post tax’. The test measures the sum of elasticity of the input prices with respect to ROA. The equation of the equilibrium-test is as follows:
Equation ( 5 )
ln(ROA it) = α + β1 ln(W1,it) + β2 ln(W21,it) + β3 ln(W3,it) +
+ γ1 ln(Y1,it) + γ2 ln(Y2,it) + γ3 ln(Y3,it) + γ4 ln(Y4,it) + δE + ε it
The outcome of the test indicates whether the sample is in equilibrium or in disequilibrium. The ‘E-statistic’, short for Equilibrium-statistic, is made up of the sum of elasticities β1, β2 and β3. An outcome of the E-statistic where E=0 indicates equilibrium
and means that ROA is not correlated with the input prices. An outcome below zero (E<0) indicates disequilibrium.
Performing the equilibrium test (in both forms), by running equation 5, reveals that the test is performed under conditions of equilibrium. This is proven by a Wald-test, the Wald-test does not reject the hypothesis of equilibrium (HO: β1 + β2 + β3 = 0).34
The development of foreign ownership
At the same time that the level of competition has decreased in the market, the presence of foreign banks has increased. The ownership dummy E is used to calculate the percentage level of foreign presence in the market (based on this dataset), both measured in numbers of banks owned by foreign banks (variable Y6) as well as assets owned by foreign banks (variable Y7). Table 5 and graph 6 show this development, which reveals a substantial increase of foreign presence in the dataset.
Table 5:
Year Y6 Y7 1997 0.34 0.32 1998 0.43 0.44 1999 0.45 0.49 2000 0.54 0.50 2001 0.55 0.52 2002 0.57 0.60 2003 0.59 0.62 2004 0.68 0.69 Graph 6: Development of Y6 & Y7 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 1997 1998 1999 2000 2001 2002 2003 2004 Year % Y6 Y7
Appendix 8 shows three pooled OLS regressions, one regression with dummy variable E and two regressions where E is replaced by Y6 and Y7 respectively. An OLS regression is different from the FE-model and this results in different H-statistics. Therefore we are not interested in the H-statistics but we look at the sign of E, Y6 and Y7 with respect to total (interest) revenue.
Equation ( 6 )
ln(P, it) = α + β1 ln(W1,it) + β2 ln(W21,it) + β3 ln(W3,it) + γ1 ln(Y1,it) + γ2 ln(Y2,it) +
+ γ3 ln(Y3,it) + γ4 ln(Y4,it) + δ1 E (or Y6 or Y7)+ ε it
The control variable E, Y6 and Y7 are all significant at the 1% level and have a strong positive relationship with the dependent variable. This means that when the presence of foreign banks in the market (both in numbers and assets) goes up (interest) revenues also go up. The values of the H-statistic range from 0.62 –0.41.35
5. Conclusions
This paper has examined the situation of the banking sector in the accessory countries that joined the EU in 2004. The Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia and Slovenia have undergone major changes in recent years; changing from a plan-economy to a market-economy as well as meeting all the requirements of EU legislation. And also foreign banks who saw and were given opportunities to enter these markets, changed but above all helped develop these markets. The time period 1997 – 2004 is examined by making use of the ‘Panzar-Rosse’-model. Panel data has been used to run a multiple regression with the use of a fixed effects model including time dummies and interaction terms in order to be able to see how the level of competition evolved over time. The results lead to conclude that competition is in a downward trend, this trend is confirmed by several calculations in this research. In the most recent years the level of competition has become so low that a state of monopoly can not be ruled out anymore. This is in contrast with several other studies (like Bikker and Haaf, 2002a) which concluded that the level of competition was increasing.
At the same time, the presence of foreign banks has increased substantially, doubling in the period under research, both in numbers and in assets, from generally 35% to 70%. In all calculations, including one of the forms of foreign ownership, it can be seen that the influence of foreign ownership is positive and, in almost all cases, statistically significant with total (interest) revenue. Graph 4 shows the positive effect of foreign ownership on the level of competition (+0.08). Results indicate that E has a positive relationship with the level of competition and foreign banks earn more (interest) revenue compared to domestic owned banks. This conclusion is in contrast with Yeyati and Micco (2003), who conclude that foreign ownership has a negative relationship with competition.
has seen many new entrants in the market, many of them are foreign. These new entrants increase the level of competition in the market, which cancels out some of the earlier mentioned lower levels of competition. But at the same time these new foreign banks mature the market, thereby introducing new products. These new products expand the size of the market and this means extra growth and lower levels of competition. In the end, the many new (foreign) entrants do not weigh up to the growth and development of the banking market; resulting in lower levels of competition.
