University of Groningen Faculty of Business and Economics
“Government Intervention to Increase the Financial Stability in the European Banking sector: Distortive Effects of Competition”
Master Thesis Finance Alberdien Meindertsma (s2044528)
Date: 18 June 2014 Word count: 8128 Supervisor: L.J. Dam
Keywords :Banking, Bailouts, Competition Jel Code: C30, D82, G21, G28, L51
Abstract
1
1. Introduction
In the recent financial crisis many government interventions in the appearance of bank bailouts have taken place. As a response on this bank bailouts during the crisis, the European Prime-Ministers have signed an agreement of the Banking Union in December 2013. The pillars of this agreement are that there is more supervision in the European banking sector, another strategy is used for rescuing banks when they become in distress and a new deposit guarantee system is used. In this new agreement, when banks becomes in distress, there will be a bail-in, which means that first shareholders, bond holders and big savers suffer for the losses and the recovery of the bank. If this is not sufficient, money will be subtracted from an national emergency account, which has to be filled by the banking sector themselves. 1
When banks become in distress the regulator faces the choice to interfere with a bailout or to let the bank fail. The regulator tries to implement policies which maintain confidence in the financial sector and acts in a way that the contagions of the crisis will be minimized (Hryckiewicz 2012). Implementation of this stability policies, by giving capital injections and guarantees, have also side effects. A lot of research has been done on the concept of moral hazard in the banking industry. This moral hazard effect implies that when a bank has a high expected bailout probability, it misuses the fact that they will be saved by the regulator and as a result, increase their excessive risk taking behavior. In the European banking industry the presence of moral hazard is shown by several authors (Dam an Koetter 2012, Rottman 2012). Besides moral hazard, there is another side effect of government intervention, which is the distortion of the competitive environment. It has been shown that that banks generate more market power when they are implicitly guaranteed with a bailout of the government (Koetter and Noth (2012)) and that non-guaranteed banks face more competition when bailouts take place (Grop Hakenes and Schnabel (2010)).
In the European banking sector, to my knowledge, there is virtually no analysis done yet of the effects of the distortion of the competitive environment by the regulators’ intervention. If it turns out that there is a distortion of the competitive environment in the form of more market power of one group related to the other groups, this can have adverse consequences for the new agreement of a Banking Union. The implementation of this Banking Union may lead to more stability in the European banking industry, since the banking industry itself has to pay for potential losses. However the union may also lead to even more risk taking behavior, since no one is prepared to pay for another bank receiving
2 ‘free’ market power, which creates the wrong incentives with respect to the risk taking behavior of the banking industry.
This paper analysis the impact of government intervention regarding their bailout policy on the competitive environment in the European banking industry. Using the method of Koetter and Dam (2012) the bailout probability is estimated, which is analyzed in relationship with the Lerner index, a measure of competition. This relationship is tested from different perspectives and different variables are used. Furthermore, if a relationship is found, it is examined in a next step if the distortion of competition is caused by increasing risk taking behavior. As such, this second question also addresses if bank bailouts cause moral hazard.
My results show that government intervention in the form of bailouts indeed affect the competitive environment in the European banking industry. On average, there is a significant negative relationship with the bailout probability and the Lerner index, which means that if a company has a higher expected bailout probability his market power will decline, which implies a more competitive market. This increase in competition can be attributed to the increasing competition from the non-guaranteed banks. This effect is however dependent on the size of the bank. Especially smaller and medium banks enjoy more market power if they are confronted with a higher bailout probability, in contrast to the larger banks, whose market power decreases. Finally, it is analyzed if this distortion of competition is due to the presence of moral hazard. It is shown that moral hazard exists in the European banking sector and that the notion of the ‘Too Big to Fail’ is underlined.
In the next section, the literature regarding the competition in the banking industry and the effects of government intervention will be discussed. In section 3 and 4, the methodology and information about the data respectively will be given. In section 5 the results of my analysis are presented and finally a conclusion is given with a short discussion.
2. Literature review
2.1 Competition in the banking industry
3 maximizing. Too high prices are charged which leads to a dead weight loss compared to a perfect competitive market. If competition increases, profits of the banks are put under pressure. As a result the bank may take more excessive risk and stop screening their clients thoroughly. In effect, the bank operates riskier, which negatively influences the stability of the bank and therefore the financial sector. An agency problem comes into existence which is called the risk shifting problem defined by Jensen and Meckling (1976). When due to higher competition the margins of the firms become lower, the equity holders want to replace the asset base in a more risky asset base, to be more sure of a profit.
2.2 Risk taking behavior
According to Grop, Hakenes and Schnabel (2010) there are two channels which determine the risk taking behavior of the banks. First, due to bank guarantees, the creditors of the bank have no incentives to monitor the guaranteed bank on their risk taking behavior, since they know that the bank is protected. Less monitoring by investors, increases the excess risk taking behavior of banks. On the other side, the bank’s risk taking behavior is also determined by the effect of guarantees on the banks’ charter value and his profit, called the charter value theory. Both channels determine the decision of increasing the risk taking behavior and the importance of each channel is dependent of the bank characteristics.
Keeley (1990) also discusses this risk shifting problem when competition increases . He argues that the choice of risk taking behavior of the banks is not only determined by the level of competition nowadays, but also the competition a bank will face in the future. The bank faces a tradeoff between the expected gains due to more riskier behavior and the expected losses on the charter value of the bank, which is the present value of the future rents. Knowing that it faces less competition in the future, which increase the charter value of the bank, they are moving away from risky behavior. It could be argued that banks, knowing that they will have a bailout in the future, are sure of increased market power due to lower funding costs. Since they are sure of less competition in the future they make the tradeoff which is described by Keeley (1990) and will not take excessive risk.
