The Competition-Stability Relationship in the Banking Sector

Transcription

1 Master Thesis Finance The Competition-Stability Relationship in the Banking Sector An event study analyzing the changes in risk following changes in competition policy for European banks Name: Nadieh Elferink ANR: S Supervisor: Dr. O.G. de Jonghe Graduation date: October 18, 2011 Department of Finance Tilburg School of Economics and Management

2 Abstract This research proposes to measure the risk effects of pro-competitive changes in merger control in order to shed some light on the competition-stability relationship in the banking sector. Four different methodologies are used to measure both wealth (abnormal returns) and risk (total risk, systematic risk and tail risk) effects of an increase in competition for a dataset of 328 banks in fifteen European countries in the period It is found that risk increases after the legislative change for some countries, while it decreases for others. To investigate why risk decreases for some banks, while it increases for others, cross-sectional regressions are done with eight different bank characteristics. This analysis shows that changes in systematic risk and tail risk are significantly influenced by the level of systematic risk of a bank before the change in merger control. In addition to this it is found that the other bank characteristics do not explain much or any of the changes in risk and return. Keywords: competition, stability, merger control, banks, risk effects. 2

3 Acknowledgements First of all, I would like to thank my supervisor of Tilburg University, Dr. O.G. de Jonghe, for his support and advice during the process of writing my Master Thesis. I truly appreciate his useful suggestions and comments, which guided me through the process of writing my Thesis. Furthermore, I would like to thank my friends and family for supporting me during the moments that I needed to relieve my feelings. A special word of thanks goes to Sjoerd Knufing, who was always there to listen, to cheer me up and to support me in getting the job done. Nadieh Elferink BSc. Tilburg, September

4 Table of contents PAGE 1) Introduction 6 2) The competition-stability trade-off Theory Competition-fragility hypothesis Competition-stability hypothesis 8 3) Empirical evidence on the competition-stability relationship Country-specific studies Cross-country studies 13 4) Using an exogenous event to analyze the impact of competition 17 on risk-taking 4.1 Competition policy Research setup Data Methodology Total Systematic Tail Abnormal Returns 24 5) Results Total Systematic Tail Abnormal Returns 28 6) Explaining bank-level variation Variables Expectations Capital Diversification Bank Size Loan Loss Provision Efficiency Return on Assets Asset Composition High Beta 37 7) Results bank-level analysis Ordinary Least Squares model Abnormal Returns Total Systematic Tail 45 4

5 7.2 Probit model Abnormal Returns Total Systematic Tail Robustness check 54 8) Conclusion 55 References 57 Appendix 61 5

6 1) Introduction The relationship between competition and stability in the banking sector is a topic that has been widely discussed over the past decades. Since the current economic crisis, which started as a financial crisis, it is again an issue that is high on the agenda of policymakers and academics. It is a widely recognized idea that the banking system is special from the perspective of stability. It is, however, a lot less clear what implications this special status has for market structures and competition policies. Not much research has been done yet on the implications different market structures and competition policies have on the stability of the banking sector. In this research, this issue will be addressed. Many studies have looked at the competition-stability relationship using different countries, different years and different measurements. No consensus has been reached yet on the sign of the relationship. Some studies argue competition harms stability; other - more recent studies - say competition has a beneficial effect on stability. Moreover, since different measures of competition and stability are used, comparing the existing literature is difficult. In this research, a new approach will be introduced. Changes in competition policy regarding the banking sector will affect the risk-taking behavior of banks and as such the stability of the whole sector. One of the main areas of competition policy is merger control, which affects the size and market power of firms. To analyze the impact of competition on the risk-taking behavior of banks, a change in merger control regulation will be used as an exogenous event to see whether the risk has increased or decreased after the legislative change. This will be done for fifteen industrialized countries in Europe over the last three decades. The paper is structured as follows. In the next section the existing theoretical literature on the competition-stability relationship will be summarized. The literature will be organized into two opposing views: the competition-fragility and the competition-stability hypothesis. Section 3 turns to the existing empirical literature, where a distinction will be made between country-specific and cross-country studies. In section 4 the application of competition policy in the banking sector will be discussed and the new approach, including data and methodology, will be introduced. Section 5 will discuss the tests and results. After that, in section 6 explanations for the bank-level variation will be discussed. In section 7, a crosssectional analysis will be performed. Section 8 will conclude. 6