References
Bain, J. S. (1951). Relation of profit rate to industry concentration. Quarterly Journal of Economics, 65, pp. 293–324.
Baumol, W.J. (1982). Contestable markets: An uprising in the theory of industry structure. American Economic Review, Vol. 72, pp. 1–15.
Bikker, J.A. and Groeneveld, J.M. (2000). Competition and concentration in the EU banking industry. Research Series Supervision No.8, Netherlands Central Bank.
Bikker, J.A and Haaf, K. (2002a). Competition, concentration and their relationship: An empirical analysis of the banking industry. Journal of Banking & Finance, 26, pp. 2191-2214.
Bikker, J.A. and Haaf, K. (2002b). Measures of competition and concentration in the banking industry: a review of the literature. Economic and Financial Modeling, Vol. 9, pp. 53–98.
Bikker, J.A. and Bos, J.W.B. (2004). Trends in competition and profitability in the banking industry: a basic framework. DNB Working Paper, No. 018.
Bikker, J.A., Spierdijk, L. and Finnie, F. (2006). Misspecifications of the Panzar-Rosse model: Assessing competition in the banking industry. DNB Working Paper, No. 114. Bikker, J.A. and Wesseling, A.A.T. (2003). Intermediation, integration and internationalization: a survey on banking in Europe, DNB Occasional Paper no 3, De Nederlandsche Bank.
Chun, S. E. and Kim, B. H. (2004). The Effect of Financial Restructuring on the Degree of Competition in the Korean Banking Industry.
Claessens, S., Demirgüç-Kunt, A. and Huizinga. H. (2001). How Does Foreign Entry Affect Domestic Banking Markets? Journal of Banking & Finance, 25, pp. 891-911. Claessens, S. and Laeven, L. (2003). What drives bank competition? Some international evidence. Policy Research Working Paper Series 3113, The World Bank.
De Bandt, O. and Davis, E.P. (2000). Competition, contestability and market structure in European banking sectors on the eve of EMU. Journal of Banking & Finance, 24, pp. 1045-1066.
De Haas, R.T.A. and Van Lelyveld, I.P.P. (2002). Foreign Bank Penetration and Private Sector Credit in Central and Eastern Europe. DNB Staff Reports 91.
Demsetz, H. (1973). Information and efficiency: Another viewpoint. Journal of Law and Economics, 10, pp. 1–22.
Drakos, K. and Konstantinou, P. (2005). Competition and contestability in transition banking: An empirical analysis. South-Eastern Europe Journal of Economics, 2, pp. 183-209.
Hasan, I. and Marton, K. (2003). Development and efficiency of the banking sector in a transitional economy: Hungarian experience. Journal of Banking & Finance, 27, pp. 2249-2271.
Jansen, D. and De Haan, J. (2003). Increasing concentration in European banking: a macro-level analysis. Research Memorandum, WO 743, De Nederlandsche Bank.
Jiang, G., Wong, J., Tang, N. and Sze, A. (2004). Banking sector competition in Hong Kong - Measurement and evolution over time. Hong Kong Monetary Authority.
Mason, E.S. (1939). Price and production policies of large-scale enterprise. American Economic Review, 29, pp. 61–74.
Molyneux, P., Lloyd-Williams, D.M. and Thornton, J. (1994). Competitive conditions in European banking. Journal of Banking & Finance, 18, pp. 445-459.
Nathan, A. and Neave, E.H. (1989). Competition and contestability in Canada’s financial system: empirical results. Canadian Journal of Economics, 3, pp. 576-594.
Panzar, J. and Rosse, J. (1987). Testing for ‘Monopoly’ Equilibrium. Journal of Industrial Economics, 35, pp. 443-456.
Peltzman, S. (1977). The gains and losses from industrial concentration. Journal of Law and Economics, Oct. 1977, pp. 229–263.
Shaffer, S. (1982). A non-structural test for competition in financial markets. Proceedings of a Conference on Bank Structure and Competition. Federal Reserve Bank of Chicago, pp. 225-243.
The Economist (2006). Tussles with Brussels. The Economist, 378, issue 8461.
Vesala, J. (1995). Testing competition in banking: Behavioral evidence from Finland. Bank of Finland Studies, E:1, Helsinki.
Vives, X. (2001). Competition in the Changing World of Banking. Oxford Review of Economic Policy, 17, pp. 535-545.
Yeyati, E.L. and Micco, A. (2003). Concentration and foreign penetration in Latin American banking sectors: impact on competition and risk. Working Paper, No. 499, Inter–American Development Bank, Research Department.