4 possible contagion effects of bankruptcies (Morrison 2007, Hughes and Mester 1993). The extent to which the creditors of the bank are monitoring the risk taking behavior depends on the incentives to monitor and the availability of the information. Bushman and Williams (2007) analyze to what extent financial accounting information contributes to the risk taking behavior of banks. They used earnings smoothing via the loans provision as an indicator of the transparency of the banks’ risk taking behavior. In countries where more earning smoothing takes place, banks exhibit more shifts in their risk taking behavior. They conclude that this practice leads to less ability of information for investors and creditors to judge and monitor the risk taking behavior of the banks. So if banks are not monitored closely by their investors, they are increasing their risk taking behavior. Furthermore, if creditors and investors expect that the ongoing existence of banks is guaranteed and potential losses are compensated, the incentives to closely monitor the bank’s behavior decreases.
2.3 Effect of government intervention on competition
An increase in risk taking behavior also has effects on the competition between banks. Norden, Roosenboom and Wang (2012) investigate how government interventions in the banking industry in the United States influence the stock returns of corporate borrowers of that bank. They conclude that government interventions have a positive influence on the stock return of the borrowing firms. As part of the Troubles Asset Relief Program (TARP) the American government supported banks with additional equity with the purpose that the banks were able to provide loans to their customers, which was aimed at boosting the economy. (U.S. Department of Treasury, October 14, 2008). Norden et al (2012) found that especially the smaller and riskier clients had increased stock performance. This is evidence for a competitive advantage of the banks which will have government support. They have more liquidity and can take more risk.
The risk taking behavior of banks is also analyzed by Hakenes and Schnabel (2004). They analyze the effects of government bailouts on the competition in the banking industry. They use a theoretical model to analyze the effects on the competition in a transparent and opaque market for both the bailed out bank and the non-guaranteed bank. The key ingredient of their model is the risk shifting problem, which is first described by Jensen and Meckling (1976). When due to higher competition the margins of the non-guaranteed bank become lower, the equity holders wants to shift the asset base into a more risky asset base, to increase profits. Hakenes and Schnabel (2004) find that the banks which do not have a guarantee will increase their risk taking behavior. The increased competition due to the government guarantees forces the smaller banks to increase their risk taking behavior.
5 of insured competitor banks (MSI) and the risk taking of the banks. They find that a higher MSI increases the banks risk taking behavior. The banks own bailout probability does not significantly influence the risk taking behavior or it even decreases the risk taking behavior significantly. They assigned this results as evidence of the charter value theory. Furthermore, they distinct their sample in public banks and private banks. They found that private banks take less risk if they have a higher bailout probability, in contrast to public banks, which increase their risk taking behavior if they have a higher bailout probability. A possible explanation for this phenomenon is that public banks care less about their charter value than private banks. The results they found have important implications for government policy. Government guarantees for banks which are in distress during the financial crisis, can lead to less stability in the banking sector in the future. They argue that the expectations regarding a bailout probability of the banks have to be reduced. The point is that not only the real bank bailout probability, but also the expected bailout probability can affect the competitive environment in the banking industry.
2.4 Distortive effects on competition by government intervention in the European banking sector From the literature discussed it can be concluded that government intervention in the form of bailouts indeed distorts the competitive environment in the banking sector. To my knowledge, no research has been done yet to the distortive effects of government intervention in the European banking sector. Differences between areas, for example the United States and Europe, can be expected due to differences in regulation and the level of competition in the banking industry. Bikker and Haaf (2000) did some research on structural changes in the European banking sector and compared the level of competition between sectors. They show that there is much diversification in level of concentration of banks across the world. Therefore, since there are different levels of competition between countries and continents, it is useful to investigate each area separately. One can imagine that the current level of competition in the banking industry also determines the effects of government intervention on the competitive environment regarding their bailout policy.
3. Methodology
3.1 Estimation of effects of government intervention
6 which is used to determine the Lerner index, the Cost to Income Ratio is used to estimate this measure. The Lerner index can be calculated as:
𝐿𝐼 = 1 − 𝐶𝑜𝑠𝑡 𝑡𝑜 𝐼𝑛𝑐𝑜𝑚𝑒 𝑅𝑎𝑡𝑖𝑜 (1)
In the second step, the relationship between the Lerner index and the lagged expected bailout probability, 𝜋𝑖,𝑡, and the lagged banks’ characteristics, 𝑋′ 𝑖,𝑡 , are estimated using Ordinary Least
Squares. Lagged variables on the bank level and macroeconomic level are used since it is assumed that current expectations on the bailout probability will lead to changes in the risk taking behavior, which is revealed at the subsequent year. The relationship is estimated as:
LI = b0 + 𝑏1𝜋𝑖,𝑡−1 +𝑏2 𝑋′ 𝑖,𝑡−1 (2)
3.2 Estimation of the bailout probability and distress probability
To determine the bailout probability for each bank the two step model of Dam and Koetter (2011) is used. The model of Dam and Koetter (2012) is also used for analyzing moral hazard. A simple regression on the relationship between a bank’s risk taking behavior and bailout characteristics would create endogeneity problems. To prevent this Dam and Koetter (2011) developed a two-step model, which could deal with this problems. In their model they assume that a bank chooses its risk position at T=0. The consequences of this risk position are revealed at T=1, which will result in a distress or in a sound state. If a bank will end up in the distress state, it has two options. Either the bank will be saved by the regulator or the bank will fail and exit from the market. The regulator does not announce a priori which banks they save when these become in distress, since it would increase moral hazard. However, based on observed bank characteristics and historical bailouts, a bank can form expectations on their bailout probability. Knowing the level of this bailout probability may influence the decision about their risk position at T=0. The bank chooses the risk position which maximizes his profit. To check if there is moral hazard involved, the relationship between the expected bailout probability and the distress probability has to be estimated in a second step.