7 2) The competition-stability trade-off - Theory Up to now, academics have not been able to come to an agreement on the relationship between bank competition and stability in the banking system. Different models, measurement methods and variables have been used to test this relationship. Many studies have only focused on one country or the comparison of two countries and do not control for the regulatory framework (Beck, 2008). From these studies, no clear conclusion on the relationship between competition and stability arises. Only since recently, a number of crosscountry studies have been done to assess the relationship between competition and stability. The evidence of these studies points mostly at a positive relationship between bank competition and stability, but yields mixed results on the relationship between concentration and stability (Beck, 2008). Furthermore, some studies use concentration as a proxy for a lack of competition (e.g. Smith, 1984), while other studies state concentration is not a consistent signal of competition (e.g. Matutes & Vives, 1996, Beck, Demirgüç-Kunt & Levine, 2006). Two main hypotheses have been brought forward in the existing literature: the competitionfragility hypothesis and the competition-stability hypothesis. In this section, the theoretical models discussing this relationship will be surveyed. 2.1) Competition-fragility hypothesis The competition-fragility hypothesis states that more competition among banks leads to more fragility. This view is also called the charter value view of banking, theoretically modeled by Marcus (1984), Chan, Greenbaum and Thakor (1986) and Keeley (1990) (Beck, 2008). This view sees banks as choosing the risk of their asset portfolio. As Beck (2008) states, banks have more incentives to take excessive risks in a more competitive environment where there is more pressure on profits. This results in higher fragility. Furthermore, banks earn less informational rents from their relationship with borrowers in a more competitive environment, which reduces their incentives to screen borrowers. This also harms stability. On the other hand, in an environment with limited competition, banks have better profit opportunities and more capital at hand and therefore fewer incentives to take excessive risks. This stimulates financial stability. Carletti (2005) states that under this hypothesis the general view is that higher deposit rates increase the probability of bank runs and that lower margins worsen the problem of excessive risk-taking. The general view under this hypothesis is thus that deregulation, resulting in more entry and competition, will lead to more fragility. 7

8 According to Vives (2010) competition may increase instability by (1) exacerbating the coordination problem of depositors/investors and fostering runs and/or panics which may be of systematic nature, and (2) increasing the incentives to take risk and raise failure probabilities. The latter channel is what we have seen in the arguments above. Regarding the first channel, Vives (2010) argues that runs can happen independently of the level of competition, but more competitive pressure worsens the coordination problem of depositors/investors and increases the potential instability and the probability of a crisis. Allen and Gale (2000, 2004) show another source of financial fragility: the possibility of contagion, which means that a small shock initially affects only one region or sector, but then spreads from bank to bank throughout the rest of the system and affects the entire economy. They explain that when there is perfect competition, each bank is small and a price taker. It assumes its actions have no effect on the equilibrium. No bank then has an incentive to provide liquidity to a troubled bank. Saez and Shi (2004), however, have argued that if there are a limited number of banks, they may have an incentive to act strategically and provide liquidity to the bank that is in trouble. Another argument that has been brought forward in the competition-fragility discussion is that more concentrated banking systems have larger banks, which allows them better diversifying their portfolios (Beck, 2008). However, this argument uses the effect of concentration on stability, not the competition-stability relationship. Nonetheless, it is an important side effect of market structure. In line with this is also the argument that a more concentrated banking system implies a smaller number of banks, which the supervisors can easier monitor and thus improves the overall banking system stability (Beck, 2008). 2.2) Competition-stability hypothesis The second hypothesis is the competition-stability hypothesis, which has arisen more recently and says that more competitive banking systems result in more, rather than less stability (Boyd & De Nicoló, 2005). This view is also called the risk-shifting view. Boyd and De Nicoló (2005) argue that the borrowers choose the riskiness of their investment undertaken with bank loans, rather than banks choosing the riskiness of their assets (as was stated under the charter value view). They show that more concentration in the loan market results in higher costs of borrowing for consumers, which decreases the success rate of their investment and makes it more likely that the borrower will default on its obligation. This effect is further reinforced by moral hazard on the part of the borrowers who increase their own risk of failure when confronted with higher interest costs. Boyd and De Nicoló (2005) thus find a positive 8

