Does Competition in Banking explains Systemic Banking Crises?

Does Competition in Banking explains Systemic Banking Crises? Abstract: This paper examines the relation between competition in the banking sector and the financial stability on country level. Compared to previous research, it takes a different approach in that it uses realized systemic risk in the form of systemic banking crises instead of the total systemic risk. Theory provides us with two opposing theories regarding the role of competition on stability. Previous studies presented mixed results which leaves us with unresolved questions which this paper tries to answer. The results show that there is evidence for both views, but without giving an all comprehending answer. Key words: financial stability, bank competition, systemic banking crisis Student number: Name: Study Program: Supervisor: S2020807 Roy Hamstra DDM UU IFM Martien Lamers Word count: 11,108* *including graphs, appendices, and tables which are inserted as pictures

1. Introduction After the financial crisis of 2007/2008 the debate about the role of banks became very important for policymakers and society. The crisis showed the world how important banks are in modern day society and that distress in banks have broader consequences than for normal companies. The bankruptcy of Lehman in 2008 is the most well-known example in recent history and the consequences for the real economy were significant across the world. Consequently policymakers wanted to prevent a next crisis from happening and so they asked regulators to develop a new set of rules. In most developed countries this new regulation would become Basel III which among others imposes higher capital requirements as a buffer against financial distress to achieve the main goal; the increase of financial stability. To achieve this goal, one must understand which forces drive the financial stability of a country. One of these factors suggested by researchers is the effect of competition. Competition has the disadvantage that it increases the default risk of companies, which in the case of banks, has significant consequences for the economy. The aim of this paper is to research the effects from competition on financial stability. This is an important issue because policy makers can have a direct influence on the competition with their power to stop mergers and acquisitions or on the contrary force Mergers & Acquisitions. In most markets, competition is considered as a positive characteristic because it lowers prices, increase product quality, and drives innovation. However in banking, competition could also cause instability. Two of the main views regarding this issue are the competition fragility and competition stability theory. While the first one argues that a lower degree of competition improves the stability, the latter states that less competition decreases stability. The competition fragility theory first suggested by Keeley (1990) states that more competition leads to lower margins and a higher risks preference. Higher risks combined with lower profits causes instability of the system. On the other hand, Boyd and De Nicoló (2005) point out that high margins in monopolistic markets create a new problem which they call the risk shifting problem. The high margins cause borrowers to compensate this margins with riskier projects which results in a higher risk for banks. In addition to these opposing views, the relationship is also influenced by other factors like the development of regulations, financial markets, and other institutions. This paper takes a different approach compared to the previously named papers in that it uses the occurrence of systemic banking crises on a country basis, instead of a bank s z-score which 1

measures the distance to default. The difference is that I take the realized excess risk on a national perspective instead of the theoretical risk on an individual bank level. Risk per se is not a bad thing and is part of the doing business, however excess risk is. The main input are the database of recent financial crises and the World Bank data on country level bank characteristics. The main questions this paper tries to answer is whether the level of competition has a significant effect on the financial stability measured by systemic banking crises. I also check the effect on the z-score to check whether this dataset is in line with the theories. In addition to the competition-financial stability relationship this paper also tries to give an explanation to the following question; did the level of institutional development have a significant effect on the likelihood of a systemic banking crisis? And what is the effect of competition on the severity of a crisis? The answers to these question could help policy makers improve their actions before and during crisis times. The results show evidence for the competition fragility view where financial stability is measured in terms of systemic banking crises as well as in the z-score. Furthermore the results show that institutional development and non-interest income play a moderating role in the competition stability relation. Due to the quality of the available data there is little evidence for any relation between competition and the severity of a crisis which offers opportunities for future research. The paper proceeds as follows: it starts with a review of the current literature including hypothesis. Section 2 provides an explanation of the methodology including the data used for the analysis, and how each variable is measured or calculated. Section 3 continues with the analysis of the regressions and present the results which will be linked back to the hypothesis formulated in the literature section. Section 4 provides an overview of the limitations and possible further research areas. Section 5 offers concluding remarks and implications. 2. Theory This section elaborates on the current literature. It gives an overview of what is already known on the various subjects and links the different parts with each other to form hypothesis which will be tested later on. 2.1. Competition in banks and financial stability According to research, competition is, in most circumstances, desirable for consumers because it lowers prices, drives innovation, and increases customer power (Kasman and Kasman, 2015). 2

However the default risk increases when competition increases. The default risk or even the bankruptcy of a firm is in most cases not too harmful for society in general. So why do we care so much about the bankruptcy of banks? Banks are a special kind of company because they perform a crucial role in the economic system. Almost every firm or household depends on banks to perform various financial services. Bank failure could have enormous consequences when looking at for instance Lehman Brothers in 2008. Banks have a direct effect on the financial stability in a country on sometimes even in multiple countries. Financial stability is one of the key goals for organizations like the IMF and the European Central Bank. It is therefore important to understand which factors affect the financial stability and in what way. The relation between bank competition and financial stability has been studied over the years but is still incomplete. Over the years researchers have come up with several theories regarding this relation. The franchise value theory and the BDN model are the most established theories and the following section provides an analysis of both views. 2.1.1. Competition fragility The competition fragility theory is also known as the franchise value theory which originated from the research by Keeley (1990). As the name suggests, this traditional view states that more competition leads to fragility of the financial system. The franchise, or charter, value is the value of a bank s ability to continue its operations in the future and is therefore dependent on the future profits of the bank. Factors which contribute to the franchise value are for example but not limited to: a bank s growth opportunities, pricing power, and goodwill. This value can be seen as an intangible asset which has no value in case of bankruptcy (Ren and Schmit, 2009). Banks and their shareholders have therefore incentives to protect this value by taking low risks to sustain their operations and thereby protecting the franchise value. Keeley (1990) was one of the first who provided empirical evidence for a negative relation between bank competition and their franchise value. He found that the franchise value of US banks decreased during the period between 1950 and the end of the 1980s. In that same period, the US banking sector experienced a large degree of deregulation which led to an increase in competition. Furthermore he documents that banks responded to the decreasing franchise value by taking on more risk. But what causes a lower franchise value to decrease the financial stability? To answer this question we have to take a step back. Consider a market with a low degree of competition. In this market, banks have a lot of market power which they use to set high interest rates to their customers. These monopoly rates are a main driver of the bank s franchise value. Because there is low competition customers have no choice but to accept the high rates. In addition to this, 3

banks want to protect their franchise value and therefore have a low risk preference (Berger, Klapper, Turk-Ariss, 2009). This combination results in less risky, high quality loans with high margins. When competition increases, the pricing power of banks decrease and they need to increase their risk preference in order to sustain revenue. This results in lower margins and riskier loans. The combination of riskier loans and low margins increases the systemic risk of banks and consequently reduce the franchise value of a bank (Jimenez, Lopez, and Saurina, 2012). These authors also show that more competition leads to lower capital ratios which also leads to higher systemic risk. The literature provides us with several studies which support the franchise value theory. Repullo (2004) presents a model of imperfect competition where banks can invest in a risky or less risky asset. He shows that banks have an incentive to invest in the risky asset when margins are small and the franchise value is low. To overcome this problem he suggests to imply capital requirements to counter this risk taking incentive. Beck, Schepens, and De Jonghe (2013) provide empirical evidence where competition measured by the Lerner index has a negative effect on the financial stability measured by the z-score. However this relation is altered by the development of institutions. I will analyze the effects of institutional development later on. Considering the previous the following hypothesis can be derived: Hypothesis 1: Competition has a negative effect on financial stability measured in terms of the occurrence of financial crisis 2.1.2. Competition stability The competition stability theory is an alternative theory about the relation between bank competition and stability. The basis of this theory is suggested by Allen and Gale (2000) and is further developed by Boyd and De Nicoló and also referred to as the BDN model (2005). This theory states that while banks have the incentive to protect franchise value, there is also a risk incentive mechanism working the opposite direction. This mechanism is the moral hazard problem. Similar to the competition fragility theory this theory states that banks will reduce risks and increase their interest rates when competition is low. This action however produces two kinds of problems. The first problem is that high interest rates (slightly) increase the default probability of the borrower because it increases the difficulty of repaying the loan. The second problem is more dangerous as borrowers determine their project risk based on the rate charged by the bank. Higher rates results in riskier projects because riskier projects produce higher returns. These problems will increase the systemic risk banks face as the risk of their loan 4

portfolio will increase. More competition leads to lower interest rates which reduces the moral hazard problem and thereby increases the financial stability. Martinez-Miera and Repullo (2010) extended the BDN model and the most significant contribution is the introduction of the margin effects. The MMR model is based on two opposing forces namely the margin effect and the risk shifting effect. The margin effect works as follows: when competition is low, margins are high, and this causes banks to build up a buffer from the profits derived by the high margins. The risk shifting problem refers to the risk preference of borrowers based on the interest rates. MMR show that there is a trade-off between both forces and that the margin effect is more dominate in competitive markets while the risk shifting problem dominates in monopolistic markets. As a consequence the financial stability is low in markets with high or low degrees of competition. In markets with very low competition, the risk shifting problem is too large to be compensated for with the buffers from the high margins, while high competitive markets suffer very little from the risk shifting problem but the margins are too low for banks to be profitable. The most stable situation is a situation where the risk shifting problem and the margin effect are relatively equal. There will be some risk shifting problems but the buffers are sufficient to cope with it. This results in a U- shaped relation between competition and risk taking and an inverted U-shape between competition and financial stability. In practice this means that competition increase stability when competition is low till a certain point where after more competition leads to less stability. Schaek, Cihak, and Wolfe (2009) provided empirical evidence in favor for the competition stability theory. They document that markets with higher levels of competition measured by the Panzar and Rosse (1987) H-statistic are less likely to experience a systemic banking crisis. Similar to the competition fragility theory there can be an alternative hypothesis derived according to the BDN and MMR theory. Hypothesis 2: Competition has an inverted U-shaped relation with financial stability measured by the likelihood of a financial crisis 2.2. Concentration Multiple studies have shown that the concentration of the market has a significant influence on the financial stability (e.g. Beck, Demirgüç-Kunt, Levine, 2005; Fu, Lin, Molyneux, 2014). Concentration is usually measured as the asset concentration of the three largest banks of the country. This percentage gives an indication of the market power of these largest banks. While competition and concentration usually go hand in hand, this is not always the case. High bank 5

concentration does not exclude a high degree of competitiveness. Research from Asia shows that a higher concentration is bad for the financial stability (Fu, Lin, Molyneux, 2014). Another paper examining the Turkish banking sector found similar results where a higher degree of concentration has a positive effect on the non-performing loan (NPL) ratio which as a consequence increases the default risk and make the financial system unstable (Kasman and Kasman, 2015). Their conclusion is therefore that policymakers could stimulate mergers between small banks to reduce the concentration and give these smaller banks more chance of survival. Beck, Demirgüç-Kunt, and Levine (2005) found similar results in which lower levels of bank concentration result in more financial stability. This is in line with the competition stability view. They used the likelihood of a systemic banking as a proxy for financial stability. However they did not investigate the mechanisms driving this relationship. An explanation of the negative effects in case of high concentration can be found in the moral hazard problem. These negative effects are caused by the fact that the banks are too big or too important to fail and are therefore insured of subsidies by their governments (Fu, Lin, Molyneux, 2014). This could provide incentives for managers to take more risk because they know they will be bailed out in case things turn out negative. It can be said that these too big to fail banks have a put option on their bank where there is only upside potential. When the situations turns out bad they know that they will be bailed out by the government. However one should bear in mind that these bail-outs also improves the financial stability because it reduces the change of bank runs and default. Concluding this leads to the following hypothesis: Hypothesis 3: Bank concentration has a negative relation with financial stability 2.3. Development of financial markets, institutions, and regulators While both theories suggest different relationships and propose a competing view on competition in the banking sector, researchers have shown that both theories can be united and that a combination provides the best explanation (Beck, De Jonghe, and Schepens, 2013). In addition they provide evidence that the framework in which a bank operates influence the competition stability relation. These factors are the development of the financial market, institutions, and regulatory supervision. As an example they show that an increase in competition between banks in, for instance, countries with higher developed financial markets have a smaller negative impact than in countries where financial markets are less developed. This condition enables banks which operate in an environment with higher developed markets, 6

institutions, and regulatory supervisors to achieve additional competition benefits while maintaining financial stability. The development of financial markets can have an enhancing effect on the financial stability. Higher developed markets are usually larger and more liquid than less developed markets and provide entrepreneurs more options to obtain funds (Beck, Schepens, and De Jonghe, 2013). Therefore entrepreneurs can easily switch between bank- and market-based funding which reduce the dependability on banks and therefore increase the financial stability. Furthermore the authors state that higher developed markets are associated with more transparency and disclosure obligations which in turn could reduce the bank risk behavior. The effects of competition are also moderated by the development of institutions and regulatory supervision like customer protection (Anginer, 2014). Better rules and supervision can decrease the negative side-effects from too much competition like risk taking behavior and allows more competition without the decrease of financial stability. One of the most popular methods of supervisors is the deposit insurance scheme where deposits are insured up to a certain amount. This measure is intended to prevent bank runs and thereby increasing financial stability (Beck, Schepens, and De Jonghe, 2013). However as already mentioned, this measure could by itself also reduce the stability because it offers incentives for bank managers to increase their risk taking behavior. Because it costly as well as difficult for individual investors to control bank behavior, a proper working supervisor with sufficient power could help reduce risk taking behavior and again increasing stability while maintaining the same level of competition. Likewise Beck, Schepens, and De Jonghe (2013) state that better institutions increase the information about borrowers which forces borrowers to behave less risky in order to obtain future loans. This can be combined with the MMR theory where the stability depends on the risk taking behavior of the borrowers. When borrowers take less risk this increases the stability and if everything else remains equal, countries with better institutions are more stable than countries with less developed institutions. Based on these different theories a number of hypotheses can be derived; Hypothesis 4: The effect of competition on stability is negatively affected by the development of a country s financial market. Hypothesis 5: The effect of competition on stability is negatively affected by the development of a country s institutions. 7

Hypothesis 6: The effect of competition on stability is negatively affected by the development of the regulatory institutions of a country. In the remaining of this paper when is spoken about institutional factors, it refers to all the factors; institutions, financial market, and regulators unless stated otherwise or is self-evident due to the context. 2.4. Systemic banking crises and competition This paper looks at realized systemic risk as an inverted proxy for financial stability. This realized excess risk is measured by using real life systemic banking crises during the sample period similar to the approach of Schaek, Cihak, and Wolfe (2009). The use of systemic banking crises is justified due to the mixed results from previous research which used the theoretical systemic risk. As a consequence it could be beneficial to find out why some risks materialize and why others do not. Both the competition fragility and stability views can and will be used in regard to systemic banking crises. The difference between the z-score used in other research and systemic banking crises used in this paper is that we can test the competition fragility and stability theories in practice. However one must bear in mind that realized risk does not take the unrealized risk into account which means that some countries could have faced risk which did not materialized in a systemic banking crisis. For further implications of this limitation I refer to the limitation section. The use of systemic banking crises instead of any other banking crises is necessary because competition has an effect on the entire market and standalone bank failures can be caused by many other reasons and therefore cannot be solely attributed to competition. In the remaining of the paper the broad definition of systemic banking crisis from Laeven and Valencia (2008) is used. In their definition there is a systemic banking crisis when a country s corporate and financial sectors experience a large number of defaults and financial institutions and corporation face great difficulties repaying contracts on time. Ex-ante we can expect a few results; the costs of a crisis is higher in more developed markets because they are more complex which results in higher costs. Further can be expected that developed countries have better financing options to react to a crisis which also attribute to the higher costs (Laeven and Valencia, 2012). 8

3. Methodology 3.1.1. Data The main data source of this paper is the Global Financial Development Database (GFDD) (2015) from the Worldbank. This database consists of characteristics of countries financial systems for 203 countries. Even though the time span of the database runs from 1960 till 2013 there is a lot of missing data especially in the period until 1998. Therefore I examine the period from 1999-2013. Furthermore some countries provided none or very little information which therefore have been excluded. A full list of countries can be found in appendix A. The remaining sample provides sufficient data for each variable, country, and year for a proper analysis and results can be generalized due to the variation in countries over multiple years. An overview of all variables, how they are measured, and which sources are used can be found in table 1. 3.1.2. Database systemic banking crises We combined the previous mentioned database with a recent database regarding systemic banking crises over the years. Laeven and Valencia (2012) created a database on the timing and resolution of all important banking crises from 1970-2012. This database contains 147 systemic banking crises as well as 218 currency crises and 66 sovereign crisis from 162 countries. While the aim was to examine policy implication during crisis times, the data provides a broad scope of information about the different aspects of each crisis which in combination with other data can be used for examining other factors than initially intended like competition. The information most important for the purpose of this research is when each crisis occurred. Furthermore it would be interesting to see whether competition has an effect on the severity of a crisis but this is not possible with the information available. 3.2. Variables 3.2.1. Z-score The z-score provides an inversed proxy of a bank s probability of failure (Berger, 2009). It combines the profitability, leverage, and return volatility into one index where a higher score means a lower probability on default. The reasoning behind this lies in the fact that high profitability and low volatility and leverage are presumed as characteristics of a non-risky and healthy bank. It measures the number of standard deviation the bank is from insolvency. 9

The equation is as follows: Z it = ROA it+ EQ it TAiT ROA (1) σ it i stands for bank i while the T stand for the time. ROA is return on assets for a certain year and EQ/TA is the percentage of equity on total assets whereas σ is the volatility of the return on assets for bank i at time T. The score used from the GFDD represent the median. 3.2.2. Lerner Index The Lerner index is a measure of competition based on the pricing power of a bank. It is a proxy for the current and future profits stemming from this pricing power (Beck, Schepens, and De Jonghe, 2013). It is measured by subtracting the marginal cost (MC i,t ) from the ratio of total operating income to total assets (P i,t ) before dividing it by again the total operating income to total assets and where i stand for a specific bank at time t. This results in a score between 0 and 1. Where 0 is perfect competition and 1 a monopoly. In perfect competition the price is similar to the marginal costs because banks compete on prices which drives the margin down and therefore the score will be 0. Similar to the z-score we use the median at a country level. The Lerner index can be written as: Lerner i,t = P i,t MC i,t P i,t (2) To examine whether the relation is non-linear, as proposed by the MMR model, the Lerner index needs to be adjusted. For this purpose the squared Lerner index is used in combination with the normal Lerner index. When both coefficients have opposing signs this is evidence of a non-linear relationship. 3.2.3. H-statistic and Boone indicator The Panzar and Rosse (1987) H-static is another measurement of competition. Fairly similar to the Lerner index this statistic measures the sum of all elasticities of banks revenues relative to their input prices (GFDD, 2015). In case of perfect competition the statistic should be 1 where the increase in input prices is equal to the increase in total revenue. This is due to the fact that there are no profit margins because they have been competed away. More roughly can be said that a score between 0 and 1 refers to monopolistic competition where the increase in revenue is partly explained by the increase in input prices. A score below 0 can be seen as a monopoly where firms can set their prices independent of the input price. Furthermore it is possible that 10

scores are greater than 1 which is the case in some oligopolistic markets (GFDD, 2015). A country s score is again the median of all scores in the same country. The Boone indicator provides the elasticity between marginal costs and profits (GFDD, 2015). In the GFDD this indicator is the coefficient of the log of profits regressed over the log of marginal costs. A country s score is the median of all scores in a country. The model is based on the efficiency of firms (Boone, 2004). Firms who are more efficient measured by lower marginal costs, gain more market share or profit. An advantage of this approach is that it is possible to examine market segments instead of the whole firm (Leuvensteijn, Bikker, Adrian, van Rixtel, and Sorensen, 2007). 3.2.4. Asset concentration Concentration shows a lot of similarities with competition but it refers to the market power of the largest banks. In situations with a few large players and a lot of small players the competition could be relative high but the market power of those large banks is too large to speak of fair competition. The asset concentration used in this research is the percentage of assets for the three largest banks compared to total assets. This is like the previous variables the median of each country. 3.2.5. Systemic banking crisis Systemic banking crises is a dummy variable which results in a zero when there is no systemic banking crisis and in a one for the years there is a systemic banking crisis. Once again we use the definition of Laeven and Valencia (2012) to identify a systemic banking crisis when a country s corporate and financial sector experience a large number of defaults and face great difficulties in repaying contracts on time. This data is provided out of the systemic banking crises database (2012). 3.3. Institutional, market, and regulatory development For these factors this paper follows the research of Beck, Schepens, and De Jonghe (2013) to a great extent because it offers a clear overview of the most important factors involving these factors. The estimation of the scores is done in the same way as their research and for an overview of which precise questions of the surveys are used I refer to their article. Discussions about the validity of the variables used is outside the scope of this research. The purpose is to check whether their results using the z-score hold with my sample and to see what the effect of institutional development is on systemic banking crises 11

For the institutional framework we use the depth of information sharing which provides an index for the amount of information credit agencies have. A higher value translates to more information. Credit agencies play an important part in determine whether a borrower receives a loan and the information sharing between both provides a good proxy of institutional development (Beck, Schepens, and De Jonghe, 2013). The development of the financial markets will be measured by using the stock market turnover. This is the ratio of the total value of shares traded to average market capitalization. A higher ratio means a more liquid market which in turn refers to a higher financial market development. Capital stringency refers to an index which is between 0-8 depending on how strict the rules regarding capital requirements are. The higher the score the stricter the rules. Stricter rules are intended to decrease bank risk taking and should therefore limit the effects of competition. Deposit insurance is measured by dividing the insured deposits by GDP. A higher value relates to a more generous deposit insurance. In addition this paper examines whether multiple supervisors play a role. Multiple supervisors can have an advantage over a single supervisor because of different approaches. This is a dummy variable which takes the value of 0 if case of a single supervisor and 1 in case of multiple. The last factor is activity restrictions which is an index between 4 and 16 where a higher score is the results of more restrictions. This index consists out of the four areas with a score of 1 till 4 where a higher score means more restrictions. The areas are insurance, securities, real estate, and voting shares in non-financial firms. To see whether these variables influence the effect of competition on stability, they are multiplied with the Lerner index. This provides a coefficient which enhances the effect of competition when it is positive and reduces the effect when negative. 3.4. Control variables This paper uses several control variables to control for factors which can influence the relation but are not relevant for this study. Three variables are being used and they are GDP per capita, stock market return, and non-interest income. They are provided by the GFDD and GDP per capita is measured as the weighted average while the stock market return and non-interest income are the median of each country per year. GDP per capita and stock market return are used to control for country specific economic development in a year and non-interest income gives an indication of the earnings model of banks in a country. Traditional banks earn an income through lending out deposited money and then receiving interest which can be seen as a relative safe strategy while other banks rely more on fee based activities. These activities include the securitization of mortgages which led to the financial crisis in 2008 and are more 12

volatile. According to Köhler (2014) more non-interest income increases systemic risk and to exclude this effect from the regression, non-interest income is included as a control variable. Furthermore the analysis makes use of time- as well as country fixed effects. These are used to exclude any variation which is caused by either time or country of origin. These fixed effects can be seen as a dummy where they are 1 in a specific year or country and zero in all the others. This procedure is done for every year and country so in the end there is a dummy for every time-country pair. Table 1 Summary of sources Variable Source Explanation Competition measures Lerner index GFDD Aggregated median at country level H-statistic GFDD Aggregated median at country level Boone indicator GFDD Aggregated median at country level Bank concentration (%) GFDD Aggregated median at country level Institutional development Depth of information sharing Doingbusiness.org Getting credit database score Stock market turnover ratio (%) GFDD Aggregated median at country level Capital stringency (0-8) Bank regulation and Index of capital requirements supervision database strictness Coverage limit / GDP per Capita (in %) Single bank supervisory = 0 Activity restrictions 3.5. Method Bank regulation and Coverage limit divided by GDP per capita supervision database Bank regulation and Dummy variable of 0 when single supervision database supervisor and 1 if multiple Bank regulation and index of activity restrictions imposed supervision database by regulator Depend variables Bank Z-score GFDD Aggregated median at country level Systemic banking crisis Systemic banking crisis database Systemic banking crisis dummy based on general definition by Laeven and Valencia (2008) Control variables GFDD Weighted average of country level GDP GDP per Capita in constant 2005 US Dollars Non-interest income GFDD Aggregated median at country level Stock market return (%) GFDD Aggregated median at country level This table provides an overview of each variable, their data source, and how they are measured This paper uses two regression methods to analyze the relationship between the variables explained in the previous parts of this study. The main methods are the ordinary least squares (OLS) regression and a binary logit model for the relationship between the different variables and the systemic banking crisis dummy. Whereas the OLS regression results in a coefficient 13

which gives the marginal increase of the dependent variable when the independent variable increases with one unit, the logit model gives a probability instead of the marginal change. Furthermore the R-square from the logit regression is a pseudo R-square which differs from the R² of an OLS regression. The interpretation is similar to the normal R² but the value is in most cases significantly lower. The equation in the first part of the analysis is as follows: Financial stability i,t = c + β competition i,t + γ control variables i,t + μ i + τ + ε i,t (3) Where β and γ are the coefficients for the competition and control variables, μ refers to the country fixed-effects, τ to the time fixed effects, and ϵ is the error term. Financial stability can either be the z-score or the systemic banking crisis dummy and competition is one of the competition variables. As always refers i and t to a specific country and time period. For the second part some extra variables are included namely the institutional development variables. Therefore the equation will be extended with an extra variable z and looks as follows: Financial stability i,t = c + β competition i,t + γ control variables i,t + z Lerner index i,t Institutional development i,t + μ i + τ + ε i,t (4) 3.6. Correlations Table 2 Correlation between competition measures Variables Lerner H-statistic Boone indicator Concentration Lerner 1 H-statistic -0.190 1 - (0.000) - Boone indicator 0.167-0.067 1 (0.002) (0.210) - Concentration 0.206-0.026 0.109 1 (0.000) (0.635) (0.043) - Table 2 presents the correlations between the different measures of competition. The first value is the correlation between the variables while the p-values are below between parentheses. The results are cross-country over the period 1999-2013. The Lerner index is a measure of competition defined as the pricing power of a bank. The H-statistic measures the elasticity between revenue and input prices while the Boone indicator measures the elasticity between marginal costs and profits. Concentration is the percentage of market share of the three largest banks measured in terms of assets. 14

The correlations between the various competition measures are shown in table 2. It can be seen that the Lerner index is significant correlated with all the other measures which is positive because in further analysis the Lerner index can be used instead of all. The highest correlation is between concentration and the Lerner index and has a value of.206 which does not provide any signs for multicollinearity and therefore can be ignored. Table 3 presents the correlations of the various institutional development variables. While some variables are significantly correlated, others are not. More important is there are no signs of multicollinearity with a highest significant correlation of.224 between depth of information sharing and stock market turnover. Therefore we can ignore the correlation between the independent variables. Table 3 Correlation between institutional development factors Variables Depth of information sharing Depth of information sharing Stock market turnover 0.224 1 Capital stringency 1 - Stock market turnover Capital stringency Coverage Limit Single bank supervisor Activity restriction (0.018) - -0.174 0.025 1 (0.068) (0.793) - Coverage Limit 0.093 0.000-0.045 1 (0.332) (0.997) (0.639) - Single bank supervisor -0.029-0.033 0.003 0.102 1 (0.763) (0.734) (0.974) (0.288) - Activity restriction -0.138-0.322 0.185 0.555-0.179 1 (0.150) (0.001) (0.052) (0.563) (0.061) - Table 3 presents the correlations between the institutional development factors. These factors can be divided into institutional factors; depth of information sharing, financial market development; stock market turnover, and supervisory development; capital stringency, coverage limit, single bank supervisor, and activity restrictions. These factors could influence the effect of competition on financial stability. The correlations are given with underneath their corresponding p-value between brackets. The values are crosscountry over the period between 1999-2013 3.7. Summary statistics The summary statistics can be found in table 4 and the results will be discussed here. Each variable is categorized into a type. The four types are; competition measures, institutional factors, control variables, and the dependent financial stability measures. For each variable the table shows the mean, standard deviation, minimum, maximum, and number of observations. The data also shows that there are no abnormal values, the dummy variables are all between 0 and 1 and also the index variable show no signs of errors. 15

Table 4 Summary statistics Variable Mean Standard dev. Minimum Maximum Observations Competition measures Lerner index 0.276 0.130 0.001 0.939 1883 H-statistic 0.607 0.271-0.562 2.028 354 Boone indicator -0.052 0.185-2.082 5.968 2346 Bank concentration (%) 74.414 21.519 7.248 100 2186 Institutional development Depth of information sharing 4.366 3.169 0 8 2790 Stock market turnover ratio (%) 47.210 62.926 0.009 511.672 1402 Capital stringency (0-8) 4.796 1.773 1 8 2055 Coverage limit / GDP per Capita (in %) 734 2017 20 8799 289 Single bank supervisory = 0 0.254 0.435 0.000 1 2009 Activity restrictions 10.597 2.195 4 16 1935 Depend variables Bank Z-score 15.294 10.639-21.224 74.129 2600 Systemic banking crisis 0.054 0.226 0 1 2639 Control variables GDP per Capita 11078 18007 133 158803 2856 Non-interest income 39.014 16.104 1.425 95.742 2628 Stock market return (%) 12.810 37.725-63.163 402.463 1145 This table describes all the variables used in this research stated per type. The first variables are the measures of competition used throughout this paper. They refer to the market power in a specific country. The next set of variables are the moderating factors which influence the relation between competition and stability. Further downwards there are the depend variables; the z-score or distance to default and the systemic banking crisis which is a dummy variable. Last there are the control variables, they are necessary to isolate the explanatory variables. 4. Analysis In the following section there will be an overview of the results. It starts with the analysis of the different competition measures on both the z-score as well as systemic banking crisis. After that the model will be expanded to include the institutional development factors and at the end there will be an analysis of every factor. 4.1. Competition and stability The results of the regression analysis of the competition measurements and financial stability measured through the z-score and systemic banking crisis can be found in table 5. In this regression I used the different competition measures in combination with the control variables. 16

This shows that the Lerner index and the H-statistic have a significant influence on the z-score, also known as the distance to default. The first has a positive effect while the latter has a negative effect. This means that a higher Lerner index increases the distance to default and a lower risk of default is similar to a higher financial stability. A higher Lerner score means more pricing power for a firm. This result is in compliance with the competition fragility theory, this means that more competition leads to less stability. Secondly the H-statistic is significant and negative, this result is similar to the Lerner index because the H-statistic has a reversed scale where 1 is perfect competition and 0 a monopoly. A lower H-statistic, less competition, increases the z-score and therefore the financial stability. The Boone-indicator and the concentration ratio both provide no significant results. The squared Lerner index gives no significant relation. There are no signs of non-linearity in the relation between the Lerner index and the z-score. When the competition measures are regressed against the occurrence of a systemic banking crisis all measures except the Boone-indicator provide significant results. The Lerner index and the H-statistic have their signs reversed which is consistent with the competition z-score relation. A higher Lerner index or a lower H-statistic reduces the occurrence of a systemic banking crisis which is a dummy of 1 when a crisis occurs and 0 in absence. Furthermore we see that concentration is negatively related to systemic banking crises, this implies that more concentration decrease the occurrence of a banking crisis. Again, this is consistent with the competition stability view where more concentration or market power leads to more stability. The squared Lerner index, in combination with the normal index, is significant and negative which is similar to the standard Lerner index. This result shows that an increase in the Lerner index, which is a decrease in competition, lowers the chance on a systemic crisis. The control variables have especially a significant influence in the systemic banking crisis regressions. These results are expected because GDP and stock market return are direct consequences of a financial crisis. It is however strange that GDP per capita has a positive effect on the occurrence of a systemic banking crisis. One possible explanation could be that the crisis years are only a small portion compared to no crisis years. The 2008 crisis affected mostly high GDP countries which could have led to an overrepresentation of high GDP countries and therefore explain the positive relation. Also the percentage of non-interest income has a significant and positive relation to the occurrence of a systemic banking crisis. This could also be explained by the 2008 crisis. This crisis started in the mortgage securitization sector where banks were highly involved in and earned a non-interest income. When the crisis hit the 17

financial world, banks which were involved in these non-interest income activities suffered the most and thereby explain why the percentage of non-interest income is positive related to the occurrence of a systemic banking crisis. Concluding can be said that there is evidence for the competition fragility theory and no evidence for the competition stability. Also there is no evidence for hypothesis 3 which states that concentration has a negative relation to financial stability while the regression shows a positive relation. 4.2. Institutional development The results of the regressions between the institutional development factors and financial stability can be found in table 6. First are all factors individually tested in combination with the control variables in order to provide an answer for hypotheses 4, 5, and 6 and secondly they are all combined into one regression. They are similar to the previous regression two part. In the first part, the z-score is taken as the dependent variable whereas in the second part, the banking crisis dummy is used. As already mentioned, this part leans heavily on Beck, De Jonghe, and Schepens (2013). The first observation is that only the depth of information sharing, activity restriction, and stock market turnover seem to have a significant impact on the Lerner-competition relation. Depth of information sharing seems consistent with expectations to lower the effect of competition, more information sharing however decreases the financial stability according to the results. Activity restriction has a positive effect on the competition stability relation, this means that when there are more restrictions the effect of the Lerner index becomes larger. The stock market turnover which is a proxy for the development of financial markets is positive and the implications of this are similar to the previous positive relation. That these relations are different from the expectations does not automatically provides evidence that they should rejected as will explained further on. 18

the control variables. Furthermore there is information about which effects are included as well as the R-squared and a pseudo R-squared for the logit regression. Last there is the method used, OLS or a Logit regression. Method OLS OLS OLS OLS OLS Logit Logit Logit Logit Logit The table above provides the regression results between the competition measures and both the z-score as well as the systemic banking crisis dummy. The variables Lerner, H-statistic, Boone indicator, Concentration, and Lerner square are the competition measures and GDP per Capita, Market return, and non-interest income are Lerner 5.571*** 5.562*** -2.767*** -2.861*** (1.967) (1.978) (0.799) (0.784) H-statistic -4.052*** 1.767* (1.244) (1.022) Boone indicator -1.214-0.5883 (0.907) (0.554) Concentration 2.101-0.856** (1.815) (0.435) Lerner² 0.853-2.369** (2.135) (0.973) GDP per Capita (%) 0.000* 0.000 0.000* 0.000 0.000* 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Market return -0.582 0.052-0.468-0.372 0.581-1.736*** -1.072-1.969*** -2.013*** -1.723*** (0.628) (0.642) (0.601) (0.613) (0.628) (0.413) (1.272) (0.417) (0.411) (0.414) Non-interest income -0.6762-9.655*** 0.3820-0.0157-0.663 1.807*** 3.098** 2.107*** 1.838*** 1.788*** (1.804) (3.679) (1.696) (1.710) (1.805) (0.646) (1.573) (0.628) (0.638) (0.649) Constant 12.589*** 30.432*** 13.188*** 11.925*** 12.500*** -2.276*** -4.011*** -3.194*** -2.359*** -2.121*** (2.250) (7.256) (2.082) (2.485) (2.262) (0.345) (1.767) (0.297) (0.437) (0.346) Number of observations 1015 222 1091 1064 1015 896 123 935 935 896 Time fixed effects yes yes yes yes yes yes yes yes yes yes Country fixed effects yes yes yes yes yes yes yes yes yes yes Control variables included yes yes yes yes yes yes yes yes yes yes R-squared 0.778 0.983 0.770 0.775 0.778 0.102 0.250 0.092 0.090 0.115 Table 5 Competition variables Variables Z-score Systemic banking crisis 1 2 3 4 5 6 7 8 9 10 19

Both in the depth of information sharing as the activity restriction the Lerner index is significant. However this is in both cases the opposite sign compared to the institutional factor. In the case of depth of information sharing, the Lerner index is positive and information sharing is negative while activity restriction is positive and the Lerner index is negative. When taking to the extreme, when there is no information sharing, less competition is necessary to increase stability because the negative value of information sharing does not need to be taken into account due to the value of zero. However it becomes interesting when there is a lot of information sharing because this lowers the financial stability in both high and low competition markets. This leads to the consequences that information sharing has a negative effect on the financial stability but this effect is larger when competition is less severe. In the second case the Lerner index is negative which evidence for the competition stability hypothesis is, but more activity restrictions have a positive effect on, in this case, the z-score. So when taken to the extreme, when there is no activity restrictions more competition is preferred because only the (negative) value of the initial Lerner coefficient has an effect. However when there are a lot of restrictions less competition is preferred. When we look at the economic implications of the result we see the following. In the case of low competition we do not need information sharing, one explanation could be that there is no need for sharing because the large banks have enough information about their customers and there is no need for sharing. Why more information sharing has a negative effect on the financial stability remains unclear and needs to be further investigated. When there is high competition the impact of information sharing is much smaller than in low competition and eventually diminishes. The case of activity restrictions is much clearer and as expected. More restrictions are preferred when competition is low and less restrictions when competition is high. This makes sense when looking at diversification theory, less restrictions provides banks the opportunity to diversify and thereby reducing their risk while restrictions decrease the diversify possibilities and increasing the risk. Especially in high competition markets this forms a problem because banks have less options to compensate for this risk due to their small market and pricing power compared to banks in low competition markets. This leaves still 3 factors insignificant, there are however possible explanations for this. One of the most used and well known measure to increase financial stability is the deposit insurance coverage. However there is no significant relation to either the theoretical stability measured by the z-score or the practical measure the systemic banking crises. Two explanations seem 20