The first step is to determine the expected bailout probability, 𝜋𝑖,𝑡, , using the bank characteristics,
𝑋′𝑖𝑡, and macro-economic variables, 𝑀′𝑖𝑡 and political factors, 𝑍′𝑖𝑡. The bank’s characteristics are
7 Where 𝐼𝑖,𝑡 equal to one means that a bank have had a bailout from the regulator and 𝐷𝑖,𝑡 equal to one
means that the bank was in distress.
The expected bailout probability is estimated only making use of the banks in the sample which were in distress during the time period used in the sample. A dummy variable is specified in the case that the bank has received a bailout from the government. When the equation for the bailout probability is estimated, the expected bailout probability is calculated for each bank.
The second step is to estimate the distress probability, P(zi,t) ,using the banks expected bailout probability, the banks characteristics, 𝑋′𝑖𝑡 , and the macro economic variables, 𝑀′𝑖𝑡. The distress
probability reflects the bank’s risk position. The distress probability can be calculated as
'
1
1
'
1
2
z
itD
it itX
itM
it (4)If it can be shown that the expected bailout probability calculated in equation (3) influences the risk taking behavior in a positive way in equation (4) moral hazard is identified.
3.3 Sensitivity analysis
To test the robustness of the results, different bank specific variables are used to estimate the bailout probability, the distress probability and the relationship between the Lerner index and the bailout probability. Furthermore, the sample is divided in groups of different sizes, since the “Too Big to Fail” concept is underlined by different authors and could be expected to be present in my sample.
4. Data
4.1
Sample and events
8 either case, the bank exits from the market. The sample consists of 71 distressed banks, of which 52 banks have received a bailout. Virtually all bailouts occurred during the financial crisis. An overview of the list of distressed banks is given in Appendix Table 1 and Table 2. In Table 1 it can be seen that about 70% of failures and bailouts of bank occurred during the years 2008 and 2009. This was the period when the banking industry in Europe suffered the most from the financial crisis.
Table 1 Distressed banks: Amount of bailouts and failures of European banks during the years 2001 until 2013
Year Distress Bailout Failure
2001 0 0 0 2002 0 0 0 2003 0 0 0 2005 0 0 0 2005 0 0 0 2006 0 0 0 2007 1 1 0 2008 31 24 7 2009 21 16 5 2010 4 2 2 2011 6 2 4 2012 6 5 1 2013 2 2 0 Total 71 52 19
Note: The variables can be defined as follows: Distress can be defined as the occurrence of a bailout or exit of a bank from the market. A bailout is a situation in which the government interferes with a bailout and as a result the bank is able to survive. A failure is a situation in which a bank exits from the market or mergers with another bank. The data about bank bailouts and bank failures are gathered from Rotmann (2012) and are updated using news websites.
4.2 Bank level and macroeconomic variables
Several variables are used to estimate the bailout probability and the distress likelihood, which can be seen in Table 2. These variables are both on the bank level and the macroeconomic level. The characteristics of the bank indicate the riskiness of the bank, while macro-economic variables give both information on the environment in which the bank operates and the likelihood that the government interferes.
9 consequences for the financial system if a large bank fails when it considers a bailout. The regulator is confronted with the “Too Big to Fail” concept, which means that larger banks are sure of a bailout since a failure has a lot of negative externalities for the financial system (Morrison 2011).
Table 2 Descriptive statistics: The descriptive statistics of European bank characteristics and European macro-economic
variables for the period of 2001 until 2013
Note: The variables represented in the table are used for determining the relationship with competition and the bailout
probability, estimating the bailout probability and examining the relationship between the bailout probability and the likelihood of running into distress. The descriptive statistics of the bank-level characteristics are gathered from Bankscope and the Macro-economic variables are gathered per country from Eurostat, which is provider of high quality statistics about information in European countries.
Macro-economic level variables are chosen which reflect the state of the economy. These variables can influence the behavior of the bank and the government regarding their bailout decision. Furthermore, in a bad environment it is more likely that a bank becomes distressed. Similar to Dam and Koetter (2012) the GDP growth rate and Unemployment rate are used, which are gathered from Eurostat. There is a wide dispersion in the unemployment rate and GDP growth between the countries and across time. The more southern countries, for example Greece and Spain, face a much higher unemployment rate than the Northern countries in Europe, such as the Netherlands and Sweden. After the financial crisis the unemployment rates increased enormously in the Southern part of Europe up to an unemployment rate of 25% in Spain.