9 relationship between concentration and bank fragility. More recent extensions of the model of Boyd and De Nicoló (2005) that allow for imperfect correlation in loan defaults however show that the relationship between competition and risk is U-shaped (Martinez-Miera & Repullo, 2008). Caminal and Matutes (2002) show that less competition can lead to less credit rationing, larger loans and a higher probability of failure if loans are subject to multiplicative uncertainty (Beck, 2008). Furthermore, advocates of the competition-stability hypothesis argue that concentrated banking systems generally have fewer banks and that policymakers are more concerned about bank failures when there are only a few banks. Based on these assumptions, banks in concentrated systems will tend to receive large subsidies through implicit too-big-to-fail or to-important-to-fail policies. These implicit policies increase risk-taking incentives and hence decrease banking system stability (e.g. Mishkin, 1999). More competition could thus resolve the too-big-to-fail problem. A counter-argument to what was said in part 2.1, where it was stated that a smaller number of banks can be easier monitored and thus improves stability, is that a smaller number of banks tends to imply banks are larger, and larger banks are positively correlated with complexity. This means that large banks are harder to monitor than small banks (Beck, 2008). A final argument of Beck (2008) predicting a positive relationship between concentration and fragility is that the recent consolidation trend has led to financial conglomerates which offer a broad array of financial services, which were previously offered by specialized institutions. This is another complicating factor for bank supervisors that could harm stability. 9

10 3) Empirical evidence on the competition-stability relationship Having summarized the theoretical models discussing the relationship between competition and stability, this section will let the data speak. It surveys the empirical literature, which is relatively small. According to Carletti and Hartmann (2002) overall four types of studies can be distinguished. The first type regresses measures of bank risk on measures of market power. The second type assesses the potential diversification or risk reduction effects of combining different businesses in a merger or increasing bank size in other ways. The third type measures changes in bank stock return correlations as an indicator of the implications of consolidation for systematic risk. The fourth type discusses the relative efficiency and risk in bank sectors of different countries that are more or less competitive. But as Beck (2008) argues, up until recently the literature either only focused on one country or the comparison of two countries. Thus, the four types of study that Carletti and Hartmann (2002) mention are mainly country-specific studies. More recently, due to the availability of large cross-country, time series data sets, cross-country studies have come to play to assess the relationship between competition and stability. Another type of study has arisen with this new data availability, namely one that assesses why competition affects stability. Therefore, in this section a separation will be made between country-specific and cross-country studies and the existing literature of both types of study will be discussed shortly. 3.1) Country-specific studies In this subsection the existing country-specific literature will be classified in the four types of studies discussed by Carletti and Hartmann (2002). Type 1 studies discuss the effect of market power on risk. Keeley (1990) uses two pooled estimations in his study. First, capital-to-asset ratios for 85 large U.S. bank holding companies between 1971 and 1986 are regressed on their Tobin s Q (as a measure of market power) and a set of controls. The parameter of Q in this regression is positive and highly significant, which indicates that a more concentrated market (more market power) is associated with increased capital cushions in banks. Second, the effect of Tobin s Q on interest rates on large certificates of deposits (CDs) for 77 large bank holding companies between 1984 and 1986 is assessed. This time, the parameter of Q is negative and highly significant, which indicates that increased market power is associated with lower risk premiums, reflected in CD rates. Both estimations thus suggest that increased market power leads to a reduction in risk. In other words, more competition has lead to greater fragility during the 1980s, which is in support of the competition-fragility hypothesis (the charter 10

11 value view). Jiménez, Lopez and Saurina (2007) find for a sample of Spanish banks for the years 1988 to 2003 that banks with higher market power, which is measured by the Lerner index, have lower non-performing loans. This provides evidence for the charter value view. They do not find a significant relationship between market structure, as measured by concentration ratios, and non-performing loan ratios. Gan (2004) investigates the relationship between banking market structure and financial stability, but tests this relationship in two steps. He first tests whether competition reduces franchise value. Second, he tests whether reduced franchise value, combined with government deposit insurance guarantees, induces risk-taking. He used the Texas real estate collapse in the 1980s and its impact on the thrift industry as a natural, exogenous, experiment to test these hypotheses. Gan (2004) finds evidence for both his hypotheses: he shows that market concentration is positively related to franchise value and financial stability. Again, this provides evidence for the charter value view. Concluding, this type of studies seems to point at the competition-fragility hypothesis. Type 2 studies discuss the effect of mergers on diversification and risk and the effect of large banks on risk. If bank mergers diversify risk, then an increase in market power through concentration would be associated with lower risk and higher bank stability (Carletti & Hartmann, 2002). Paroush (1995) provides a description of how asset-side risk concentration can be diversified through a bank merger. By using the merger of Manufacturers Hanover Trust Co. and Chemical Bank in the U.S. as an example, he points at higher bank stability caused by increases in market power coming from diversification gains after mergers. Benston, Hunter and Wall (1995) test whether the behavior of takeover bid prices reflects risk diversifications as a motive for the acquisition or rather the increase of the deposit insurance put option value. They use data for 302 U.S. bank merger between 1981 and 1986 and find some evidence that is consistent with the former hypothesis (the behavior of takeover bid prices reflects risk diversification as a motive for the acquisition) and not with the latter. They also caution however that there is an older literature that argues that post-merger institutions change their portfolios to take on new risks and transform the diversification benefits into increased cash flows at ultimately unchanged risk of failure. Craig and Santos (1997) compare pre- and post-merger risk characteristics of 256 acquisitions by U.S. bank holding companies between 1984 and They find increased post-merger profitability and reduced postmerger risk. This would imply mergers decrease risk. Boyd and Runkle (1993) measure the sign of the relationship between bank size and bank risk for 122 U.S. bank holding companies for the years 1971 to They find that larger banks benefit from diversification advantages 11

