Diversity and competition in the European banking
sector: the impact on financial soundness
Leonie Annemiek van de Meerakker S1991809
Thesis MSc. Finance June 2015
Supervisor: dr. R.M. van Dalen
ABSTRACT
This thesis investigates the impact of the presence of non-commercial banks on the stability of the financial sector in European countries. In line with previous literature, the findings provide evidence that cooperative and savings banks are on average more stable than commercial banks. Yet, the presence of cooperative and savings banks induces commercial banks to take more risks. Furthermore, the results indicate that an increase in competition positively affects banks’ capitalisation. However, this effect is moderated by the degree of diversity of business models of banks in a country. These findings have important implications for policy makers as the ECB, since it implies that they need to take great care in their will to harmonise the financial sector.
JEL classification: G10; G21; G28; G32; G38; P13
2
Contents1. Introduction ... 3
2. Literature review ... 5
2.1. Stability of cooperative and savings banks ... 5
2.2. Competition-fragility versus competition-stability ... 7
2.3. Diversity in the financial system ... 10
3. Methodology ... 13
3.1. Measuring bank stability ... 13
3.2. Empirical approach ... 15
3.3. Bank controls... 16
3.4. Country controls... 17
4. Data and descriptive statistics ... 18
3
1. IntroductionSince November 2014, the European Central Bank (ECB) is the banking supervisor in Europe. According to the Dutch Financial Daily (16 February 2015, p.11) cooperative banks across the euro zone are concerned about the ECB aiming to harmonise the banking sector. After all, uniformity and univocal regulation make supervision less complicated. The ECB is taking a closer look at cooperative and savings banks in several European countries to understand exactly how their business model and their way of decision-making works differently from listed banks (Reuters, 24 March 2015). It is important that the ECB is careful about aiming to harmonise banks, since cooperative banks enhance the financial stability of the system (Groeneveld, 2015). Furthermore, the European Association of Co-operative Banks stated that during the global financial crisis, the cooperative bank sector demonstrated its robustness and resilience, as well as its ability to be a key driver for the real economy (EACB, 2013).
Cooperative banks, accounting for around one fifth of the European Union (EU) bank deposits and loans, are a driving force for socially committed business at a local level (Fiordelisi and Mare, 2014). However, there has been little attention to the importance of cooperative banks in the empirical literature. Compared to their market share, the diversity that cooperative and savings banks add to the banking sector and their implication for financial stability in the EU has received insufficient attention in the literature.
4
Figure 1. Cooperative and savings banks’ market share from selected countries in 2013.This figure shows the three countries with highest market share of cooperative banks (Austria, Netherlands and France), three countries with the highest market share of savings banks (Germany, Sweden and Spain) and three countries with the highest market share of commercial banks (United Kingdom, Greece and Croatia). Market share is calculated based on banks’ total assets retrieved from Bureau van Dijk’s Bankscope Database.
Since cooperative and savings banks are key to the European economy, it is important to investigate their financial stability. Not only compared to commercial banks, but also to examine the effect of the presence of cooperative and savings banks on commercial banks. This relationship has not yet been examined in the post-financial crisis literature. The most recent paper studying this relationship is by Hesse and Cihák (2007). These authors find a negative relationship between the market share of cooperative banks and the stability of commercial banks.
This thesis analyses financial stability of banks employing different business models and the impact of competition from non-commercial banks on commercial banks. To analyse this relation a sample with more than 23,000 observations from cooperative, commercial and savings banks in 23 European countries between 2006-2013 is used. For each bank and year the Z-score is computed. This score is a measure of financial stability, which in fact is an insolvability indicator. Additionally, the three components of the Z-score (return on average assets, capital asset ratio and the standard deviation of the return on average assets) are analysed separately.
This study finds that on average cooperative and savings banks have a higher Z-score, which implies that these banks are more stable. This is primarily due to a lower volatility of returns compared to commercial banks, which offsets the relatively lower capitalisation and lower returns of these
non-0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Au stri a Croa ti a Fran ce G erm an y G re e ce N eth erl an d s Sp ai n Sw ed e n U n ite d K in gd o m M ar ke t Sh ar e
5
profit maximising banks. Furthermore, the results show that a higher presence of cooperative and savings banks lowers the stability of commercial banks. In addition, this study examines the impact that diversity has on the competition-stability relation of the financial system. The empirical results show that increasing competition in a countries banking sector positively influences the soundness of banks, but this effect is moderated by the degree of diversity of business models of banks in a country.
The remainder of this thesis is organised as follows. Chapter 2 reviews the relevant literature and the research hypotheses are formulated. In chapter 3 the methodology is defined. Chapter 4 shows the descriptive statistics and the data employed in the empirical analysis is described. In chapter 5, the main results are presented. Chapter 6 concludes and offers final remarks.
2. Literature review
In this section the most important literature with regard to the effect of cooperative and savings banks’ competition on the stability of commercial banks is discussed. Section 2.1. discusses why the unique business model of cooperative and savings banks results in being on average more stable than commercial banks. Next, in section 2.2. the two contrasting views of the effect of competition on financial stability are assessed. Finally, section 2.3. elaborates on the literature that advocates diversity in the banking sector and explains how the presence of cooperative and savings banks can have an impact on commercial banks.
2.1. Stability of cooperative and savings banks
6
There are several reasons why cooperative banks tend to have less incentive to take excessive risk. First, cooperative banks have an objective to maximise consumer surplus, since they return part of their revenues to their members (which are their customers) in the form of surplus (Ayadi et al., 2011). Hence, they are not under pressure to maximise profits, which can induce risk taking at commercial banks.
Another reason for less risk taking is the diversity of ownership. Cooperative banks are inclined to follow a long-term horizon in their decision making process, because they are owned by a diversity of members. Furthermore, they do not receive short-term pressure from shareholders, which leads to lower risk taking (Beck et al., 2009).
Third, cooperative banks have a strong local presence, which reduces asymmetric information. They have better access to the needs and are better equipped to assess the risk profile of lenders (Groeneveld, 2015). Moreover, there is less risk taking in cooperative banks compared to commercial banks since the principal-agency problems are nearly absent. This is due to the fact that the ownership of cooperative banks is characterised by the one-vote principle, each member has the right to vote, which is not dependent on how much capital he or she owns. As a final note; many cooperatives are part of substantial networks, of which the mutual support they receive should not be underestimated (Groeneveld and De Vries, 2009).
Because of their risk-averse attitude, cooperative banks have in general a lower average return compared to their commercial equivalent. However, in years of distress like the 2007-2008 global financial crisis, cooperative banks were able to use some of the consumer surplus as a cushion to mitigate the negative impact on returns (Groeneveld and de Vries, 2009). This led cooperative banks to weather the storm largely unharmed, in contrast to the large commercial banks that experienced large losses due to risky investments in the years preceding the financial crisis (Bülbül, Schmidt and Schüwer, 2013).
7
Furthermore, savings banks can be – like cooperative banks – part of larger network systems. Compared with commercial banks, their ownership stakes are not publicly available. An important difference between cooperative and savings banks is that savings banks are often publicly owned.
Several studies empirically show that savings banks are more stable (measured by the Z-score, the explanation of this score will follow in the methodology part of this study) than commercial banks (Cihák and Hesse, 2007; García-Marco and Robles-Fernández, 2008; Beck et al., 2009). Furthermore, these studies show that savings banks do not perform worse than their commercial equivalents. Moreover, savings banks are small and operate locally which allows them to provide financing to smaller businesses in less developed regions. They thus fulfil an important role in the development of and contribution to the stability of the financial system (Ayadi et al., 2009).
In summary, previous literature shows that cooperative and savings banks are on average more stable than commercial banks, because of less risk-taking, a cushion of consumer surplus and (nearly) absent principal-agent problems. Therefore, the first hypothesis will be as follows:
H1: cooperative and savings banks are on average more stable compared to commercial banks.
2.2. Competition-fragility versus competition-stability
While many studies empirically investigate whether an increase in competition increases financial stability, most of them assess this relationship between commercial banks, without taking into account the non-commercial banking sector (Fiordelisi and Mare, 2014). Up to now there is no academic consensus whether a competitive banking system (in general) leads more financial stability. Theoretically, there are two opposite views: the competition-fragility and the competition-stability view.
The dominant view in the academic literature and in the actual supervision of banking systems worldwide is that increased competition can threaten the solvency of particular institutions and hamper the aggregate stability of the banking sector (Jiménez, Lopez and Saurina, 2013). This so-called
competition-fragility view states that competition leads to more risk taking and therefore reduces
8
to take more risk, since their profits are under more pressure. This results in higher fragility. Examples of risk taking are taking on more credit risk in the loan portfolios, lower capital levels, or both.
In addition, in a highly competitive banking market, the rate of individual bank failure rises. This intensifies the risk of bank runs and a higher risk of contagion (OECD, 2010). When financial systems have more entry restrictions and are therefore less competitive, banks are better able to benefit from opportunities to make profit and have higher capitalisation ratios. This results in less incentive to take risks (Fiordelisi and Mare, 2014). Beck, De Jonghe and Schepens (2013) find a significant and positive relationship between market power and bank soundness. This indicates that an increase in competition, which erodes banks’ pricing power, increases banks’ risk taking behaviour and is hence damaging financial stability.
Cihák and Hesse (2007) find in their IMF working paper that a high presence of cooperative banks reduces the stability of weak commercial banks in OECD countries. This is due to cooperatives using their lower average cost of capital to follow expansion strategies that can be threatening to commercial banks. This supports the competition-fragility view with respect to cooperative banks having a negative impact on the stability of other banks.
Opposing to the fragility view, several other papers support the
competition-stability view. This view suggests that higher competition in the financial sector enhances its competition-stability due
to banks becoming more relationship oriented. For example, Boyd and De Nicolò (2005) argue when only a few banks have considerable market power, this could result in higher bank risk and therefore the interest rates charged to loan customers will rise. This will lead to financing of risky projects due to adverse selection and moral hazard, which will in turn result in a less stable banking system. On the other hand, they show that lower lending rates reduce the costs of borrowing for entrepreneurs and increase their success rate, resulting in lower credit risk for banks in a more competitive sector.
9
Moreover, the aggregate level of non-performing loans (NPL) is lower in competitive environments. In addition, Schaeck and Cihák (2012) provide evidence that banks from 10 European countries maintain capital levels that exceed the regulatory requirements when there is more competition. Their theory of competition-stability is based on the intuition that bank capital is apparent to customers and a bank with higher levels of capital is more attractive to borrowers. According to this view, competition imposes discipline on banks, which results in higher capital asset ratios and thereby enhancing financial stability. In line with the competition-stability view, Fiordelisi and Mare (2014) find evidence that higher market power negatively affects financial soundness among cooperative banks. Furthermore, consistent with the competition-stability view, banking systems in a more competitive market are less vulnerable to financial crises (Schaeck, Cihák and Wolfe, 2009).
Berger, Klapper and Turk-Ariss (2008) argue that the two views do not have to result in contradicting predictions when analysing the impact of competition on financial stability. They state when banks are operating in a more concentrated banking system their portfolios of loans might be riskier, but they are able to offset this with higher capitalisation or other mitigating techniques. Consistent with the competition-fragility view, they find evidence that less competition does increase the risk of portfolios of loans. However, this risk can be offset by higher equity capital ratios, which is in line with the competition-stability view. Likewise, several other studies (Hakenes and Schnabel, 2010; Martinez-Miera and Repullo, 2010) empirically show that the relationship between bank competition and risk-taking is U-shaped. In this model, increasing competition in a highly competitive banking system increases bank failure risk, thereby reducing financial stability. However, in more concentrated banking markets, bank failure risk declines when competition is increased. Jiménez et al. (2013) test the theoretical non-linear relationship between bank competition and risk-taking and find clear support for this U-shaped relationship in the Spanish banking system. Evidence for the U-shaped relationship is also provided by Liu, Molyneux and Wilson (2012), who assess a regional analysis in Europe on competition to examine bank risk.
Considering the papers that argue there is an U-shaped relationship between competition and financial stability and the fact that the European banking sector is not highly competitive, the European banking sector could benefit from more competition. Therefore, the second hypothesis is:
10
2.3. Diversity in the financial systemAs stated in the previous subchapter, many studies assess the competition-stability relation in the European banking sector. However the empirical research on the effect of diversity in the financial system on this relation is still scant. The attention to cooperative and other non-commercial banks (as savings banks) has risen after the beginning of the 2007-2008 global credit crisis. Several studies examine whether the specific characteristics of non-commercial banks armoured them against the global financial crisis, since they weathered this storm relatively well (Becchetti, Ciciretti and Paolantonio, 2014; Fiordelisi and Mare, 2014).
According to Llewellyn (2014) there is a public policy interest in fostering diversity in the financial system. There are considerable advantages that can be gained from greater diversity in ownership and business models. Llewellyn argues that a diverse banking system with contrasting business models is likely to be more competitive and systematically more sound than a system with only one dominating business model. As put by Ayadi et al. (2011), when the market environment is uncertain, diversity has advantages since the best suited form of business model in different circumstances cannot be predicted. Haldane and May (2011) draw a comparison between the banking system and the dynamics of ecological food webs and with networks within which infectious diseases spread. They state that in the run-up to the recent financial crisis, regulatory intervention is focused on increasing capital requirements and diversification within the bank to reduce risk. However, banks risk management systems became more homogeneous, since they all rely on the same models to measure risk. This came at the cost of a less diverse banking system, thereby increasing systematic risk. In their view, there has been too little incentive from regulators to encourage the diversity of the structure of the balance sheet, business models and risk management systems. They advocate that supervisors give more attention to this systematic diversity objective. This has also been recognised by Altunbas, Manganelli and Marques-Ibanez (2011) in an ECB Eurosystem working paper. Their paper suggests that regulators should intensify their focus on the impact of different business models on risk. Accordingly, a resolution from the European Parliament on 5th June, 2008 already contained the following statement:
“The diversity of legal models and business objectives of the financial entities in the retail banking
11
enriches the sector, corresponds to the pluralist structure of the market and helps to increase competition in then internal market.”An earlier report of the Centre for European Policy Studies (Ayadi et al., 2009) argues as follows:
“Legal, political and risk-related considerations serve to highlight the need for a European banking model based on diversity...”
In line with this view, Bülbül et al. (2013) plea for diversity in the banking system. They state that it is important to secure and preserve banks that follow a non-commercial business model. According to this paper, these banks – savings and cooperative banks – and the networks of these non-commercial banks, up to now managed to perform well as successful enterprises with their social customer surplus-maximising focus. Seen as endangered species that will not brought back to life when they would disappear, policy makers should support these banks that serve the clients’ interest.
Despite the fact that these studies promote diversity in the financial system, there is not much empirical research to support this argument. In order to provide valuable regulation, it is important for policy makers to assess how diversification could affect the safety and soundness of the overall banking system (Fiordelisi and Mare, 2014). As mentioned in the previous paragraphs, there are several papers elaborating on the fact that cooperative banks are relatively more stable than commercial banks (e.g. Hesse and Cihák, 2007; Beck et al, 2009; Ayadi et al. 2011; Groeneveld and de Vries, 2009; Fiordelisi and Mare, 2014). Additionally, there has been an intensive debate in academic as well as policy and regulatory circles whether financial stability is greater in a more competitive environment (Hakenes and Schnabel, 2010; Martinez-Miera and Repullo, 2010; Boyd and De Nicolò, 2005; Schaeck and Cihák, 2010). However, the impact of diversity of the banking sector and its moderating effect on the competition-stability relation in European countries has not yet been examined.
12
low profitability and the risky assets on the balance sheet grow faster than the capital, which will negatively influence the banks’ soundness. Or, these banks sacrifice lower profitability in exchange for delivering value to its clients and they will offer services at below market prices.
Furthermore, cooperative banks have a comparative advantage in gaining trust because they are owned by their members. This can reduce the profitability of other financial institutions, which in turn negatively impacts their stability, since it decreases the solvability of these institutions. Besides the trust advantage, cooperatives can offer services at below market prices since they have a lower cost of capital compared to other financial institutions. This is a result of the fact that they only need to remunerate their member shares part of equity, which they do not remunerate generously (Fonteyne, 2007). These low pay-out ratios imply that highly profitable cooperative banks can experience fast organic growth of their capital, which will be at the expense of commercial banks market share.
As stated in the above paragraph, the empirical research up to now points in the direction that if commercial banks are surrounded by non-profit maximising financial institutions, this could induce them to take more risk. Therefore, the third hypothesis of this study is:
H3: a higher presence of cooperative and savings banks decreases the stability of commercial
banks.
The above section points out that several studies report the advantages of diversity in the business models of banks. This thesis examines if diversity in business models of banks has an effect on the competition-stability relation. However, this moderating effect has not yet been examined in the existing literature. Therefore, it is not clear whether diversity will have a positive or negative effect on the competition-stability relation. The fourth hypothesis elaborated on in this thesis will be:
H4: diversity increases the soundness of the financial system and this has a moderating impact on
the competition-stability relation.
13
banking system is in favour of its stability or harms the soundness of a countries’ financial system. Moreover, diversity of banking business models in a financial system can enhance the soundness of the financial system. Finally, this thesis examines if this diversity of business models impacts the relation between competition and stability in the banking sector.
3. Methodology
This chapter explains the methodology that is used to test the four hypotheses. Section 3.1. describes the measure of financial stability for banks that is used in this study. Next, in section 3.2. the empirical model is presented. Furthermore, section 3.3. and section 3.4. gives an outline of respectively the bank and country control variables that are used in the empirical model.
3.1. Measuring bank stability
The financial stability of an individual bank is measured by the Z-score in this study. The Z-score is an indicator often used as a measure of individual bank risk in empirical research (e.g. Beck et al., 2009; Beck et al., 2013; Boyd and Runkle, 1993; Fiordelisi and Mare, 2014; Hesse and Cihák, 2007; Laeven and Levine, 2007; Nicolò, 2000; Roy, 1952). This score employs the banks’ distance-to-default, which is a proxy for financial soundness. The Z-score measures the number of standard deviations a return realisation has to fall in order to deplete equity, under the assumption of normality of banks’ returns. A higher Z-score corresponds to a lower upper bound of insolvency risk and therefore implies a lower probability of insolvency risk (Fiordelisi and Mare, 2014). The Z-score is the primary dependent variable for the regression model that will be used in this thesis. The Z-score is defined as
𝑍 − 𝑠𝑐𝑜𝑟𝑒 = (1)
where ROAAit is the return on average assets (profit/average assets). CARit is the capital asset ratio, which means the ratio of total equity over total assets. SDROAAit denotes each bank’s standard deviation of the ROAA. The subscripts i and t refer to bank and year, respectively. The SDROAA is calculated using a three-year rolling forward standard deviation of the ROAA1. For example, the SDROAA
14
for 2006 is the standard deviation of the ROAA from 2004-2006 and the SDROAA for 2007 is the standard deviation of ROAA from 2005-2007, and so forth. An advantage for using the rolling time period method is that it allows for time variation in the denominator of the Z-score. Moreover, it allows examining the denominator separately in the regression. Furthermore, the banking sector and economy are constantly changing. Therefore, it is better that in this way both periods report a different SDROAA for a given bank, since the circumstances in 2006 and 2013 are not comparable. A disadvantage of this method is that very low values of SDROAA will be obtained, resulting in high values of the Z-score. Moreover, due to data availability, the Z-score can be only computed for the period 2006-2013, while when calculating the standard deviation of return over the complete sample period a Z-score for the period 2004-2013 could be obtained.
The Z-scorewill increase with the bank’s profitability and capital ratio, but will decrease if the volatility of the banks’ profit increases. Hence, from an economic perspective the Z-scoremeasures the probability of a bank becoming insolvent when the value of assets becomes lower than the value of debt (Fiordelisi and Mare, 2014). The Z-score is computed for every bank and year of the dataset.
The three-year rolling forward method used to compute the SDROAA causes relatively low values for this score, which in its turn results in extremely high values for the Z-score. This is an important limitation of using the Z-score as a measure of financial stability. Therefore, following Houston, Lin, Lin and Ma (2010) and Köhler (2015), next to the Z-score the three components of the Z-score (CAR, ROAA and SDROAA) are analysed separately in order to assess which component of the Z-score is primarily driving the relationship between the independent variables and the Z-score. Since the Z-score is highly skewed, similar to Fiordelisi and Mare (2014), Houston et al. (2010), Köhler (2015) and Laeven and Levine (2009), the natural logarithm of the Z-score as a measure of risk is used in the empirical analysis to smooth out higher values within the distribution.
15
3.2. Empirical approachTo investigate the impact from diversity and competition on financial soundness of the European banking sector, regressions of the following form will be used:
𝑍 − 𝑠𝑐𝑜𝑟𝑒 = 𝛼 + 𝛽 𝑆 + 𝛽 𝐶 + 𝛽 𝑀𝑆, + 𝛽 𝑀𝑆, 𝐶 + 𝛽 𝐻𝐻𝐼, + 𝛽 𝐵𝑙𝑎𝑢, + 𝛽 𝐵𝑙𝑎𝑢, 𝐻𝐻𝐼, + ∑ 𝛽 𝐵, + ∑ 𝛽 𝑀, + ∑ 𝛽 𝐿 + ∑ 𝛽 𝑌 + ɛ (2)
The dependent variable is the Z-score or one of the three components of the Z-score. To distinguish the impact of bank type on the Z-score (H1), dummy variables are incorporated in the model for savings
and commercial banks, respectively 𝑆 and 𝐶 . Several variables include the subscript j, which refers to country. All time-varying variables are lagged to eliminate bias due to time-constant and omitted variables. This is denoted in the regression by the subscript t-1. The total sum of the market share of
cooperative banks and the market share of savings banks (in terms of assets) for each year and country is
calculated and denoted by 𝑀𝑆, in equation (2). To examine whether the presence of savings and cooperative banks negatively influences the stability of commercial banks (H3), the commercial bank
dummy 𝐶 is interacted with the lagged market share variable 𝑀𝑆, . The Herfindahl-Hirschman Index (denoted by 𝐻𝐻𝐼, ) is included as a measure of concentration in the banking sector of a country to examine whether competition has a positive influence on banking stability (H2). This index is defined as
𝐻𝐻𝐼, = ∑ 𝑆 (3)
where 𝑆 is the market share (in terms of total assets) of bank i and N is the total number of banks. The index ranges from zero to one with a higher value indicating a higher concentration of the market. To measure whether the competition effect is moderated by the degree of cooperative banks, the market share of cooperatives and savings banks variable is interacted with the Herfindahl-Hirschman Index. Furthermore, a diversity score is calculated for every country and year, using the Blau’s (1977) Diversity Index formula. This formula is commonly used to measure diversity, but has not yet been used to calculate the diversity among the business models of banks in a country. Blau’s Index is defined as
16
where 𝑃 is the percentage of market share in each category (cooperative, savings or commercial banks) in each country and year. The values of the Blau Index for the diversity in business models of banks range from 0.00 to a maximum of 0.67, which occurs when the three categories of banks are equally represented in a country (measured in terms of assets). The lagged HHI Index is interacted with the lagged Blau Index in the model, to examine on a systematic (country) level the impact of diversity on the effect that competition has on financial stability (H4).
This empirical model follows the checklist for multiplicative interaction models by Brambor, Clark and Golder (2006). Accordingly all constitutive terms and interaction terms are included in the model. The term 𝑀, comprises the lagged macroeconomic control variables for country j at time t-1. The vector 𝐵, comprises lagged bank specific control variables for bank i at time t-1. These control variables will be discussed below. To control for country fixed effects, dummy variables for 22 of the 23 countries are included in the regression. These country dummy variables are denoted by 𝐿 . In addition, year dummies are included in the regression to capture period-specific effects. The year dummies are denoted by 𝑌 in the regression. Finally, ɛ is the residual.
3.3. Bank controls
To examine the impact of cooperative banks’ competition on commercial banks’ financial stability, it is obviously necessary to control for macroeconomic and bank-specific factors that are likely to influence the financial stability of commercial banks. Hence, these control variables help to mitigate omitted variable biases.
The bank specific control variables included are: bank size, a cost inefficiency measure, asset
composition and a measure of income diversification. First of all, bank size is an important variable since
17
with larger assets can have an advantage in their ability to diversify their risks (Beck and Laeven, 2007), which results in more stable earnings (increasing SDROAA). This implies a positive relation between size and stability. The log of total assets is used to control for bank size in the regression, taking into account the non-normal distribution of asset size. The cost-to-income ratio is used to account for management
inefficiency. Less efficient banks are more likely to experience distress, therefore cost inefficiency is
negatively associated with the Z-score according to Beck et al. (2009). Asset composition is computed as the loan-to-asset ratio. It is important to control for this bank specific variable since banks with higher loans relative to assets tend to be riskier (Berger et al., 2008). In addition, to control for the level of diversification in a banks’ income, a measure of income diversity is calculated. Income diversity is defined in accordance with Laeven and Levine (2007):
1 −
(5)
This equation measures the degree to which banks diversify from traditional lending activities (which generate interest income) to other income generating activities. Köhler (2015) shows that retail-banks will be significantly more stable and profitable if they are more diversified. These benefits are particularly large for savings and cooperative banks.
3.4. Country controls
Several country-level variables are included to control for differences in economic development and institutions across countries. The macroeconomic factors included are: the real GDP growth rate, the
real long-term interest rate, the annual change of inflation and the exchange rate appreciation. First of
18
Therefore, the effect of inflation rates depends on the net effect from increasing net interest margins and funding costs. Finally, the exchange rate appreciation is included as a control variable. A change in the value of domestic currency could alter the currency value of a banks’ assets, therefore it has an effect on the CAR which has in its turn an impact on the Z-score. Moreover, Domaç and Peria (2003) find evidence that fixed exchange rate policies would reduce the probability of a banking crisis. Thereby arguing that volatility in exchange rate appreciation is negatively correlated with financial stability.
4. Data and descriptive statistics
This chapter gives an outline of the data and its summary statistics. In section 4.1. the sample that is used in this study is defined. Next, section 4.2. presents several descriptive statistics from the data.
4.1. The sample
The bank-data used in this study are compiled mainly from the BankScope Database provided by Bureau van Dijk. BankScope has a unique coverage of banks, accounting for over 90% of all banking assets in each country. The macroeconomic control variables in this study are retrieved from Eurostat, which is the statistical office of the European Union. An overview of the variables definitions and the data sources is presented in Appendix A.
19
Table 1Number of banks in the sample by type of business model across European countries in 2013.
Country Cooperative Banks Savings Bank Commercial Banks Total
Austria 71 67 45 183 Belgium 3 4 21 28 Bulgaria 1 1 17 19 Croatia 1 1 22 24 Cyprus 1 0 8 9 Czech Republic 2 0 16 18 Denmark 5 31 30 66 Finland 1 3 21 25 France 67 21 84 172 Germany 930 486 89 1,505 Greece 1 0 4 5 Hungary 1 0 13 14 Italy 394 29 63 486 Luxembourg 1 2 44 47 Malta 0 1 5 6 Netherlands 1 1 24 26 Poland 1 1 32 34 Portugal 2 4 14 20 Romania 0 2 17 19 Slovenia 2 1 6 9 Spain 50 12 21 83 Sweden 0 51 17 68 United Kingdom 0 2 88 90 Total 1,535 (52%) 720 (24%) 701 (24%) 2,956 4.2. Descriptive statistics
Table 2 shows the summary statistics of all key variables. Looking at Table 2, the mean log Z-score is 4.57 and its standard deviation is 1.53. This relatively high standard deviation and the wide range of Z-scores (which can also be observed from Table 3 and 4), suggest that there is considerable cross-sectional variation in the level of bank risk.
The summary statistics shown in Table 2, 3 and 4 report higher values for the Z-score and the logarithm of the Z-score compared to the values reported by Laeven and Levine (2009), Hesse and Cihák (2007) and Houston et al. (2010). This is due to the different calculation of SDROAA, using a three-year rolling time window. Other researchers calculate the standard deviation of the return on average assets for the entire sample period, as explained in section 3.1.
20
consists of 23 European countries. The mean ROAA is 0.42% and its standard deviation over the sample period is 0.39%. The average total assets of banks in the sample are 11.2 billion Euro. The total assets of banks across the sample have a fairly high standard deviation. The average asset composition is 0.59 and the average income diversity is 0.58, which is comparable to the descriptive statistics reported by Laeven and Levine (2007). This indicates that on average banks generate about 70% from their income from lending activities and 30% from non-lending activities, which means they do diversify but their main source of income is from lending. The minimum and maximum income diversity are respectively -72.72 and 22.4 where it would be expected to be 0.00 and 1.00. However, this is due to the fact that this income measure is based on calculations with data retrieved from BankScope using equation (5).
Table 2
Summary statistics of the key variables.
To reduce the influence of outliers, the banks with the upper and lower 1% of ROAA, SDROAA and CAR are excluded.
Variable Mean Median Std. Dev. Min. Max. Obs.
Explained variables Z-score 342 81 845 0.05 19660 22,116 Ln Z-score 4.57 4.40 1.53 0.02 9.88 22,116 CAR 8.93% 7.76% 5.21% 1.90% 66.44% 22,116 ROAA 0.40% 0.31% 0.59% -3.35% 3.45% 22,116 SD ROAA 0.42% 0.21% 1.24% 0.00% 0.58% 22,116 Bank-level data
Total assets (th. Euro) 11,200,000 584,000 8,620,000 1,246 2,590,000,000 22,116
Log of total assets 20.43 20.21 1.79 14.04 28.58 22,116
Cost-to-income ratio 0.67 0.67 0.17 0.00 5.00 22,116
Asset composition 0.59 0.61 0.18 0.00 1.00 22,116
Income diversity 0.58 0.53 0.68 -72.72 22.40 22,116
Country-level data
Market share of cooperatives 0.16 0.14 0.10 0.00 0.44 22,116
Market share of savings banks 0.25 0.20 0.19 0.00 0.52 22,116
Market share of commercial banks 0.59 0.57 0.19 0.19 1.00 22,116
HHI Index 0.14 0.12 0.06 0.06 0.50 22,116
Blau Index 0.48 0.54 0.16 0.00 0.61 22,116
21
cooperative banks account on average for 16% of the total market share of banks. An average country-year observation of the HHI Index of 0.13 indicates the banking sector of this sample of European countries is fairly competitive. The Blau Diversity Index of 0.48 across this sample confirms what has also been indicated by the distribution of the market shares across types of banks. Namely, that the sector is quite diverse across the European countries in the sample, but one type of business model has a considerably higher market share.
Table 3 shows that the mean Z-scores are considerably higher for cooperative and savings banks’ than for commercial banks, which suggests a higher stability for these banks. To test whether these differences are also statistically significant, a t-test is executed and presented in the results section. Looking at the breakdown of the components of the Z-score, it is clear that the difference in stability is driven by the significantly lower standard deviation of returns for cooperative and savings banks. A logical explanation might be that the customer-oriented strategies of cooperative and savings banks allows them to better absorb profitability shocks and use their customer surplus as a cushion in less stable times. The higher mean Z-score that is found for cooperative and savings banks is consistent with the existing literature (Ayadi et al., 2009; Ayadi et al., 2011; Beck et al., 2009; Bülbul et al., 2013; Cihák and Hesse, 2007; Garcia-Marco and Robles-Fernández, 2008; Groeneveld and De Vries, 2009).
Table 3
Breakdown of the explained variables from the different specialisations and size of the banks.
Large (small) banks are defined as having assets larger (smaller) than €1 billion. To reduce the influence of outliers, the banks with the upper and lower 1% of ROAA, SDROAA and CAR are excluded.
Z-score ln Z-score CAR ROAA SD ROAA
All Banks 342 4.57 8.93% 0.40% 0.24% Commercial Banks 94 3.61 10.47% 0.54% 0.45% Cooperative Banks 347 4.73 8.62% 0.38% 0.18% Savings Banks 574 5.13 8.12% 0.29% 0.17% Large Banks 353 4.55 7.74% 0.39% 0.22% Commercial Banks 88 3.57 8.28% 0.55% 0.39% Cooperative Banks 331 4.82 8.00% 0.38% 0.14% Savings Banks 661 5.39 6.94% 0.22% 0.11% Small Banks 336 4.58 9.64% 0.40% 0.25% Commercial Banks 103 3.68 14.00% 0.52% 0.56% Cooperative Banks 351 4.71 8.77% 0.39% 0.19% Savings Banks 466 4.80 9.57% 0.37% 0.25%
22
across countries. The high reported Z-score of Germany stands out, as presented this score is driven by the low standard deviation of returns. The lowest Z-score is reported for Greece, which is driven by the high standard deviation of returns.
Table 4
Breakdown of the explained variables between countries in the sample.
To reduce the influence of outliers, the banks with the upper and lower 1% of ROAA, SDROAA and CAR are excluded.
Country Z-score ln Z-score CAR ROAA SD ROAA
Austria 154 3.89 8.86% 0.35% 0.59% Belgium 58 3.46 7.70% 0.59% 0.56% Bulgaria 40 3.30 11.75% 0.77% 0.96% Croatia 74 3.48 13.03% 0.30% 1.11% Cyprus 34 2.86 8.05% 0.43% 1.06% Czech Republic 103 3.83 8.94% 1.01% 0.43% Denmark 54 3.31 12.26% 0.38% 1.08% Finland 126 4.04 7.88% 0.50% 0.33% France 159 4.20 10.21% 0.59% 0.63% Germany 566 5.27 7.33% 0.29% 0.16% Greece 43 3.25 10.46% 0.30% 2.23% Hungary 39 3.15 10.70% 0.43% 1.30% Italy 116 4.01 11.02% 0.48% 0.51% Luxembourg 82 3.58 8.12% 0.67% 0.68% Malta 95 4.14 14.33% 0.84% 1.26% Netherlands 89 3.62 7.75% 0.37% 0.71% Poland 84 3.64 11.16% 0.97% 0.87% Portugal 65 3.45 12.18% 0.41% 1.16% Romania 25 2.87 11.90% 0.43% 1.91% Slovenia 48 3.33 8.28% 0.21% 2.04% Spain 158 4.17 9.37% 0.47% 0.56% Sweden 73 3.60 14.37% 0.93% 0.70% United Kingdom 78 3.58 12.22% 0.39% 1.06% 5. Results
In this section the empirical results of the regression model will be discussed. Section 5.1. presents the preliminary results and explains whether there is evidence found to support the four hypotheses. Section 5.2. elaborates on the robustness of the results.
5.1. Preliminary results
Cost-23
to-income ratio, income diversity and size are negatively correlated with the log Z-score. Furthermore, the reported coefficient for the HHI index is negative, indicating that banking concentration is negatively correlated with stability. On the other hand, the coefficient of the Blau Index as a measure of diversity shows a significant positive relation with the log Z-score, this probably stems from a the fact that the diversity index is negatively correlated with the standard deviation of the return. However, these correlation findings do not imply causal relations and do not account for other control factors, there will be turned to panel regressions later in this results section. As expected, the correlation coefficients for the cooperative and savings bank dummies are positively related to the log Z-score, while the commercial bank dummy shows a negative correlation with this stability measure.
Table 5
T-test equality of means for different specialisation of banks.
This table shows the mean and standard deviation of the natural logarithm of the Z-score, CAR, ROAA and SDROAA by classification of banks’ business models (commercial, cooperative and savings banks). For every mean variable the t-statistic, its p-value and the degrees of freedom are reported. To reduce the influence of outliers, the banks with the upper and lower 1% of ROAA, SDROAA and CAR are excluded.
Specialisation
Cooperative Bank Savings Bank
Mean St. Dev. Mean St. Dev. t-Statistic p-value of t-test df
Ln Z-score 4.73 1.42 5.13 1.61 -16.46 0.000 17,772
CAR 8.62% 3.37% 8.12% 4.28% 8.25 0.000 17,772
ROAA 0.38% 0.41% 0.29% 0.46% 12.82 0.000 17,772
SDROAA 0.18% 0.28% 0.17% 0.36% 1.26 0.209 17,772
Cooperative Bank Commercial Bank
Mean St. Dev. Mean St. Dev. t-Statistic p-value of t-test df
Ln Z-score 4.73 1.42 3.61 1.24 -49.98 0.000 17,645
CAR 8.62% 3.37% 10.47% 8.37% 20.96 0.000 17,645
ROAA 0.38% 0.41% 0.54% 0.93% 15.03 0.000 17,645
SDROAA 0.18% 0.28% 0.45% 0.57% 43.19 0.000 17,645
Commercial Bank Savings Bank
Mean St. Dev. Mean St. Dev. t-Statistic p-value of t-test df
Ln Z-score 3.61 1.24 5.13 1.61 -54.66 0.000 10,693
CAR 10.47% 8.37% 8.12% 4.28% 18.31 0.000 10,693
ROAA 0.54% 0.93% 0.29% 0.46% 17.01 0.000 10,693
SDROAA 0.45% 0.57% 0.17% 0.36% 30.63 0.000 10,693
24
deviation in returns of cooperative and savings banks do not significantly differ from each other. However, the main variable of interest, the log Z-score, does differ significantly for the different bank specialisations.
As shown in Table 5 the mean ROAA is significantly higher for commercial banks than for cooperative and savings banks in the sample, indicating that commercial banks are more profitable. This finding is in line with the literature (e.g. Beck et al., 2009; Fiordelisi and Mare, 2014; Cihák and Hesse, 2007). The CAR is significantly higher for commercial banks than for non-commercial banks, which is also in line with the existing literature that states that cooperative and savings banks are less capitalised (Cihák and Hesse, 2007).Considering the fact that the mean CAR and ROAAare both significantly higher for commercial banks, the lower mean log Z-score for these banks is driven by the standard deviation of the return on average assets, which is almost twice as high as the mean of the total sample. The high standard deviation of commercial banks’ returns on average cannot be offset by its higher CAR and ROAA, resulting in a lower log Z-score. This evidence supports the first hypothesis: cooperative and savings banks have a higher Z-score and therefore it can be stated that they are on average more stable than commercial banks. This is in line with the findings of the existing literature (Ayadi et al., 2009; Ayadi et al., 2011; Beck et al., 2009; Bülbul et al., 2013; Cihák and Hesse, 2007; Garcia-Marco and Robles-Fernández, 2008; Groeneveld and De Vries, 2009).
25
Table 6The impact of diversity and competition on the individual stability of banks.
This table reports the Panel Least Squares regression results including bank and year dummies. Log Z-score is computed as the natural logarithm of (CARit + ROAAit)/SDROAAit. Where CARit is the capital asset ratio, ROAAit the return on average assets, and SDROAAit the standard deviation of the return on average assets, computed using a three year rolling time window. A higher Z-score implies more stability. An F-test is conducted on the log Z-score, CAR, ROAA and SDROAA to examine whether the means significantly differ over time and country. The reported statistic of the p-value is 0.0000, therefore country and year dummies are included in the regression. Values are computed by the heteroskedasticity-robust standard errors clustered by banks. To reduce the influence of outliers, the banks with the upper and lower 1% of ROAA, SDROAA and CAR are excluded. *, **, *** Respectively, represent statistical significance at the 10%, 5% and 1% levels.
Ln Z-score CAR ROAA SD ROAA
Savings bank dummy 0.4439*** 0.0054*** -0.0009*** -0.0001
Commercial bank dummy 0.3869*** -0.0086*** -0.0009** 0.0001
Sum of market share of cooperatives and
savings banks 0.8737 0.0386 0.0110* 0.0024
Commercial bank dummy*
Sum of market share of cooperative and
savings banks -2.4562*** 0.0747*** 0.0050*** 0.0039***
HHI Index 0.4530 -0.1338*** -0.0016 -0.0027
Blau Index -0.3520 -0.1465*** -0.0167 -0.0042
Blau Index*HHI Index 6.3125** 0.2662*** 0.0233 -0.0071
Log of assets -0.0106 -0.0104*** -0.0003*** -0.0002*** Cost-to-income ratio -0.7073*** -0.0253*** -0.0080*** 0.0030*** Asset composition 0.0366 -0.0070*** -0.0007 -0.0005*** Income diversity -0.0225 -0.0009 0.0002 0.0001 GDP growth 2.5968*** -0.0527 0.0338*** -0.0106 Interest rate -10.0651*** -0.3360*** -0.0659*** 0.0320*** Inflation -2.5794 0.0046 0.0003 0.0096
Exchange rate appreciation 0.4490 -0.0349 0.0118*** -0.0065***
Constant 4.4781*** 0.4101*** 0.0185*** 0.0068
Observations 21902 21902 21902 21902
R-squared 0.3175 0.2817 0.2573 0.2438
26
Figure 2. Presence of non-commercial banks and stability.This figure shows the impact of the total sum of market share of cooperative and savings banks for different specialisation of banks on the soundness of banks with a different business model.
Figure 2 presents the regression results for the interaction effect of the total sum of market share of cooperative and savings banks and the commercial bank dummy on the average stability of banks in this European sample. This figure shows that if the sum of market share of cooperative and savings banks will rise, ceteris paribus, the stability of a commercial bank will drop. Contrary to the stability of commercial and savings banks, the stability of those banks will rise when their total market share rises.
Table 7
The impact of the market share of cooperative and savings banks on stability.
This table reports the marginal effect of the total sum of market share of cooperative and savings banks on the ln Z-score for non-commercial and commercial banks. *, **, *** Respectively, represent statistical significance at the 10%, 5% and 1% levels. Standard errors are in parentheses.
Commercial bank dummy Marginal effect
0 0.4145**
(-0.197)
1 -2.1193***
(-0.127)
Brambor et al. (2006) state that if a multiplicative interaction model is employed, it is not sufficient to simply report the coefficients and the significance of the variables. In accordance with this paper, the marginal effect of the total sum of market share of cooperative and savings banks for non-commercial and non-commercial banks, along with the two corresponding standard errors is reported in
3 3.5 4 4.5 5 5.5 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 Ln Z -sco re
Sum of market share cooperative and savings banks
27
Table 7. This table shows that the positive impact of the total sum of market share of cooperative and savings banks for non-commercial banks as displayed in Figure 2 is significant at a 5% level. The same applies to the negative impact on commercial banks, which is even significant at a 1% level. Together with the regression results reported in Table 6 and Figure 2, this table gives reason to accept the third hypothesis of this thesis: a higher presence of cooperative and savings banks will on average decrease the stability of commercial banks. This finding is in accordance with the existent literature on this subject (Barth et al., 1999; Cihák and Hesse, 2007; Fonteyne, 2007; Goodhart, 2004).
Figure 3. The impact of diversity on the competition-stability relation with respect to a banks’ capital asset ratio (CARit). For different diversity ratio’s (Blau Index), the impact of rising concentration on the capital asset ratio is shown.
The coefficients of the Blau Index as a measure of diversity and the HHI Index as a measure of competition are not statistically significant. However the interaction effect of these two variables has a significant positive impact on the log Z-score at a 5% confidence level. In contrary, both the two separate explanatory variables and the interaction effect show significant coefficients for the regression with the CAR as dependent variable. To clarify the interaction effect of these variables on the capital asset ratio of a bank in a more tangible way, they are graphically displayed in Figure 3. This figure shows the effect of rising concentration of banks in banking sectors for different values from the Blau Index, ranging from no diversity (Blau Index = 0) to maximum diversity (Blau Index = 0.67), on the capitalisation of individual
0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.0 2 0.0 4 0.0 6 0.0 8 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34 0.36 0.38 0.4 0.42 0.44 0.46 0.48 0.5 CAR HHI
28
banks. The average Blau Index reported for the full sample is 0.48 and the average level of competition is a HHI Index of 0.14 (Table 2). The HHI Index on the horizontal axis ranges from 0 to 0.5, since the maximum HHI reported for the sample in this study is 0.50. The six lines display the effect of competition on stability for different levels of diversity. Independent of the level of diversity, a rising HHI index has a negative effect on the capital asset ratio of banks.
According to Brambor et al. (2006) and Golder (2006) the regression results from interaction effects as presented in Table 6, are informative but can only shed limited light on the third and fourth hypotheses. The results show that decreasing competition has a negative effect on the CAR when the Blau Index is zero. However, the results do not indicate whether competition has a statistically significant impact on the CAR when the Blau Index is greater than zero. Figure 4 graphically shows how the marginal effect of the HHI index on the CAR changes across the observed range of the Blau Index. The solid upward sloping line shows how the marginal effect of competition changes with the level of diversity. This line displays the following equation:
𝜕𝐶𝐴𝑅
𝜕𝐻𝐻𝐼 𝐼𝑛𝑑𝑒𝑥= 𝛽 + 𝛽 𝐵𝑙𝑎𝑢 𝐼𝑛𝑑𝑒𝑥 (6)
29
Figure 4. The marginal effect of competition (measured by the HHI Index) on the CAR. Depicts the conditional relationship between competition and the capital asset ratio as a function of the level of diversity. The moderating effect is statistically significant when the upper and lower bounds of the confidence interval are both above or below the zero line.It follows from Figure 4 that only the three top lines displayed in Figure 3 are significant. Looking at these three lines for different ranges of diversity; if the level of diversity is higher, the downward sloping curve is less steep than in case of minimum diversity. This provides evidence that diversity moderates the competition-stability relation. In summary, these results indicate that in accordance with the competition-stability view decreasing banking competition has a negative impact on the capitalisation of banks, which has in its turn a negative effect on the Z-score (Boyd and De Nicolò, 2005; Schaeck and Cihák, 2010; Uhde and Heimeshoff, 2009). Therefore, the result indicates higher fragility in a more concentrated banking system. The result is significant up to a certain level of diversity, which is the case for most of the countries in the sample. These outcomes thus partly provide evidence to support the second hypothesis. Furthermore, the results indicate that diversity does have an impact on the competition-stability relation. However, no empirical evidence is found in this study that diversity increases the soundness of the financial system. Accordingly, the results provide evidence for a part of the fourth hypothesis of this thesis.
In general, the bank and country control variables in the panel regression have the expected signs. Larger banks tend to be more risky, which supports the expectations discussed in section 3.4. Although the coefficient for the log of assets does not have a significant impact on the log Z-score, they are significant for the different components of the Z-score. Larger banks tend to have a lower CAR, ROAA
30
and on the other hand lower standard deviation of the ROAA. This implies they are better able to diversify their risks (lower SDROAA), but this does not offset the lower CAR and ROAA, suggesting too-big-to-fail policies. Inefficient banks in terms of their cost-to-income ratio have a higher Z-score, which is due to the fact that they are less likely to cover their costs in case of a crisis (Cihák and Hesse, 2007). Furthermore, inefficient banks can be ruled out of business when competition increases (Schaeck and Cihák, 2010). Asset composition has a slightly negative effect on the CAR. As expected, since banks with a larger percentage of loans (relative to total assets) have lower capitalisation (Berger et al., 2008). On the other hand, when the loan-to-asset ratio rises, the volatility of returns decreases. Neither of these contrasting effects on the two components of the Z-score is dominant, since the effect on the log Z-score is not significant.
As expected, GDP growth has a positive effect on bank stability, which is mainly driven by the significantly positive increase of the ROAA. This indicates that a banks’ investments opportunity can rise under economic growth and a lower insolvency of borrowers. The regression results show that lower interest rates have a positive effect on bank soundness. This is in line with the findings of Beck et al. (2009), who find a significant negative correlation between the interest rate and a banks’ Z-score. The reader should note that although the coefficient seems quite large, if the interest increases by 1%, this would reduce the (non-logged) Z-score with approximately 1 unit. Among the country control variables, inflation and income diversity have insignificant coefficients. Exchange rate appreciation has a positive impact on the ROAA and a negative impact on the SDROAA. This would suggest an increase of the Z-score. However, although the coefficient on the log Z-score is positive, it is not significant.
5.2. Robustness checks
31
relationship between bank-size and financial stability. Furthermore, a similar effect is found with the loan-to-asset ratio. This ratio has a significant negative effect on bank soundness for small banks, while it has a significant positive effect on stability for large banks. This suggests that small banks will benefit from more diversification in assets, whereas large banks seem to be over-diversified. Similar results are found for income diversification, although not significantly positive for large banks. These results are in line with the findings of Köhler (2015) who states that the business models of banks should be taken into account when pursuing a strategy to become more stable. In general, commercial banks are large banks, while cooperative and savings banks are smaller. Cooperative and savings banks are more retail-oriented which often means they will benefit from more diversification. Commercial banks are in general more investment-oriented and are found to be over diversified and more risky if they increase their share of diversification.
To further assess the robustness of the results, a sample is selected of 14 European countries to estimate the regression. This selection is based on the EU-15 countries, excluding Ireland. The definition of EU-15 counties is that these are member countries of the European union prior to 1st May 2004. The sample includes: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden and the United Kingdom. The regression results are presented in Appendix F. Again, the main regression results do not differ considerably from the original samples results, confirming its robustness.
32
6. ConclusionBy looking at a sample of nearly 3,000 banks in 23 European countries, this thesis finds evidence that savings and cooperative banks are on average more stable than commercial banks. The higher stability of these banks is due to lower volatility of the returns on assets, which more than offsets their relatively lower return and capital asset ratio. There are several possible explanations for the lower volatility of returns of non-commercial banks among which a more risk-averse attitude and the use of a capital cushion to face less stable times. This finding is consistent with earlier studies (Ayadi et al., 2009; Ayadi et al., 2011; Cihák and Hesse, 2007; Garcia-Marco and Robles-Fernández, 2008; Beck et al., 2009).
On the other hand, a higher presence of cooperative and savings banks induces commercial banks to take more risks. This supports existing theories that the presence of non-profit maximising banks can weaken commercial financial institutions (Goodhart, 2004; Cihák and Hesse, 2007). The empirical results from panel estimations hold when employing alternative samples or an alternative calculation of the Z-score. Within the backdrop of large literature looking at the competition-stability versus the competition-fragility relation in financial systems, this study explores how competition affects banking stability. Evidence is found that increasing banking concentration within a country has a negative impact on the capital asset ratio of banks, suggesting a decrease in competition induces banks to take more risk. This supports the competition-stability view (Boyd and De Nicolò, 2005; Schaeck and Cihák, 2010; Uhde and Heimeshoff, 2009).
33
financial sector, since this will increase systematic risk. Moreover, a more competitive banking sector benefits stability since it imposes discipline on banks. Therefore, consolidation of banks should not be stimulated, from a regulatory and supervisory perspective.
The explanatory variable from the regression is based on the Z-scores of individual banks. Using the Z-score has some drawbacks, since it is based on accounting data and focuses on capital and profits instead of liquidity or asset quality and therefore is kind of a primitive measure of (in)stability. Furthermore, the Z-score is highly skewed, since the three-year rolling forward standard deviation of returns results in very disparate scores. To assess stability at a more complex level, one can analyse the non-performing loans ratio, which is an indication of lending risk. Unfortunately, it was not possible to examine this ratio in this thesis, since there was little data available to compute this ratio. In addition, a probability of distress-score can be computed as a measure of actual insolvency risk, which focuses on the tail risk of banks. Beck et al. (2009) employed this time-consuming method in their study. Furthermore, the probability of default on individual loans or pools of transactions, the losses-given-default and the correlation across losses-given-defaults can be estimated to assess a banks’ stability on a more detailed level (Dermine and Neto de Carvalho, 2006). Finally, this thesis assesses the effect of certain variables on the individual bank stability. To assess the financial stability on a country level, one could look at the aggregate Z-score to get one stability ratio for every country (Uhde and Heimeshoff, 2009).
34
AppendicesAppendix A
Description and sources of the individual variables.
Variable Description Source
Ln Z-score
Equals log of (ROAA+CAR)/(SDROAA), where ROAA=p/AA is return on average assets per bank per year and CAR=E/A is capital asset ratio per bank per year, SDROAA is the standard deviation of ROAA, computed using a three year rolling forward time window.
Bureau van Dijk BankScope, own calculations
Size Log of the total assets of a bank (In EU€ billion).
Bureau van Dijk BankScope Asset composition Ratio of loans to assets (percent).
Bureau van Dijk BankScope
Cost-to-Income Ratio Ratio of costs to income (percent). Bureau van Dijk BankScope
Income Diversity 1 −Net Interest Income − Other Operating Income Total Operating Income
Bureau van Dijk BankScope, own calculations based on Laeven and Levine (2007)
Herfindahl Index (HHI) Sum of squared market shares (in terms of total assets) on a country level, with higher values indicating greater market concentration.
Bureau van Dijk BankScope, own calculations GDP Growth Gross domestic product at market prices (percentage of change over the precious
period).
Eurostat
Inflation Annual average rate of change of harmonised indices of consumer prices (HICP). Eurostat Real Long-Term
Interest Rate
Nominal long-term interest rate (EMU convergence criterion bond yields), adjusted for lagged GDP deflator (percent).
Eurostat
Exchange Rate
Appreciation Real Effective Exchange Rate (REER) – 42 trading partners, index (2005=100).
Eurostat
Share of Cooperative
banks Market share of cooperative banks (in terms of total assets) in a country per year.
Bureau van Dijk BankScope, own calculations
Share of Saving banks Market share of savings banks (in terms of total assets) in a country per year.
35
Appendix B36
Appendix CCorrelation table.
This table reports the correlation coefficients between the log Z-score, the components of the Z-score and selected key variables over the entire sample period (2006-2013). Log Z-score is computed as the natural logarithm of (CARit + ROAAit)/SDROAAit. Where CARit is the capital asset ratio, ROAAit the return on average assets, and SDROAAit the standard deviation of the return on average assets, computed using a three year rolling time window. A higher Z-score implies more stability. To reduce the influence of outliers, the banks with the upper and lower 1% of ROAA, SDROAA and CAR are excluded. *, **, *** Respectively, represent statistical significance at the 10%, 5% and 1% levels.
Variable Ln Z-score CAR ROAA SD ROAA
Commercial bank dummy -0.3395*** 0.1595*** 0.1320*** 0.2907***
Cooperative bank dummy 0.1053*** -0.0521*** -0.0220*** -0.1570***
Savings bank dummy 0.2107*** -0.0959*** -0.1041*** -0.1020***
37
Appendix DRobustness tests: regression results small banks.
This table reports the Panel Least Squares regression results including bank and year dummies. Log Z-score is computed as the natural logarithm of (CARit + ROAAit)/SDROAAit. Where CARit is the capital asset ratio, ROAAit the return on average assets, and SDROAAit the standard deviation of the return on average assets, computed using a three year rolling time window. A higher Z-score implies more stability. An F-test is conducted on the log Z-score, CAR, ROAA and SDROAA to examine whether the means significantly differ over time and country. The reported statistic of the p-value is 0.0000, therefore country and year dummies are included in the regression. Values are computed by the heteroskedasticity-robust standard errors clustered by banks. Sample of small banks with assets smaller than €1 billion. To reduce the influence of outliers, the banks with the upper and lower 1% of ROAA, SDROAA and CAR are excluded. *, **, *** Respectively, represent statistical significance at the 10%, 5% and 1% levels.
Ln Z-score CAR ROAA SD ROAA
Savings bank dummy 0.4246*** 0.0055*** -0.0008*** -0.0003***
Commercial bank dummy 0.3607*** 0.0004 -0.0007 0.0004
Market share of cooperatives and
savings banks 0.9091 0.0461** 0.0126* 0.0022
Commercial bank dummy*Market share of cooperative and savings
banks -2.0113*** 0.1068*** 0.0052*** 0.0029***
HHI Index 1.4174 -0.0806 0.0092 -0.0031
Blau Index -0.5334 -0.1490*** -0.0126 -0.0016
Blau Index*HHI Index 1.3878 -0.0204 0.0097 -0.0001
Log of assets 0.0399** -0.0140*** -0.0004*** -0.0004*** Cost-to-income ratio -0.6427*** -0.0273*** -0.0082*** 0.0034*** Asset composition -0.3042*** -0.0247*** -0.0005 -0.0011*** Income diversity -0.1249*** 0.0007 -0.0001 0.0007*** GDP growth 6.2796*** 0.1113 0.0422*** -0.0225** Interest rate -5.0028* -0.3717*** -0.0295 0.0087 Inflation -3.6194 -0.0178 -0.0168 0.0127
Exchange rate appreciation 0.3986 -0.0408* 0.0121** -0.0080***
Constant 2.3287*** 0.5173*** 0.0233*** 0.0095***
Observations 13568 13568 13568 13568