To examine the relationship between government intervention and the competitiveness of banks the Lerner index is used as a measure of competition. The Lerner index measures the market power of the bank, by measuring the difference between the price and the marginal costs, relative to the marginal
Variables Mean Std. Dev. Min Max Obs
Bank-level variables
Net Loans to Total Assets 55.952 23.662 -0.039 100 45430
Capital Ratio 18.242 20.000 -101.380 869 18578 Equity to Assets 12.299 19.004 -775.000 138.017 48020 Total Assets 5.876 0.911 1.262 9.413 48065 Return on assets 0.616 6.097 -348.070 274.540 47821 Lerner Index 30.736 36.877 -884.19 100 46396 Macro-economic
Growth Domestic Product Growth 2.231 3.716 -17.7 11 204
10 costs. The higher the Lerner index, the higher the banks market power and the better it is able to be competitive. The Lerner index is chosen since it measures the competitiveness of each bank and is dynamic in time. However, since the Lerner index is not available on Bankscope and the Lerner index using the method of Koetter and Noth (2012) is hard to estimate, an alternative measures is used to come to the Lerner Index by transforming the Cost to Income ratio.
5. Results
First the main question of this paper, namely whether government intervention disturbs competition is analyzed. The first step is to estimate the bailout probability for each bank using the method of Dam and Koetter (2012). The analysis is done twice with different variables. In specification [1] the Total Capital Ratio is used as a measure of the bank’s buffer and leverage position. In specification [2] the Equity to Total Assets parameter is used instead of the Total Capital Ratio. This distinction in variables is made since the number of observations will increase when the variable Equity to Total Assets is used. The variables represent more or less the same risk and more observations make the analyses more reliable. The coefficients of these equations are estimated using the Probit model and the Ordinary Least Squares model. The latter is used since it is better in explaining the economic significance, however, since the distribution of the bailout probability does not met the conditions that the bailout probability of a bank lies between zero and one. Therefore, more weight is attached to the results of the probit model. The ordinary least square results are shown in the Appendix Table 3.
5.1 Bailout probability
The results of the estimation of the bailout probability are shown in Table 3. Based on a subsample of failed banks and banks which had a bailout during the sample period, a bailout probability can be extrapolated which can be calculated for each bank in the sample. So, for all banks a bailout probability will be calculated, even for those banks which did not have had a bailout yet. What can be seen from the table is that size is an important variable of estimating the bailout probability. There is a positive significant effect of the size of the banks, measured as the log of the total assets, on the bailout probability of the bank. The larger the bank, the more certain it is that they will have a bailout, which could be a support for the ‘too big to fail’ concept.
11 that such a statement is appropriate to conclude, since the likelihood that profitable banks will become distressed is low. As a robustness check, the variable return on assets is omitted from the equation and the results are shown in Appendix Table 4. There are no significant changes in signs or significance level, so the results are robust.
12 Table 3 Bailout estimation: Results of the estimation of the bailout probability of the subsample consisting of distressed
banks for the years 2007 until 2013.
Note: The bailout probability is estimated from a subsample which consists of banks which have had a bailout or banks which failed during the sample period. By extrapolating this data, the bailout probability can be calculated for each bank in the sample, even if the bank did not have had a bailout. The explanatory covariates are variables which are likely to contribute to the level of the bailout probability of each bank. The variable political orientation government dummy is equal to one if the government is right hand orientated. If there is a year of elections, the election year dummy is equal to one. The number of observations between specification [1] and specification [2] differ due to differences in the availability of information regarding the variables Equity to Total Assets and Total Capital Ratio. */**/*** denote the significance at the 10%/5%/1% levels respectively.
Specification [1] [2]
Dependent variable Bailout Bailout
Explanatory covariates
Bailout Probability
Total Capital Ratio (%) 0.046
(0.59)
Equity to Total Assets (%) 0.106
(0.84)
Net Loans to Total Assets (%) 0.021
(0.99)
0.016 (0.78)
Total Assets (th EUR) (log) 1.384***
(2.82) 1.210*** (2.79) Return on Assets (%) 0.145* (1.77) 0.136** (1.87) GDP Growth Rate (%) -0.000 (-0.00) -0.002 (-0.01) Unemployment Rate (%) 0.257** (2.51) 0.217** (2.50)
Political Orientation Government (dummy) 0.978
(1.54)
0.931 (1.52)
Election Year (dummy) 0.070
13 5.2 Distortive effects regarding competition
In Table 4 the results are shown of how the regulators intervention distorts the competitive environment for both specifications. The competitive environment is measured using the Lerner index of each bank. The relationship between the bailout probability and the Lerner index is therefore analyzed. For both specification [1] and [2] there is a negative significant relationship, of b=-8.301 and b=-2.163 respectively. If a bank faces a higher chance of being bailed out, their competitiveness will be lowered. They do not profit from the status of having a higher bailout probability. All the coefficients of the variables in both specification are consistent except for the unemployment rate. Specification [1] shows a positive significant relationship between the Unemployment Rate and the Lerner index and specification [2] shows a negative significant results. The difference in these coefficients can be explained by the fact that different samples are used for both specifications since data about the
capital ratio is less available.
Table 4 Distortive effects: Results of the regression of the relationship between the Lerner index and the Bailout probability
Note: The relationship between the bailout probability and the Lerner index, as a measure of competition, is estimated for the sample. This relationship is tested twice using the variables Total Capital Ratio and Equity to Total Assets as measures of the risk buffer of a bank separately. */**/*** denote the significance at the 10%/5%/1% levels respectively.