12 (less volatile stock returns), but these advantages do not translate into a lower probability of failure. This could be due to the significantly higher leverage of larger banks. Boyd and Graham (1991, 1996) study whether large (more diversified) banks fail less often than small banks for the years 1971 to They find that over the entire sample the cumulate number of failures over all banks is 17% for large banks and 12% for small banks. When they divide the sample in sub-periods they find that the failure rate is higher for large banks during the earlier periods (1971 to 1986), while during the later period it is higher for small banks. The authors emphasize however that the results on the earlier periods may appear reversed when saved banks are not included in the failure category. De Nicoló (2000) studies the relationship between bank size and measures of charter value and insolvency risk in a sample of publicly traded banks in 21 industrialized countries for the years 1988 to He finds a positive and significant relationship between the size of banks and the probability of failure for banks in the U.S., Japan and several European countries. It is difficult to draw firm conclusions from these studies, but there are some indications that in more recent times concentrations resulting from mergers in the U.S. may have been associated with lower risk of individual banks after mergers, which is in line with the charter value view. However, other results indicate that larger banks do not necessarily fail less often than do smaller banks. Type 3 studies discuss the effect of consolidation on interbank linkages, which is an indicator of systematic risk. De Nicoló and Kwast (2002) argue that interbank linkages, as measured by correlations of stock returns, provide an indicator of systematic risk potential. They analyze the dynamics of the stock return correlations of a sample of U.S. large and complex banking organizations (LCBOs) over the years 1988 to 1999 and find a significant positive trend in stock return correlations. In addition, by estimating measures of the consolidation elasticity of correlation, they relate the return correlations of the firm to their consolidation activity. Consolidation at the sample appears to have contributed to the inter-dependencies, however also factors other than consolidation have been responsible for the upward trend in return correlations. These findings are consistent with the view that systematic risk in the U.S. banking system has increased over time, partly as a consequence of consolidation. Even though consolidation does not have to be an indication of less competition, this study does give a different picture from the individual bank risks discussed in the context of the charter value view or the diversification literature. 12

13 Type 4 studies compare pairs of countries to discuss whether countries with more competitive banking sectors face a greater risk of bank instability. Bordo, Redish and Rockoff (1996) compare the Canadian (few large banks) and the U.S. (many small banks) banking systems performances between 1920 and They observe a much greater stability of Canadian banks compared to U.S. banks and speculate that this might partly be due to the oligopolistic market structure of the Canadian banking sector. However, there is practically no evidence of higher monopoly rents for Canada compared to the U.S. Their observations suggest that due to their balance sheet structures, Canada had both a more stable and a more efficient (and not less competitive) banking sector than the U.S. during this period. Hoggarth, Milne and Wood (1998) compare the banking systems of the U.K. and Germany. Banking profits in the U.K. were consistently higher than in Germany, but also much more variable. Higher U.K. profitability can be explained by higher non-interest income and lower staff costs. Greater German stability can be explained by lower and more stable inflation as well as less competition. These two countries reflect two cases: one less competitive but more stable system (Germany) and one more competitive but less stable system (U.K.). Staikouras and Wood (2000) perform a similar analysis for Greece and Spain and find that Spanish banks as a whole are both more profitable and more stable than Greek banks, except for the Spanish commercial bank sub-group, which is less stable. This leads them to think that the Spanish banking sector is more competitive than the Greek one, which still has a larger public involvement. This is consistent with the competition-stability hypothesis, similar to the comparison between Canada and the U.S. above. The Germany-U.K. comparison on the other hand is consistent with the competition-fragility hypothesis. Overall, no single ever-valid relationship between competition and stability in the banking sector becomes clear from these studies. However, as Beck (2010) argues, two conclusions can be drawn from the existing literature. The first is that a higher degree of market concentration does not necessarily imply less competition. The second is that there is an important interaction effect between the regulator and supervisory framework on the one hand, and the market structure and competitiveness on the other hand, in their effect on banking system stability. 3.2) Cross-country studies As said, more recently several studies have attempted to provide cross-country evidence on the relationship between competition and stability. Berger, Klapper and Turk-Ariss (2008) test 13

14 the relationship between competition and stability by regressing measures of loan risk, bank risk and bank equity capital on several measures of market power, as well as indicators of the business environment. They use data for 8235 banks in 23 developed countries for the years 1999 to Their results suggest that banks with a higher degree of market power also have less overall risk exposure. This is in support of the competition-fragility hypothesis. However, the data also provides some support for one element of the competition-stability view: that market power does increase loan risk. This increased loan risk may however be offset in part by higher equity-capital ratios. Beck, Demirgüç-Kunt and Levine. (2006 a,b) use standard panel logit models to assess whether the probability that a country suffers a systemic crisis in a specific year depends on the concentration of the banking system, controlling for macro, financial, regulatory, institutional and cultural characteristics. They use data of 69 countries for the years 1980 to They find that systemic crises are less likely in concentrated banking systems. When Beck et al. (2006 a,b) analyze the channels through which concentration might be positively associated with banking system stability, they find evidence that more concentrated banking systems allow better possibilities for banks to diversify risk. They do not find evidence that it is easier for bank supervisors to monitor more concentrated banking systems or that the higher stability is the result of the market power and consequent franchise value of banks in more concentrated banking systems. This indicates that concentration does not proxy for a lack of competition. Boyd, De Nicoló and Loukoianova (2009) use a country-level dataset with individual bank data of 91 countries for the years 1980 to 2002 and find that more concentration leads to a higher probability of a systemic shock, but not a larger probability of government intervention. They claim that indicators of banking crises in the literature are actually indicators of the government response to the crisis. According to Boyd et al. (2009) the interpretation of the results in Beck et al. (2006) would be that more concentration leads to less intervention and more systemic crises, and that less barriers to entry lead to less intervention and less crises. Boyd, De Nicoló and Jalal (2009) arrive at a similar conclusion using individual bank data of a cross-sectional sample of 2500 U.S. banks in 2003 and a panel data set of about 2600 banks in 134 non-industrialized countries for the period 1993 to They find that the probability of failure is negatively and significantly related to measures of competition, so there is no trade-off between bank competition and stability, and that bank competition seems to foster banks willingness to lend. Beck, De Jonghe and Schepens (2010) assess whether the relationship between competition and stability varies across markets with different regulatory frameworks, market structures 14

15 and levels of institutional development. They use a sample of banks in 62 countries and find that, while holding the measure of competition and stability constant, support for either the competition-stability or competition-fragility view varies across countries and over time. Next, they identify the potential channels through which cross-country variation in the competition-stability relationship is created. They find that an increase in competition will have a larger impact on the risk-taking incentives of banks in countries that have stricter activity restrictions, more herding in revenue structure and unconcentrated banking markets. Schaeck, Cihák and Wolfe (2006) use the Panzar and Rosse H-Statistic as a measure for competition in 38 countries during the years 1980 to 2003 and find evidence that more competitive banking systems are less sensitive to systemic crises and that time to crisis is longer in a more competitive environment. Schaeck and Cihák (2007) identify bank capitalization as one of the channels through which competition stimulates stability. Schaeck and Cihák (2010) use a new measure of competition, the Boone (2008) indicator, to test the relationship between competition and stability. This indicator focuses on the impact of competition on the performance of efficient banks, and it provides an industrial organization based explanation for why competition fosters stability. They use two complementary datasets: a panel dataset for European banks covering the period 1995 to 2005 and a crosssectional sample of single-market banks operating in the U.S. in They first establish that the new competition indicator captures a broad variety of other characteristics of competition in a consistent manner. Second, they verify that competition increases efficiency. Third, they present novel evidence that the beneficial effect of competition on stability is due to a reallocation of profits. In other words, efficiency is the channel through which competition contributes to bank stability. Also, they find that smaller banks stability measures respond more strongly to competition than larger banks stability measures and that the soundnessenhancing effect of competition is larger in magnitude for sound banks than for fragile banks. Lastly, there is evidence that fewer regulatory restrictions regarding entry and banks activities are associated with less systemic fragility. Beck et al. (2006 a,b) and Barth, Caprio and Levine (2004) find that banking systems with more restrictions on banks activities and barriers to entry are more likely to suffer a crisis, while capital regulations are not significantly associated with the probability of suffering systemic distress. Restrictions on banks activities thus seem to harm rather than strengthen stability. This is in line with the competition-stability hypothesis. 15

16 Overall, the cross-country evidence seems to point at a positive relationship between competition and stability. This is somewhat contrasting with the mixed findings that arise from the country-specific studies. An explanation for this could be the fact that countryspecific studies do not control for the regulatory framework. 16

17 4) Using an exogenous event to analyze the impact of competition on risk-taking. Having discussed the existing theoretical and empirical evidence on the relationship between competition and stability, it is now time to start with a new approach. In this study, an exogenous change in competition policy will be used to analyze the effect of competition on the risk-taking behavior of banks. This research approach is based on the paper of Carletti, Hartmann and Ongena (2009), who analyze in an event study the impact of changes in merger control on the companies stock prices, distinguishing between the reaction of non-financial firms and banks. They then try to explain the differential effects on firm and bank stocks, by performing a cross-sectional analysis to directly test several economic hypotheses which may explain the different reactions in stock prices. In this section, the application of competition policy in the banking sector will be discussed first. Next, the new research setup will be introduced and motivated. And lastly, the data and methodology will be explained. 4.1) Competition policy Three main areas of competition policy are now fully applied to the banking sector: mergers, cartels and abuse of a dominant position (Carletti & Vives, 2007). A merger happens when two or more previously independent firms merge or when one firm acquirers the other, enabling it to exercise control. Cartels refer to any agreement and/or coordination of market behaviors between firms. An abuse of dominant position relates to any possible anticompetitive behavior exercised by one or more firms that have a dominant position in the market. Competition policy aims at preventing that any of these three elements leads to substantial restrictions or distortion of competition through, for example, prohibitions of mergers, or imposition of remedies and/or fines (Carletti & Vives, 2007). Carletti et al. (2009) have looked at the effect of legislative changes in merger control on the value of banks. They found that bank stock prices react positively to the announcement of pro-competitive changes in merger control. Transparency of the supervisory process is found to be the key driver of the positive reaction of bank abnormal returns: merger control plays an important function as a check and balance on the supervisory control. This result suggests that the effect of changes in merger control will be heavily influenced by the supervisory regime: the effect will be larger in less transparent supervisory systems. So, as Carletti et al. (2009) have found out, legal arrangements governing competition and supervisory control of bank mergers seem to have important implications for bank and firm performance in the economy. 17

18 The consensus today is that competition is not responsible for fragility in the banking sector, but the question on whether competition policy should be milder with market power in the banking industry remains open. The reason is that market power may have a moderating effect on the incentives to take risk (Carletti & Vives, 2007). Whether this is true, remains to be seen and will be the main topic of this research. 4.2) Research setup In this research, changes in competition policy (merger control) will be used to identify the effects of competition on stability. To be precise, the effects of legislative changes introducing or substantially modifying merger control regulation (which stimulates competition) on the banks risk-taking behavior will be assessed. Carletti et al. (2009) have studied the legislative changes affecting merger control regulation in 19 industrialized countries over the last three decades (1987 to 2004) and have constructed four indices that describe its main institutional characteristics. That data will also be used for this research. In an event study, the impact of the changes in merger control regulation on banks risk-taking will be analyzed. Banks risk will be measured as the volatility of the banks stock prices, as systematic risk (the beta estimated from a market model) and as tail risk (the returns in the extremes of the return distribution). The next step is to explain the differential effects on banks risk-taking. A cross-country analysis lies behind the scope of this research, but a bank-level analysis will be performed to test several economic hypotheses which may explain the positive or negative reactions (depending on the results of the first part) in banks risk-taking. This will be done by regressing the abnormal risk (the change in risk) on a number of variables capturing individual bank characteristics. This new setup, using an exogenous change in competition law to analyze the impact of competition on risk-taking, has several advantages. First, an exogenous event creates a clean environment in which to assess risk-taking behavior and allows investigating the relationship between market structure/competition and risk-taking behavior of banks. Second, it takes into account the substantial effects that that the introduction or changes in the merger policy may have on the investors expectations and thus stock prices. As Carletti et al. (2009) mention, by analyzing the period before and after a legislative change, one captures the investors 18

19 potential reaction to the entire political debate and process preceding and surrounding any important committee work. The focus of this study is not on abnormal returns, but on abnormal risk, but as stock prices are used to measure total, systematic and tail risk, it is important to take this into account. Third, with this setup one can distinguish the effects of merger control on risk-taking across different banks. This makes it possible to take into account the different banks characteristics, which can have an important interaction effect with merger control (Carletti et al. 2009). With this setup, it is tried to explain why competition increases stability in some cases, while it harms stability in others. 4.3) Data As mentioned before, the data that is used for this research is largely the same data as the paper of Carletti et al. (2009) on the impact of merger control, except that in this study only bank data is used. This research covers the years 1980 to 2005, with event date information until Only commercial banks, bank holdings, saving banks and cooperative banks in Europe are used: including the countries Austria, Belgium, Denmark, Finland, France, Greece, Ireland, Italy, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and the United Kingdom. Carletti et al. (2009) collected detailed information on the legislative changes affecting the institutional design of merger policy in each country in order to identify the event dates. For France, Norway, Spain and Sweden there were two events during the period of investigation. The event dates can be found in table 1 (column 3 to 6 of table 1 will be discussed in section 5). Balance sheet and Income Statement data are collected from Worldscope. Worldscope is a database containing accounting data on the world s leading public and private companies, compiled by Thomson Reuters and accessible via Datastream. The EU15+2 (Norway and Switzerland) was downloaded from Bankscope. Bankscope is a database containing financial information on banks worldwide based on publicly available data-sources, compiled by Fitch/Bureau van Dijck. From this sample, Luxembourg was deleted, since it was not part of the EU yet during the period Carletti et al. (2009) investigated merger control legislation in Europe. Furthermore, Germany was deleted from the sample, since no changes in merger control policy occurred there during the period of investigation. Lastly, also a few banks from the Faroes Islands were part of the sample. These islands are part of the Kingdom of Denmark, but are deleted from the sample, since their merger control legislation could differ from that in Denmark. 19

20 Table 1 Event dates and changes in risk and return. The first two columns of this table provide information on the dates of a pro-competitive change in merger control regulation for each country in the sample. For France, Norway, Spain and Sweden merger control regulation changed twice in the year 1980 to The third column contains information on the change in return variance of banks relative to the variance of a bank return index (total relative risk). Total relative risk: TRR. R j is the daily return on bank j and RB is the return on the bank index. TRR j = TRR j (after) TRR j (before), where after refers to days +10 to +260 after the regulatory change and before refers to days -260 to -10 before the regulatory change. The t-statistics test is used to test the hypothesis that TRR j = 0. The grey row tests TRR j = 0 for the full sample. The other rows test if TRR j = 0 for each country individually. The fourth column contains information on the change in Systematic. Systematic is measured as the change in the beta coefficient of the bank s stock return relative to the return on the bank index. The estimated model for the return of stock j on day t, R j,t, is,,,. D t is a dummy variable, equal to zero for days -260 to -10 before the regulatory change and equal to one for days +10 to +260 after the regulatory change. The change in beta, β j, is then defined as β j = β j (after) - β j (before) =. The t-statistics test used to test the hypothesis that = 0. This is done for the full sample (grey row) and for each country individually. The fifth column contains information on the change in Tail ; the risk in the lower tail of the distribution. For each bank, the six lowest returns in the year before the event are obtained. The highest return of these six returns is used as the benchmark and the other five returns are subtracted from the benchmark, to get excess extreme returns. Then the average of these five excess extreme returns is calculated. The same is done for six lowest returns in the year after the event. The result is an average tail risk for each bank before the event, and for each bank after the event. A t-statistic is then used to test the hypothesis that the change in tail risk is equal to zero. This is done for the full sample (grey row) and for each country individually. The sixth column contains information on the abnormal returns of the banks. Again, the model,,, is used, now D t being the variable of interest. Dt equals zero in the year before the event and one in the period after the event. The coefficient of D t answers the question whether the returns after the event are larger or smaller than the returns before the event. A t-statistic is used to test the hypothesis that, 0. This is done for the full sample (grey row) and for each country individually. Means for which the t-statistic is >1.645 are significant. Country name (1) Event date (2) Total (3) Systematic (4) Tail (5) Abnormal returns (6) Mean t-stat Mean t-stat Mean t-stat Mean t-stat Full sample 0,043 0,683-0,016-1,259-0,351-2,301 0,015 1,406 Austria Belgium Denmark Finland France France Greece Ireland Italy Netherlands Norway Norway Portugal Spain Spain Sweden Sweden Switzerland Great Britain

21 The daily stock price information (return index and volumes) was collected from Datastream. From this return index, daily returns are obtained. To arrive at the final sample, some more steps were taken. First, both listed as well as delisted banks are included in the sample, but for banks that have been delisted for a long period of time, the return index is deleted. Second, banks that have no data available for at least 100 days during the event window are deleted from the sample. Since France, Norway, Spain and Sweden have two events, they are included twice in the sample; once for the first event and once for the second event. This means that when a country does not have enough data available around the first event (which is very well possible, since most of the data is only available from 1990 onwards), but it does have enough data available around the second event; the country does not get deleted from the sample. The final sample consists of 15 countries, 328 banks and 19 events. This sample is merged with the event dates, a bank index (mnemonic BANKSEU in Datastream) and a risk free rate (the German 1 month interest rate). With the bank index, market returns are obtained. See table 2 for summary statistics of the returns and market returns. 4.4) Methodology The focus of this research is on risk-taking behavior of banks. will be measured in three ways, namely as total risk, as systematic risk and as tail risk. The total risk and systematic risk methodology is also used in the paper of Amihud, DeLong and Saunders (2002). For the analysis of risk, the banks risk one year after the regulatory change (the event) will be compared to the risk one year prior to the regulatory change. Specifically, data from 10 to 260 days after the regulatory change will be analyzed and compared to data from 10 to 260 days before the regulatory change. It will also be examined whether there are significant abnormal returns during the year after the regulatory change ) Total risk A measure of total risk that is often used is the variance of a firm s or bank s stock return. Amihud et al. (2002) examine the return variance of a bank relative to the variance of three bank return indexes and name this the total relative risk of bank j (TRR j ). In this research, the same approach will be used, but adapted to the topic that is discussed in this research. Since in this research the focus is on the effect of a regulatory change on risk and not on the effect of a bank merger on risk (as is the case in the paper of Amihud et al., 2002), it makes no sense to 21

22 Table 2 Summary statistics. Panel A of this table contain the summary statistics of the returns and market returns of all banks in the sample. Given are the mean, the standard deviation, the minimum and the maximum. Data is retrieved from Datastream. The rest of the table shows summary statistics for the bank specific variables used in this research. The bank specific data is retrieved from the Datastream Worldscope database. Panel B contains the dependent variables that have also been discussed in section 5 and earlier of this research. The means of the change in total risk, the change in systematic risk, the change in tail risk and the change in return are the same numbers as the means in columns 3, 4, 5 and 6 respectively of table 1. Panel C contains the independent variables, for which it will be assessed which effect they have on the dependent variables. For each bank, the variable information of three years before the event date is used. An average of these three years is taken to construct the value of each variable for each bank in the sample. CAP is the capital ratio and is constructed as Common Equity/Total Assets. DIV is income diversification and is constructed as 1-[(Net Interest Income Other Operating Income)/Total Operating Income]. Because of some large outliers, DIV is winsorized at p(0.05). SIZE is constructed as the natural logarithm of Total Assets. LLP is loan loss provision and is computed as Loan Loss Provision/Total Loans. EFF is efficiency and is measured as the sum of Other Operating Expenses and Interest Expense, divided by the sum of Commission and Fees, Interest Income and Other Operating Income. ROA is measured as the return on assets. Because of some large outliers, ROA is winsorized at p(0.05). COMP is the asset composition and is defined as Total Loans/Total Assets. HIBETA is a dummy variable which equals one if the bank s preevent systematic risk is above the 0.75 percentile and zero otherwise. Variable Mean St.Dev. Min Max Panel A Return Variables Return Market return Panel B Dependent Variables Total Systematic Tail Return Panel C Independent Variables CAP DIV SIZE LLP EFF ROA COMP HIBETA

23 examine the return variance of a bank relative to the variance of three bank return indexes (world, home and host), so it will only be examined relative to the variance of one bank return index (BANKSEU). In formula-form this is:, R j is the daily return on bank j and RB is the return on the bank index. The next step is to calculate the change in the total relative risk, TRR j, TRR j = TRR j (after) TRR j (before), Where after refers to days +10 to +260 after the regulatory change and before refers to days -260 to -10 before the regulatory change. A t-statistic will then be used to test the hypothesis that TRR j = 0. This will be done for the full sample, for each country and for each individual bank ) Systematic risk To measure the systematic risk of the banks the banks betas are used (e.g. Hansel & Krahnen, 2007 and Hendricks & Singhal, 1996). The change in the beta coefficient of the bank s stock return relative to the return on the market index will be calculated. The estimated model for the return of stock j on day t, R j,t, is (based on the model of Amihud et al, 2002):,,, D t is a dummy variable, equal to zero for days -260 to -10 before the regulatory change and equal to one for days +10 to +260 after the regulatory change. The change in beta, β j, is then defined as follows: β j = β j (after) - β j (before) = A t-statistic will then be used to test the hypothesis that = 0. This will be done for the full sample, for each country and for each individual bank ) Tail risk In the existing literature it is well known that financial returns tend to have empirical distributions that have fatter tails than the normal distribution (Huisman, Koedijk, Kool & Palm, 2001). This means extreme results are more likely. It is therefore important to have knowledge of the tail behavior of the returns, also called the tail index, since it says something about the extreme risks. To measure the extreme risk; the risk in the tails of the distribution, an approach is used that is derived from the Hill (1975) approach. Hill s (1975) estimator is given by 23