Specification [1] [2]
Dependent variable Lerner index Lerner index
Explanatory covariates
Bailout Probability -8.301***
(-4.64)
-2.163** (-2.15)
Total Capital Ratio (%) 0.120***
(6.16)
Equity to Total Assets (%) 0.117***
(4.97)
Net Loans to Total Assets (%) 0.226***
(16.88)
0.180*** (23.13)
Total Assets (th EUR) (log) 5.940***
14 5.3 Distortive effects regarding to competition for different sizes
Koetter and Noth (2012) found that small and medium bank profit more from a higher bailout probability, in terms of market power. This difference in relationship is also test in my sample of European banks. The definition of the European Central Bank (ECB) of a small, medium and large bank is used when the sample is divided. According to the ECB a large bank is a bank which has more than 0,5% of assets of the total assets in the European banking sector. A medium bank has assets between 0.5% and 0.005% of the total assets in the European banking sector and a small bank has less than 0.005% of the total assets.
Table 5 Competitive distortions: Results of the estimation of the relationship between the Lerner index and the bailout
probability for different sizes of banks
Note: The sample is divided in three subsamples regarding their size, for which the relationship is estimated between the Lerner index and the bailout probability for each category. */**/*** denote the significance at the 10%/5%/1% levels respectively.
Table 5 shows the results of this analysis. The coefficients for each category of banks differ a lot. Small and medium banks enjoy higher mark ups if they have a higher bailout probability with b=0.872 and b=1.870 respectively, though not significant. The larger banks have a negative significant relationship between the Lerner index and the bailout probability. If their bailout probability increases with one
Specification Small Mediun Large
Dependent variable Lerner index Lerner index Lerner index
Explanatory covariates Bailout Probability 0.872 (0.55) 1.870 (1.26) -14.501* (-1.92)
Equity to Total Assets (%) 0.035 (1.13)
0.299*** (7.05)
0.702** (2.09)
Net Loans to Total Assets (%) 0.210*** (19.72)
0.098*** (8.12)
0.244*** (4.26)
15 percent, the Lerner index decreases with 14,5%. This results are in line with the conclusions of Koetter and Noth (2012). An economic argument for the high negative coefficient for the larger banks could be that since they have a higher bailout probability, they face more regulation, which lowers their market power. The regulator could fear stability problems in the banking sector and want to prevent that the bank will fails in light of potential spillover effects. The small and medium banks are less under the attention of the regulator and as a result will be less closely monitored.
5.4 Competition effects if bailouts take place
According to Hakenes and Schnabel (2004) banks that do not receive any government support face more competition when a bailout has taken place. Banks who know that the likelihood of having a bailout is high in the case of distress, take excess risk, which leads to unfair competition. Due to lower funding costs of guaranteed banks, non-guaranteed banks face more competition. To analyze if this effect is present in Europe an analysis is done. The relationship between the Lerner index and the dummy variable of the distress situation are analyzed for the years from 2007 until 2013.
The results of this analysis are shown in Table 6. It shows that the dummy variable distressed has a significant coefficient of b=-19.576. If the dummy variable is one it is defined as a bank in distress. It
means that a bank which was is in distress, and
16 Table 6 Competitive distortions: Results of the estimation of the competitive distortions for non-guaranteed banks , in the
period where bailouts take place
Note: A ordinary least squares regression is done to test the relationship between the Lerner index and the effect of a bailout on the competitive environment. The variable Distressed equal to one represents a bank which is in distress. Distress means that the bank needs a bailout from the regulator, otherwise it will exit the market. */**/*** denote the significance at the 10%/5%/1% levels respectively.
5.5 Moral Hazard
To shed some light on whether there is indeed moral hazard in the European banking sector the relationship between the banks risk taking behavior and expected bailout probability is analyzed using the two step model of Dam and Koetter (2012). In Table 7 the results of the two step model are shown. The most interesting parameter is the coefficient of the bailout probability. If there is a positive coefficient, it means that a higher bailout probability increases the likelihood of running in distress, which is the evidence of moral hazard. For both specification [1] and [2] there is no significant relationship between the bailout probability and likelihood of distress.
Specification
Dependent variable Lerner index
Explanatory covariates
Distressed -19.576***
(-4.24)
Equity to Total Assets (%) 0.011
(0.75)
Net Loans to Total Assets (%) 0.190***
(27.47)
Total Assets (th EUR) (log) 5.863***
17 Table 7 Moral Hazard: Two stage probit estimation of the bailout and distress probabilities
Note: Using the two step model of Dam and Koetter (2012) the bailout probability and the likelihood of running into distress is estimated for each bank. The bailout probability is estimated using a sample of banks which were in distress during the time period 2007 until 2013 Distressed can be defined as a bank which needs a government bailout otherwise it will exit the market. From the estimation of the bailout probability, the data can be extrapolated and each bank can be calculated a bailout probability. If the dummy Political Orientation Government is equal to one the government is right hand orientated. If the Election Year dummy is equal to one, that elections take place in that year. The likelihood of running into distress is calculated using the bailout probability, the banks’ characteristics and macro-economic variables. */**/*** denote the significance at the 10%/5%/1% levels respectively.
What can be seen from the table is that size is an important variable of estimating the bailout probability and the likelihood of running into distress. A positive significant effect of the size of the banks, measured as the log of the total assets, on the likelihood of running into distress means that the argument ‘too big to fail’ is supported. The larger the bank, the more certain it is that they will have a bailout. The coefficients of the size in the distress equation with a size of b=0.577 and b=0.837, specification [1] and [2] respectively, confirm that the larger the size of the bank, the more it is likely to run into distress.
Specification [1] [2]
Dependent variable Bailout Distress Bailout Distress
Explanatory covariates Bailout Probability 0.277 (0.73) -0.374 (-1.25)
Total Capital Ratio (%) 0.046 (0.59)
-0.048*** (-5.22)
Equity to Total Assets (%) 0.106 (0.84) -0.006***
(-2.97)
Net Loans to Total Assets (%) 0.021 (0.99) 0.006* (1.69) 0.016 (0.78) 0.011*** (3.61)
Total Assets (th EUR) (log) 1.384*** (2.82) 0.577*** (4.44) 1.210*** (2.79) 0.837*** (8.32) Return on Assets (%) 0.145* (1.77) -0.124*** (-5.80) 0.136** (1.87) -0.027 (-4.24) GDP Growth Rate (%) -0.000 (-0.00) 0.023 (1.06) -0.002 (-0.01) 0.004 (0.19) Unemployment Rate (%) 0.257** (2.51) -0.002 (-0.12) 0.217** (2.50) 0.286** (1.95) Political Orientation Government (dummy) 0.978
(1.54)
0.931 (1.52) Election Year (dummy) 0.070
18 Since the size of the bank is an important element, some additional analysis on this variable is done. The sample is divided in the three main categories small, medium, and large in the same way as done in the analysis of distortions in the competitive environment. The results are shown in Table 8. There is not enough data available to see how the bailout probability affects the risk taking behavior for the small size banks. The medium and large size banks show however contradicting significant results. When the bailout probability increases, the risk taking behavior of the large bank also increases. This finding is supported by the literature and is interpreted as moral hazard. Medium banks however, lower their risk taking position when their bailout probability increases. This can be contributed to the charter value theory. Medium size banks face more future gains in increasing the charter value of their bank than losses due to increasing risk taking behavior.
Table 8 Moral Hazard: Two stage probit estimation of the bailout and distress probabilities for different bank sizes
Note: Since size is an important factor for determining the likelihood of running into distress, the relationship between the bailout probability and the likelihood of running into distress is estimated for different bank sizes. */**/*** denote the significance at the 10%/5%/1% levels respectively.
Specification Medium Large
Equation Distress Distress
Explanatory covariates
Bailout Probability -1.076*
(-1.81)
0.824** (1.60)
Equity to Total Assets (%) -0.064*** (-3.02)
-0.049** (-2.11)
Net Loans to Total Assets (%) 0.014*** (2.77)
0.007** (1.91)
19
6. Conclusion and discussion
This paper analyses the effect of government intervention in the form of bank bailouts on the competitive environment in the European banking sector. Following Koetter and Noth (2012) the relationship between the Lerner index, as a measure of competition, and the bailout probability is determined using ordinary least squares. To determine the bailout probability the two step model of Dam and Koetter (2011) is used. Finally it is investigated if the distortion of competition is caused by increasing risk taking behavior of banks. First the bailout probabilities are estimated on the basis of historical bank level variables and macroeconomic variables. Using this bailout probability the banks distress likelihood is estimated. If in these two equations a higher bailout probability leads to a higher distress likelihood, moral hazard is identified.
The results show that a higher bailout probability leads to a lower Lerner index. This means that when a bank has more chance to be bailed out if it is in distress, market power is declined. The reason that market power declines is due the fact that non-guaranteed banks face more competition (Hakeness and Schnabel 2004). Guaranteed banks have lower funding costs due to a high expected bailout probability, which leads to unfair competition. If the relationship of the bailout probability and the Lerner index is analyzed for different sizes of banks, it shows there is a different effect for each category. Small and medium banks enjoy higher mark ups if they have an increasing bailout probability and the market power of larger banks becomes significantly lower. This could be explained by the fact that the larger banks become more heavily regulated when their bailout probability increases, since their failure weakens the whole financial system. Furthermore it is analyzed that when a bailout takes place more competition takes place. Banks which have had a bailout have decreasing market power in contrast to non-guaranteed firms.
Finally, it is tested if this distortion of competition is caused by moral hazard. The analysis show that the bailout probabilities does not significantly influence the likelihood of running into distress. Size, however, is an important variable in determining the likelihood of running into distress. To extend this analysis, the sample is divided in different sizes. It shows that especially larger banks takes more risk if they have a higher bailout probability. This can be contributed to the ‘too big to fail’ argument. Even if the government does not announce which banks she will bail out, larger banks know that they will have a bailout when they become in distress, because of the externalities which come into existence in case of failure. It can be concluded that only the larger banks are confronted with moral hazard.
20 be limited. More reliable results can be drawn if more events are included by for example using a longer time period.
21
Appendix
Table 1 List of European banks which had a bailout for the years 2007 until 2013
Bank name and country Date of (first) bailout
Aareal Bank AG DE Feb 09
ABN Amro NL Oct 08
Allied Irish Bank IE Sep 08
Anglo-Irish Bank IE Jan 09
Banca Monte dei Paschi di Siena IT Dec 09
Bank of Ireland IE Sep 08
Bank of Scotland GB Oct 08
Bankia ES May 12
Banco de Valencia ES Dec 12
Banca Monte dei Paschi di Siena ES Jan 13
Banque Populaire FR May 09
Bayerische Landesbank DE Jan 09
BNP Paribas FR Mar 09
Bradford & Bingley GB Sep 08 Caja de Ahorros Castilla La Mancha ES Mar 09
CajaSur ES May 10
Catalunya Banc ES Dec 12
Commerzbank DE Jan 09
Crédit Agricole FR Oct 08
Crédit Mutuel FR Oct 08
Danske Bank DK May 09
Dexia BE Oct 08
EBS IE Apr 10
Fortis BE Oct 08
Groupe Caisse d’Epargne FR May 09
HBOS GB Oct 08
HSH Nordbank DE Jan 09
Hypo Real Estate DE Mar 09
IKB DE Jan 09
NCG Banco ES Dec 12
ING Groep NL Oct 08
IKB Deutsche Industriebank AG DL Jul 07
KBC Group BE May 09
Krajbanka LV Nov 11
Landesbank BW DE Nov 08
Lloyds Banking Group GB Mar 12
Lloyds TSB GB Oct 08
National Westminster GB Oct 08
Natixis FR Aug 09
Nordbanken SE Oct 08
Nordea Banken SE Oct 08
Northern Rock GB Feb 08
RBS Group GB Oct 08
22
Snoras LT Nov 11
SNS Reaal NL Feb 2013
Société Générale FR Oct 08
WestLB DE Nov 09
Note: A bailout is a situation in which the government interferes with a bailout and as a result the bank is able to survive. The list of European banks is gathered from Rotmann (2012) and is updated using news websites. The years 2007 until 2013 are used since before these years, no bank bailouts have taken place in Europe.
Table 2 List of European banks which are failed for the years 2007 until 2013
Note: A failure is a situation in which a bank exits from the market or mergers with another bank. The list of European banks is gathered from Rotmann (2012) and is updated using news websites. The years 2007 until 2013 are used since before these years, no bank failures have taken place in Europe.
Bank name and country Date of failure
Alliance & Leicester GB Oct 08
Anglo-Irish Bank IE Jul 11
Amagerbanken DK Feb 11
Bank of Scotland GB Jan 09
Bank Trelleborg DK Jan 08
Bradford & Bingley GB Sep 08
Capinordic Bank DK Feb 10
Dresdner Bank DE May 09
DSB NL Oct 09
EiK Bank DK Sep 10
Fionia Holding DK Feb 09
Fjordbank Mors DK Jun 11
Friesland Bank NL Apr 12
Forstaedernes Bank DK Oct 08
Fortis BE Oct 08
HBOS GB Jan 09
IKB DE Oct 08
Krajbanka LV Nov 11
Northern Rock GB Nov 11
Roskilde Bank DK Aug 08
SNS Reaal NL Feb 13
23
Table 3 Estimation of the two step model and the relationship of the Lerner index and the bailout probability for both the probit model and the ordinary least square model
Specification [1] Probit [2] Probit [3] OLS [4] OLS
Dependent variable Bailout Distress Lerner index Bailout Distress Lerner index Bailout Distress Lerner index Bailout Distress Lerner index Explanatory covariates Bailout Probability 0.277 (0.73) -8.301*** (-4.64) -0.374 (-1.25) -2.163** (-2.15) 0.277 (0.73) 4.130 (1.45) 0.000 (0.09) 17.440*** (13.69)
Total Capital Ratio (%) 0.046
(0.59) -0.048*** (-5.22) 0.120*** (6.16) 0.001 (0.08) -0.048*** (-5.22) 0.081*** (4.48) -0.000 (-0.73)
Equity to Total Assets (%) 0.106
(0.84) -0.006*** (-2.97) 0.117*** (4.97) 0.008 (0.27) -0.081*** (-4.06) Net Loans to Total Assets (%) 0.021
(0.99) 0.006* (1.69) 0.226*** (16.88) 0.016 (0.78) 0.011*** (3.61) 0.180*** (23.13) 0.001 (0.32) 0.006* (1.69) 0.205*** (15.47) 0.001 (0.14) -0.000 (-0.37) 0.165*** (22.39) Total Assets (th EUR) (log) 1.384***
(2.82) 0.577*** (4.44) 5.940*** (12.92) 1.210*** (2.79) 0.837*** (8.32) 6.130*** (19.62) 0.255*** (3.07) 0.577*** (4.44) 3.145*** (4.30) 0.224** (2.70) 0.005*** (4.02) 1.591*** (4.50) Return on Assets (%) 0.145* (1.77) -0.124*** (-5.80) 5.016*** (27.08) 0.136** (1.87) -0.027 (-4.24) 1.437*** (26.96) 0.034* (1.92) -0.124*** (-5.80) 4.828*** (25.27) 0.035* (1.96) -0.001*** (-4.71) 1.262*** (23.15) GDP Growth Rate (%) -0.000 (-0.00) 0.023 (1.06) 0.318*** (4.16) -0.002 (-0.01) 0.004 (0.19) 0.143** (2.40) -0.003 (-0.12) 0.023 (1.06) 0.276*** (3.64) -0.003 (-0.11) 0.000 (1.28) 0.149** (2.51) Unemployment Rate (%) 0.257** (2.51) -0.002 (-0.12) 0.278** (2.10) 0.217** (2.50) 0.286** (1.95) -0.541*** (-7.61) 0.036*** (2.83) -0.002 (-0.12) -0.145 (-0.99) 0.034*** (2.78) 0.000* (1.80) -0.954*** (-13.23) Political Orientation Government (dummy) 0.978
(1.54) 0.931 (1.52) 0.148 (1.16) 0.153 (1.25)
Election Year (dummy) 0.070
(0.11) 0.040 (0.08) -0.074 (-0.58) -0.070 (-0.60) Consolidated -14.601** (-2.47) -6.746*** (-6.65) -22.016*** (-6.35) -12.316** (-2.42) -9.492*** (-11.90) -12.161*** (-5.47) -1.759* (-1.93) -6.746*** (-6.65) -2.441 (-0.51) -1.438 (-1.60) -0.031*** (-3.59) 14.684*** (6.63) (Pseudo) R² 0.397 0.323 0.095 0.341 0.312 0.061 0.237 0.323 0.093 0.194 0.014 0.067 Number of observations 49 18360 11478 56 45039 32480 49 18360 11478 56 18360 32480
Note: In this table the results are shown for different estimations of relationships between variables for two specifications, in which the variables Equity to Assets and Total Capital ratio differs. In the columns with the dependent variable Bailout, the relationship between bank characteristics, macro-economic variables and political variables and the bailout probability are estimated. The dummy variable Political Orientation Government equal to one represents a right handed orientated government. The dummy variable Election year is equal to one if elections take place that year. In the columns Distress, the relationship between the bailout probability and the likelihood of running into distress is estimated, to investigate if moral hazard is present in the sample. In the columns with the dependent variable Lerner index, the relationship is
24
Table 4 Estimation of the bailout probability with and without the variable Return on Assets
Note: The bailout probability is estimated twice using the method of Dam and Koetter (2012) with and without the variable Return on Assets, specification [1] and [2] respectively. */**/*** denote the significance at the 10%/5%/1% levels respectively.
Specification [1] [2]
Dependent variable Bailout Bailout
Explanatory covariates
Equity to Total Assets (%) 0.106
(0.84)
0.121 (1.01)
Net Loans to Total Assets (%) 0.016
(0.78)
0.018 (0.96)
Total Assets (th EUR) (log) 1.210***
(2.79) 1.134*** (2.81) Return on Assets (%) 0.136** (1.87) - GDP Growth Rate (%) -0.002 (-0.01) 0.065 (0.68) Unemployment Rate (%) 0.217** (2.50) 0.167** (2.18)
Political Orientation Government (dummy) 0.931
(1.52)
0.833 (1.46)
Election Year (dummy) 0.040
25
References
Articles:
Beck, T., Coyle, D., Dewatripont, M., Freixas, X., Seabright, P., 2010. Bailing out the Banks: Reconciling Stability and Competition. CEPR Report,Centre for Economic Policy Research, London, UK.
Bikker, J.A. and Haaf, K. 2000. Competition, Concentration and their Relationship: an Empirical Analysis of the Banking Industry. Research series supervision ,no. 3, De Nederlandsche Bank
Bushman, R.M. and Williams, C.D. 2007. Bank Transparancy, Loan Loss Provisioning Behavior, and Risk Shifting. Working Paper, University of North Carolina- Chapel Hill
Calderon, C. and Schaeck, K. 2012. Bank Bailouts, Competitive Distortions, and Consumer Welfare. Working Paper. The World Bank, Development Economics
Dam, L., & Koetter, M. 2012. Bank bailouts and moral hazard: Evidence from Germany. Review of Financial Studies, Vol. 25, No. 8, pp. 2342-2380.
Dowd, K. 2009. Moral Hazard and the Financial Crisis. Cato Journal, Vol. 29, No. 1, pp. 141-166
Elzinga K.G. and Mills, D.E. 2011. The Lerner Index of Monopoly Power: Origins and Uses. The American Economic Review, Vol. 101, No. 3, pp. 558-565
Flannery, M. 1998. Using Market Information in Prudential Banking Supervision: A Review of U.S. Evidence. Journal of Money, Credit and Banking, Vol. 30, pp. 273-305.
Gropp, R., Hakeness, H., Schnabel, I. 2010. Competition, Risk-Shifting, and Public Bail-out Policies. Review of Financial Studies, Vol 24, No. 6, pp. 2084-2120
Hakenes, H. and Schnabel, I. 2004. Banks without Parachutes – Competititve Effects of Government Bail-out Policies. Working Paper, Max Planck Institute for Research on Collective Goods, Bonn
26 Hughes, J.P. and Mester, L.J. 1993. A Quality and Risk-Adjusted Cost Function for Banks: Evidence on the ‘’Too-Big-To-Fail’’ Doctrine. The Journal of Productivity Analysis, Vol. 4, pp. 293-315
Keeley, M.C. 1990. Deposit Insurance, Risk, and Market Power in Banking. The American Economic Review, Vol.80 No. 5, pp. 1183-1200
Koetter, M. and Noth, F. 2012. Competititve Distortions of Bank Bailouts. Working paper
Morrison, A.D. 2011. Systemic Risks and the ‘Too-Big-To-Fail’ Problem. Oxford Review of Economic Policy, Vol. 27, No. 3, pp. 498-51
Norden L., Roosenboom, P., Wang, T., 2012. The Impact of Government Intervention in Banks on Corporate Borrowers’ Stock Returns. Working paper. Erasumus University of Rotterdam
Rottman, J. 2012. Bank Bailouts and Moral Hazard in the Financial Crisis: recent evidence from the European Banking Sector. Working Paper, University of Groningen
Shabbir, G., Anwar, S., Hussain, Z., Imran, M. 2012. Contribution of Financial Sector Development in Reducing Unemployment in Pakistan. International Journal of Economics and Finance, Vol 4, No. 1, pp. 260-268
Stern, G.H. and Feldman, R.J. 2004. Too Big To Fail: The hazards of Bank Bailouts. Brookings Institution Press, Washington
Websites:
http://www.europa-nu.nl/id/vj05k11iqguz/europese_bankenunie viewed 27 may 2014
Books:
