The relation between charter value and bank risk: Is it U-shaped? Evidence from the Eurozone

(1)

The relation between charter value and bank risk:

Is it U-shaped? Evidence from the Eurozone

Master’s thesis, MSc Finance

University of Groningen, Faculty of Economics and Business

January 2016

Tim Strokappe

Student number: 2161362

First supervisor: M. Zaouras

Abstract

(2)

2

1. Introduction

Banks are an important part of the financial system and are critical to the health and efficiency of the economy. The main role of the bank is to channel funds from savings to parties which have productive investment opportunities. If banks go into default, they will not be able to deliver the latter and hence cause problems to the economy. However, banks are also in search of return. Since risk and return are intertwined, banks can yield higher returns if they exert higher risk-taking strategies. As we have experienced from the last financial crisis, the high risk-risk-taking behavior of the banks can do severe damage to the financial system, which directly affects the real economy. Hence, the risk-taking behavior of the banks can induce major problems for society and the welfare of a country. Therefore, it is important that banks act prudently by not taking excessive risks in their activities. Policy makers and regulators are interested in how they can limit the risk-taking behavior of the banks and thereby enhance overall financial stability.

An academic view that has been suggested to prohibit banks from taking too much risk is the bank charter value paradigm. The charter value of a bank captures the bank’s future economic profits and is zero in case of bankruptcy. Keeley (1990) finds that banks with large charter values, which are partly a result of a low level of competition, are more incentivized to take on a lower risk-taking strategy. This is because large charter value banks want to preserve the charter value they have build up over the years and thereby avoid high potential bankruptcy costs. Furthermore, Keeley (1990) states that banks with a low bank charter value will have less to lose if it goes bankrupt which induces more risk-taking. According to this hypothesis regulators could inhibit excessive risk-taking behavior of banks by encouraging banks to preserve high charter values.

However, the Boyd and De Nicolo (2006) model argues that decreasing competition leads to higher default risk on loans, caused by the higher interest rates banks will demand on their

(3)

3 I will use a dataset consisting of bank-specific data of listed banks from the Eurozone from 2006 to 2014. The model will be estimated using the two-step system Generalized Method of Moments (GMM) estimator, following the specification used by Niu (2012). The results of this study could have important implications for policy makers to enhance financial stability in the banking sector and thereby improve the health of the whole economy. In contrast to Niu (2012), the results of this study do not support a U-shaped relationship between charter value and the five measures of bank risk for the European banking sector. Instead, I find evidence for the negative linear relationship as

suggested by Keeley (1990) and the charter value paradigm.

The rest of this paper is organized as follows. In section 2 I discuss the existing literature on this topic in the literature review. In section 3 I present the dataset and define the variables. Section 4 contains the methodology of the empirical analysis. In section 5 I present the results. Finally, section 6 concludes.

2. Literature review

2.1 Theoretical Literature

The charter value or franchise value paradigm is a well established topic in the existing banking literature. The main theory states that banks are willing to reduce bank risk-taking in order to preserve the bank’s charter value. In case of an increase in competition within the banking market, profits will decline and hence a decline in bank charter value. According to the established literature, a decline in charter value will in most cases lead to higher risk-taking and therefore greater financial instability within the banking market.

In one of the early papers on this subject, Marcus (1984) notes that the traditional view of bank finance on wealth maximizing is too simplistic. The standard view argues that banks can maximize wealth by exploiting non-risk-rated deposit insurance by either holding an asset portfolio with a high variance in returns or by decreasing the bank’s capital, without taking the effect of potential losses in bank charter value into account. Hence, potential bankruptcy costs due to insolvency were left out of the equation. Marcus (1984) states that the charter value effect can induce drastic changes in the optimal financial policy of the banks. Banks with decreasing bank charter values will in most cases obtain a strategy with a higher risk profile as these strategies are relatively more attractive. Moreover, Marcus (1984) concludes that banks of which the goal is to maximize value, the optimal strategies are either extremely high risk strategies or extremely low risk strategies, stating a policy in the middle ground as suboptimal.

Furthermore, Chan et al. (1986) finds that banks devote less resources to screening

(4)

4 potential high-quality borrowers. The lack of quality screening leads to the selection of more poor quality assets and therefore the credit risk within the loan portfolio increases. Therefore, a decline in charter value, caused by high competition, leads to higher risk-taking within the loan portfolio. Marquez (2002) and Besanko and Thakor (1993) have similar findings as Chan et al. (1986).

Keeley (1990) states that banks will not have the incentive to take on more default risk as long as the expected loss of charter value is higher than the gain of exploiting the effects of the deposit insurance. In other words, the potential loss from bankruptcy costs would offset the gains from higher risk-taking behavior. Using a state preference model, Keeley (1990) finds that

anticompetitive regulations make bank charters valuable to the banks. Furthermore, Keeley (1990) argues that banks with more market power hold more capital and have a lower risk profile. Hence, the potential loss of bank charter will incentivize banks to limit their risk-taking behavior.

The previous studies consider the charter values as being exogenous. Suarez (1994) uses a dynamic optimization model with endogenous values to find that there is a negative relation between market power and solvency. Furthermore, Suarez (1994) states that a bank’s charter value is an important component of bankruptcy costs. Bankers may therefore be incentivized to adopt less risk-taking strategies in order to stay solvent and thereby preventing themselves from paying high bankruptcy costs.

Hellmann et al. (2000) argue that financial-market liberalization cause banks to display less prudent behavior and therefore cause more moral hazard problems. The idea is that financial-market liberalization attracts more competition which implies lower profits, which will be reflected in a lower charter value. The lower charter value induces lower incentives to create a loan portfolio of high quality and thus induces more gambling incentives. Hence, moral-hazard problems will be increasing. Furthermore, Hellmann et al. (2000) finds that capital requirements are not sufficient to overcome the moral hazard problem, because it will affect future charter values negatively. Hence, it will lead to higher risk-taking in the future.

In addition to Hellmann et al. (2000), Repullo (2004) also finds a positive relationship between competition and bank risk-taking. Repullo (2004) finds that using a dynamic model of imperfect competition in the deposits market, an increase in competition leads to lower bank margins, which hereby will negatively affect the charter value. This will in turn lead to higher risk-taking behavior in the absence of adequate regulation. In contrast to Hellman et al. (2000), Repullo (2004) finds that based capital requirements could control effectively for the shifting in risk-taking incentives.

(5)

5 a negative relationship between competition and bank failure, which is the reverse of the prediction of the charter value paradigm. Boyd and de Nicolo (2006) argue that if the market power of the bank is high, higher interest rates on the loans will be demanded. This will induce borrowers to take more risk if they adopt a loan. The paper by Stiglitz and Weiss (1981) states that if borrowers have to take on more risky projects, the portfolio of the bank becomes more risky as well. The riskiness of the portfolio can harm profits due to losses from non-performing loans and this will harm buffers against losses. Therefore, bankruptcy risk will increase as competition goes down.

Martinez-Miera and Repullo (2010) extended the model from Boyd and De Nicolo by assuming imperfect correlation between loan defaults and the probability of bank failure. Their results show that there are two opposite effects on bank failure. The first effect is the risk-shifting effect which implies that higher competition leads to lower loan rates, which will reduce the default risk of borrowers on outstanding loans. Hence, higher competition will lead to lower bank risk. The second effect is called the margin effect. This effect captures the result that when competition increases and the loan rates drop, revenues from performing loans will be lower. Hence, buffers to cover potential losses will decrease and this will lead to higher bank risk and higher probability of bank failure. Taking these two effects into account, the Martinez-Miera and Repullo (2010) model predicts a U-shaped relationship between competition and the risk of bank failure. Niu (2012) argues that given the negative relation between competition and charter value, the Martinez-Miera and Repullo (2010) model implies a U-shaped relationship between charter value and bank risk.

2.2 Empirical Literature

(6)

6 Japanese financial system. Ghosh (2009) has studied the factors which determine bank charter value and the disciplining effect it has on bank risk-taking. Ghosh (2009) finds that efficiency and bank size are important determinants of charter value and he finds a robust disciplining effect of charter value on several bank risk measures. Gan (2004) uses the Texas real estate crisis to test the relationship between bank charter value and bank risk-taking during an exogenous shock and finds that in case of an exogenous shock the slope between risk and charter value becomes more negative. Salas and Saurian (2003) examine the relationship on the Spanish banking market and find that banks with lower charter values have in most cases lower solvency rates and experience higher credit risk. The above mentioned studies all find empirical results supporting the charter value paradigm, which implies a negative relation between bank charter value and bank risk-taking.

However, Saunders and Wilson (2001) state that the self-regulatory benefits from charter value may be constrained. They find that high charter value banks show the largest declines in charter value during market declines. Therefore, Saunders and Wilson (2001) argue that the risk constraining incentive can disappear quickly during economic contractions. Jones et al. (2011) find similar results. Park (1997) states that a large charter value can fail to improve the soundness of the bank, unless it is complemented with proper regulations. In a working paper, Arping (2014) finds that banks which have lower charter values due to tighter margins, also have less room to take reckless risks and therefore indicating a positive relationship. Finally, Berger et al. (2009) find support for both the competition-fragility view as the competition-stability view. They find that a lower charter value due to higher competition, leads to higher risk-taking to increase returns, as displayed by the competition-fragility view. However, they also find that higher competition, and thus lower margins and consequently lower charter value, leads to lower risk in the loan portfolio as suggested by the competition-stability view.

2.3 Hypothesis

(7)

7 This study will examine whether there is a U-shaped relationship between charter value and bank risk for the European banking sector, like Niu (2012) finds in his study covering the American banking sector for a different time period. In addition to Niu (2012), this study will test for the linear relationship as proposed by the charter value paradigm.

3. Data

The data which constitute the dataset are obtained from Bankscope and Datastream. Bankscope offers the bank level data that is required for the empirical analysis and Datastream the stock returns for both the banks and the index. The dataset consists of all listed banks within the Bankscope database which are established in the Eurozone and meet the data criteria. There are two data criteria which determine the content of the dataset. First, the data to construct all independent variables of the specification used by Niu (2012) have to be available to examine at least one year. Second, banks have to be operating in a country which is connected to the Eurozone before 2006. Furthermore, I have corrected the dataset for outliers by excluding one percent of the extreme values in the distribution for each risk measure. This all leads to a panel of yearly data from 2006 to 2014. In table 1 I report a summary of all variables with their definitions which are used in the empirical analysis.

Table 1

Definitions of variables

Variable Definition Total Risk The standard deviation of daily stock returns for a bank

in a year.

Systematic Risk The coefficient on the market index in annual regressions of a bank's daily stock returns on the MSCI Europe index. Firm Specific Risk The standard deviation of the residuals in annual regressions

of a bank's daily stock returns on the MSCI Europe index. Nonperforming Loans Nonperforming loans/total loans.

Z-score Z = (ROA + CAR)/σ(ROA), where ROA is the return on assets and CAR is the capital ratio.

Charter Value Market value of assets/Book value of assets. Size LN(total assets).

Capital Ratio Equity/total assets Return on Assets Net income/total assets

Noninterest Income Ratio Noninterest income/(net interest income + noninterest income) .

Cost Income Ratio Noninterest expense/(net interest income + Noninterest income) .

Subdebt Subordinated debt/total assets . GDP growth The growth in GDP per year .

(8)

8 3.1 Bank Risk

I use five different measures of bank risk, following the specification used by Niu (2012). The first measure is the total risk of a bank which is the standard deviation of daily stock returns for a certain bank in a certain year. This measure corresponds to the overall bank risk, see Haq and Heany (2012) and Pathan (2009). The second measure is systematic risk which is defined as the coefficient on the market index. This measure reflects the exposure of an individual bank to market

movements. The MSCI Europe index is used as the index to determine systematic risk, which is also used by Haq and Heaney (2012). The third measure is the specific risk measure and is defined as the standard deviation of the regression residuals and reflects the unique risks a bank is exposed to.

The fourth measure is the nonperforming loans ratio. The nonperforming loans ratio is defined as the ratio between the amount of nonperforming loans against the amount of total loans. This variable is important to measure, because it reflects the exposure of credit risk on the bank’s loan portfolio. Jiminez et al. (2013) state that credit risk is the primary source of the overall risk of the bank. Furthermore, the results and models of Martinez-Miera and Repullo (2010) and Boyd and de Nicolo (2006) are based on the borrowing behavior of banks in the loan market. Besides Jiminez et al. (2013), Delis et al. (2011) and Berger et al. (2009) are two other studies that use this ratio in their studies.

Finally, I use the Z-score as a measure of bank risk. According to Lepetit and Strobel (2013), the Z-score is a simple but good measure for financial stability and insolvency risk of individual banks. The Z-score is defined as the sum of return on assets (𝑅𝑂𝐴) and the capital asset ratio (𝐶𝐴𝑅) divided by the standard deviation of the return on assets (𝜎(𝑅𝑂𝐴)). Bank insolvency implies that the losses are higher than the equity, hence the probability of insolvency is given by 𝑝(−𝑅𝑂𝐴 < 𝐶𝐴𝑅). This implies that the score is an inverse measure of insolvency probability, meaning that a higher Z-score implies lower insolvency risk.

3.2 Charter Value

(9)

9 3.3 Control Variables

Besides the charter value, there might be some other variables that have an impact on bank risk as well. I will include a number of control variables, which also have been used in other studies. The first control variable is the size variable and is defined as the natural logarithm of total assets. Saunders et al. (1990) find that large banks have more exposure to the general market and are therefore more exposed to systematic risks. Galloway et al. (1997) state that large banks might have an incentive to take on higher risks, because the probability that these banks will be bailed out by the government in case of financial distress is high. Furthermore, Hakenes and Schnabel (2011) find that under Basel II small and large banks are treated asymmetrically, which due to fiercer competition leads to a higher risk-taking incentive for small banks.

The second control variable is the capital asset ratio, which is defined as the ratio between equity and total assets. Capital affects bank risk as it can be seen as a cushion against unexpected losses. Moreover, capital reduces gambling incentives as it puts bank equity at stake (Hellmann et al. (2000)). This would imply a negative relation between capital and bank risk.

Third, I control for the return on assets measure. This variable captures the profitability of the bank. The profitability of the bank is an influence on bank risk as a part of the profits will be booked as retained earnings. Hence, high profits will increase the bank’s capital buffer and therefore reduces the risk of insolvency. Therefore, I expect a negative relationship between return on assets and bank risk.

The fourth control variable is the noninterest income ratio. This ratio is calculated as non-interest income divided by the sum of net non-interest and nonnon-interest income. Stiroh (2004) argues that reliance on noninterest income will lead to lower risk-adjusted profits and higher risks. Furthermore, Brunnermeier et al. (2012) find in their paper that banks with high non-interest income have a higher contribution to systematic risk. These results imply a positive relationship.

The fifth variable is the cost income ratio and it is a variable that measures the bank’s cost efficiency. It is defined as the noninterest expense divided by the total of net interest income and noninterest income. Berger and de Young (1997) examine whether the cost income ratio is a measure to predict future problem loans. They find that better cost efficiency measures precede reductions in the amount of nonperforming loans. Therefore, cost efficiency might be an adequate measure to predict future problem loans and problem banks.

(10)

10 relationship. He finds that subordinated debt leads to higher risk-taking incentives.

To control for the differences between countries in the Eurozone, I also add the GDP-growth and Population measure. Table 1 provides an overview of all the variables I mentioned above and the corresponding definitions.

Table 2

Descriptive statistics

Variable Obs. Mean Std. Dev. Min Max

Risk-taking Variables

Total Risk 447 0.0254 0.0145 0.0010 0.0873

Systematic Risk 450 0.8553 0.5465 -0.1457 2.2992

Firm Specific Risk 435 0.0201 0.0106 0.0010 0.0591

Nonperforming Loans 409 0.0702 0.0659 0.0005 0.3494 Z-score 453 13.2604 6.8871 0.1967 29.9139 Other Variables Charter Value 457 0.9921 0.0475 0.8818 1.4363 Size 462 11.2293 1.8110 5.7205 14.6051 Capital Ratio 459 0.1272 0.0314 -0.0510 0.2470 Return on Assets 462 0.0013 0.0128 -0.1237 0.0443

Noninterest Income Ratio 462 0.3608 0.2503 -2.3666 2.3271

Cost Income Ratio 462 0.6491 0.2981 -1.5146 3.4619

Subdebt 454 0.0188 0.0110 0 0.0564

GDP-Growth 462 0.0008 0.0287 -0.0910 0.0770

Population (x10mln) 462 4.2586 2.7341 0.4208 8.2438

This table reports summary statistics for the variables that are used in the analysis. Please see table 1 for the definitions of the variables.

(11)

11 this study substantially smaller, where Niu (2012) found an average charter value of 1.069. The other control variables for the average bank in the sample are 11.2293 in size, 0.1272 for the capital ratio, 0.0013 for the return on assets measure, 0.3608 for the noninterest income ratio, 0.6491 for the cost income ratio and 0.0188 for the subdebt measure. The country level control variables have averages of 0.08% for GDP-growth and 42.5 million citizens.

As mentioned before, there are differences in the availability of the information for each bank in Bankscope. This leads to differences in the number of banks per year in the dataset and the available years each bank has in the dataset. This results in an unbalanced dataset which is used to perform the empirical analysis. For further information of the differences, tables A1.1 and A1.2 in the appendix can be consulted.

In table 3 I present the correlation matrix. In respect to the correlations between the charter value and the other control variables we can observe the following: Firstly, the correlation between size and charter value is positive, 0.042. This is consistent with the charter value-size correlation Niu (2012) finds. Secondly, the relationship between charter value and the capital ratio is negatively correlated, -0.006. This correlation differs from the study by Niu (2012) who finds a significant positive correlation. The third relation is between charter value and return on assets. This relationship is positively correlated, 0.046, and is consistent with positive correlation Niu (2012) reports. This could be explained by that banks with better return measures have higher charter values. Fourth, the correlation between charter value and the noninterest income ratio is

insignificant and slightly positive, 0.001. The correlation measure of Niu (2012) differs substantially. He finds a significant positive relationship between charter value and the noninterest income ratio, 0.2172. Fifth, consistent with Niu (2012), I find a negative correlation between charter value and cost income ratio, -0.017. This indicates that banks which are less efficient with regard to the costs have lower charter values. And finally, the sixth relationship is between charter value and subdebt. I find a negative correlation of -0.031, indicating that more subordinated debt does not increase charter value. This is in contrast with Niu (2012) who finds a positive relationship. Furthermore, I find no significant correlations between charter value and the control variables in my dataset. Niu (2012), however, finds only highly significant correlations on the one percent significance level between charter value and the control variables. It appears that the datasets differ substantially from each other. This might be explained by the financial crisis period that this study considers, while the study performed by Niu (2012) does not.

(12)

12 Table 3

Correlation matrix

Variable TR SR FSR NPL Z CV S CAR ROA NII CI SD POP GDP Total Risk (TR) 1.000

Systematic Risk (SR) 0.463*** 1.000

Firm Specific Risk (FSR) 0.955*** 0.361*** 1.000

Nonperforming Loans (NPL) 0.430*** 0.175*** 0.456*** 1.000

Z-score (Z) -0.170*** -0.060 -0.192*** -0.363*** 1.000

Charter Value (CV) -0.000 -0.087 -0.023 0.127** 0.237*** 1.000

Size (S) 0.186*** 0.628*** 0.052 -0.079 0.052 0.042 1.000

Capital Ratio (CAR) -0.109** 0.042 -0.184*** -0.021 0.030 -0.006 0.063 1.000

Return on Assets (ROA) -0.426*** -0.146*** -0.481*** -0.401*** 0.385*** 0.046 -0.030 0.253*** 1.000

Noninterest Income Ratio (NII) 0.126*** -0.006 -0.147*** -0.202*** 0.031 0.001 0.025 0.048 0.142*** 1.000

Cost Income Ratio (CI) -0.093** -0.072 0.155*** 0.249*** -0.125*** -0.017 -0.046 -0.073 -0.215*** -0.534*** 1.000

Subdebt (SD) -0.202*** -0.158*** -0.175*** -0.034 0.178*** -0.031 -0.292*** 0.066 0.079* -0.105** 0.031 1.000

Population (POP) -0.186*** 0.034 -0.150*** -0.089* 0.020 -0.148*** 0.201*** 0.020 0.088* 0.188*** 0.058 -0.091* 1.000 GDP-Growth (GDP) -0.374*** -0.113** -0.376*** -0.270*** 0.032 0.211*** 0.060 0.101** 0.374*** 0.060 -0.029 0.093** 0.045 1.000

This table reports the correlations of the variables used in the empirical analysis. Please see table 1 for the definitions of the variables. ***Significant on 1% level.

(13)

13 are negatively correlated. These effects could be possible as size has different effects on bank risks (e.g. Galloway et al. (1997), Hakenes and Schnabel (2011)).

4. Methodology

The model that examines the relationship between charter value and bank risk is the following: 𝑅𝑖𝑠𝑘𝑖,𝑡= 𝛽0+ 𝛽1𝑅𝑖𝑠𝑘𝑖,𝑡−1+ 𝛽2𝐶ℎ𝑎𝑟𝑡𝑒𝑟 𝑉𝑎𝑙𝑢𝑒𝑖,𝑡−1+ 𝛽3(𝐶ℎ𝑎𝑟𝑡𝑒𝑟 𝑉𝑎𝑙𝑢𝑒𝑖,𝑡−1) 2 + 𝛽4𝑆𝑖𝑧𝑒𝑖,𝑡−1 + 𝛽5𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑅𝑎𝑡𝑖𝑜𝑖,𝑡−1+ 𝛽6𝑅𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝐴𝑠𝑠𝑒𝑡𝑠𝑖,𝑡 + 𝛽7𝑁𝑜𝑛𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝐼𝑛𝑐𝑜𝑚𝑒 𝑅𝑎𝑡𝑖𝑜𝑖,𝑡+ 𝛽8𝐶𝑜𝑠𝑡 𝐼𝑛𝑐𝑜𝑚𝑒 𝑅𝑎𝑡𝑖𝑜𝑖,𝑡+ 𝛽9𝑆𝑢𝑏𝑑𝑒𝑏𝑡𝑖,𝑡−1 + 𝛽10𝐺𝐷𝑃𝐺𝑟𝑜𝑤𝑡ℎ𝑗,𝑡+ 𝛽11𝑃𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑗,𝑡+ 𝜗𝑡+ 𝜇𝑖+ 𝜀𝑖,𝑡

where 𝑖 denotes the individual bank, 𝑗 denotes the country a bank is operating in and 𝑡 denotes the year. 𝜗𝑡 are year fixed effects, 𝜇𝑖 are bank fixed effects and 𝜀𝑖,𝑡 is the error term. 𝑅𝑖𝑠𝑘𝑖,𝑡 represents

one of the five different risk measures and 𝑅𝑖𝑠𝑘𝑖,𝑡−1 denotes the lagged dependent variable.

GDP-growth and Population are included to control for differences between the countries the banks are operating in.

The lagged dependent variable is included because previous papers, such as Jiminez et al. (2013), find that banks adjust their risk-taking over several periods. Using the Wooldridge test to test for autocorrelation in panel data I find that for four out of five risk measures there is significant autocorrelation present. Hence, lagged values contain useful information about current risk-taking behavior.

Secondly, I implement both charter value as well as the charter value squared variables to examine the existence of a U-shaped relationship between charter value and bank risk. If 𝛽2 appears

to be significantly negative and 𝛽3 significantly positive, the results will support the U-shaped

relationship between charter value and bank risk. The only exception is the Z-score, which is the inverse probability of bank default. Hence in case of the Z-score, a U-shaped relationship will be found if 𝛽2 and 𝛽3 will be significant positive and significant negative, respectively.

Thirdly, I include bank fixed effects to control for unobservable bank characteristics constant over time. Furthermore, year fixed effects are implemented to control for structural changes in the banking industry over time.

The model is estimated by using the two–step system GMM estimator developed by Arellano and Bover (1995) and Blundell and Bond (1998). Following Niu (2012), standard errors will be

(14)

14 variables.

The consistency of the two-step system GMM estimator depends on the absence of autocorrelation in the error terms and the validity of the instruments. Therefore, I conduct two diagnostic tests to test for these conditions. Firstly, to test the validity of the instruments I will use the Hansen J test, which is a test that measures for over identifying restrictions. If the null hypothesis is not rejected, the instruments are valid. Secondly, I will perform a Arellano-Bond test for

autocorrelation in the residuals of the first-difference equation. There is no autocorrelation in the error terms if there is order autocorrelation but no second-order autocorrelation of the first-difference equation. Both null hypotheses assume no autocorrelation.

5. Empirical Results

In this section I present the results from the regressions to test the relationship between charter value and bank risk. In section 5.1 I report the results from the test using the specification of Niu (2012), with the dataset where the outliers are excluded1_{. In section 5.2 I report some extra}

robustness checks. In section 5.3 I will present the overall results of this section.

5.1 Results

In table 4 I present the results of the five regressions I have performed using the

methodology stated above, which follows the methodology performed by Niu (2012). As mentioned in the methodology section, there should be a negative significant coefficient on charter value and a positive significant coefficient on charter value squared in order to find the U-shaped relationship between charter value and bank risk. This is only for the risk measures total risk, systematic risk, firm specific risk and the nonperforming loans. For the Z-score I expect a positive significant coefficient on charter value and a negative significant coefficient on charter value squared to find the U-shaped relationship. This is because the Z-score is an inverted measure of insolvency risk. In contrast to Niu (2012), I find for only the nonperforming loans ratio significant coefficients on charter value and charter value squared using my dataset of banks from the Eurozone2_{. This is conflicting with Niu}

(2012) who finds for all five risk measures significant coefficients, implying all U-shaped relationships. Furthermore, the significant coefficients on charter value and charter value squared I found for the nonperforming loans ratio indicate an inverted U-shaped relationship, instead of a regular U-shaped relationship. This is because the coefficient is positive for charter value and negative for charter value

1_{I have performed the same tests with the same specification using the dataset including outliers. However,}

results were insignificant for all measures.

2_{I have performed a Wald test to test for the coefficients jointly being equal to zero. The null hypothesis was}

(15)

15 Table 4

The relation between charter value and bank risk with the squared term included.

Total Systematic Firm specific Nonperforming

risk risk risk loans Z-score

Lagged dependent variable 0.342*** 0.128 0.254** 0.798*** 0.242

(0.102) (0.141) (0.117) (0.128) (0.170)

Charter value 0.120 -2.642 0.034 0.903* 6.814

(0.147) (4.997) (0.163) (0.481) (51.701)

Charter value squared -0.051 0.750 -0.015 -0.439** 1.041

(0.061) (2.095) (0.066) (0.187) (21.807) Size 0.004** 0.188** 0.001 0.023*** 1.125 (0.002) (0.078) (0.002) (0.008) (0.996) Capital ratio -0.144*** 1.382* -0.079*** -0.216** -35.632 (0.040) (0.711) (0.021) (0.096) (21.707) Return on assets -0.018 0.549 -0.109*** -0.659*** 99.007*** (0.049) (0.936) (0.038) (0.255) (20.854)

Noninterest income ratio -0.002 -0.054 -0.006 0.015 0.337

(0.003) (0.126) (0.004) (0.014) (0.667)

Cost income ratio -0.002 -0.089 -0.008 0.022* 0.609

(0.005) (0.108) (0.007) (0.011) (0.884) Subdebt -0.028 4.581 -0.090 0.703** 31.384 (0.139) (6.785) (0.080) (0.290) (19.454) GDP-growth -0.115** -1.051 -0.069* 0.026 -10.571 (0.046) (1.255) (0.038) (0.210) (11.030) Population -0.005 0.057 0.004 -0.017* 1.010 (0.004) (0.092) (0.004) (0.009) (0.914)

Year fixed effects Yes Yes Yes Yes Yes

Number of observations 376 380 360 347 380

Number of groups 58 59 58 56 60

Number of instruments 53 53 53 53 53

Hansen test (p-value) 0.307 0.685 0.588 0.883 0.854

AR(1) test (p-value) 0.000 0.004 0.001 0.141 0.003

AR(2) test (p-value) 0.127 0.1309 0.057 0.886 0.990

The results from the two-step system GMM regressions. Robust standard errors are reported within the parentheses. Please consult table 1 for the definitions of the variables.

*** indicates statistical significance at the 1% level ** indicates statistical significance at the 5% level * indicates statistical significance at the 10% level

(16)

16 value and bank risk for the European banking sector using this specific dataset which includes the financial crisis period.

Niu (2012) states that there are direct implications for regulators from the U-shaped relationship his study found. Because of a sign change of the correlation between charter value and bank risk on the charter value spectrum, the effect of regulation is dependent on the current level of bank charter value. This is in contrast to the previous studies, which argue that an adequate policy to incentivize bank risk reduction is to incentivize banks to maximize charter value. I find only significant coefficients for one risk measure and these coefficients imply an inverted U-shaped relationship instead of a regular U-shaped relationship. Hence, the results from this study using this specification do not support the U-shaped relationship, found by Niu (2012) in the American banking sector, for the European banking market in a different time period. Furthermore, the insignificant coefficients on charter value and charter value squared indicate that charter value is not an adequate measure to explain bank risk this way. However, the linear negative relationship between charter value and bank risk, as proposed by studies such as conducted by Keeley (1990) and Demsetz et al. (1996), might still be present. In section 5.2.2 I will discuss the results on the linear relationship.

In addition to the difference in findings between the charter value-bank risk relationship, there are also substantial differences in the results of the control variables between Niu (2012) and this study. Firstly, Niu (2012) finds for every risk measure highly significant coefficients on the lagged dependent variables, while I only find significant lagged dependent variables for the total risk, firm specific risk and nonperforming loans measure.

Secondly, Niu (2012) finds a negative relationship between size and total risk and firm specific risk, while I find a significant positive relationship for total risk and an insignificant positive relationship for firm specific risk. Galloway et al. (1997) suggests that the positive relationship comes from the assumption that large banks have a higher probability on government bailout in case of financial distress and therefore a higher incentive on excessive risk-taking. Consistent with Niu (2012), I find a positive significant relationship between size and systematic risk. According to Saunders et al. (1990) can this be explained by the higher sensitivity to general market movements by larger banks. Furthermore, I find a highly significant positive relationship between size and the nonperforming loans ratio, this differs substantially from the studies performed by Ghosh (2009) and Jimenez et al. (2013) who both find a highly significant negative relationship.

Third, Niu (2012) finds no significant correlations between the risk measures and the capital ratio. I however find a significant positive correlation for systematic risk. Furthermore, I find

significant negative correlations for total risk, firm specific risk and the nonperforming loans measures. The significant negative correlations on the total risk, firm specific risk and the

(17)

17 Fourth, for the return on assets measure I expect negative relationships with the first four risk measures and a positive relationship on the Z-score. In line with Niu (2012) I find negative relationships on the total risk and firm specific risk measures, where the correlation for the firm specific risk measure is highly significant. Furthermore, I find a highly significant negative correlation for the nonperforming loans measure and a highly significant positive correlation for the Z-score. The significant measures are also consistent with the expectations and the Niu (2012) study.

Fifth, the noninterest income variable did not give any significant results with the risk measures. This is in contrast to Niu (2012) who finds significant correlations on total risk, systematic risk and firm specific risk.

Sixth, for the cost income ratio there are no significant correlations except for the

nonperforming loans ratio. Niu (2012) does not find any significant results between the cost income ratio and the risk measures. Consistent with Ghosh (2009) I find a positive significant relationship for the nonperforming loans measure. This positive relationship is in line with the explanation of Berger and de Young (1997).

Finally, the subordinated debt ratio is only significantly correlated with the nonperforming loans ratio in this study. The only significant correlation Niu (2012) finds is the systematic risk

measure. Ghosh (2009) finds the same positive significant relationship between nonperforming loans and the subordinated debt ratio. Furthermore, the insignificant measures could be explained by the findings from the Bliss and Flannery (2002) study. They argue that investors, among which are subordinated debt holders, do not have the power to influence managerial actions and thereby cannot constrain risk-taking behavior.

Next to the control variables Niu (2012) uses, I also implemented the GDP-growth measure and the Population amount to control for differences between countries. A result that is noteworthy is that I find an insignificant correlation between the GDP-growth and the nonperforming loans ratio. This is noteworthy, because Beck et al. (2013) argue in their working paper that real GDP-growth was the main determinant of the nonperforming loans ratio’s in the past decade.

(18)

18 means that there is no autocorrelation in the error terms, except for the firm specific risk measure. These tests show that in general the GMM estimator is well specified.

5.2 Robustness checks 5.2.1 U-shaped relationship

To test whether the conclusion that there is no U-shaped relationship between charter value and bank risk is valid, I have conducted some robustness checks. The first correction is to exclude the capital ratio and the return on assets. I have done this as charter value is in theory highly correlated with these two control variables and this might be a reason of the coefficient of charter value not being significant. Furthermore, the Z-score is constructed using the capital ratio and return on assets measures. Hence, the results from the Z-score could be expected to be explained largely by these two variables. In table A2.2 in the appendix I present the results using this correction and there is only one significant difference from the initial regressions. That is, the significant coefficients of charter value and charter value squared for the nonperforming loans ratio have become insignificant. The other risk measures remain insignificant. These results are indicating that there is no U-shaped relationship between charter value and bank risk.

Secondly, Niu (2012) is not specific why to use some lagged control variables instead of unlagged variables. Furthermore, there are other papers concerning the charter value and bank risk relationship which do not use any lagged control variables (e.g., Demsetz et al., 1996, Ghosh, 2009). Therefore, I have performed the regressions without lagged control variables to check for

misspecification bias in the Niu (2012) methodology. I have done this specification with and without the capital ratio and return on assets as control variables. I report the full results in tables A2.4 and A2.6 in the appendix. Tables A2.4 and A2.6 report that for both specifications there are no significant coefficients on either charter value or charter value squared. These findings support the main conclusion of the results section that there is no U-shaped relationship between charter value and bank risk3_.

5.2.2 linear relationship

Studies performed by Keeley (1990) and Demsetz et al. (1996) argue that there is a negative linear relationship between charter value and bank risk. I will test for the linear relationship by excluding the charter value squared variable from the specification4_{. In table A2.1 in the appendix I}

3_{The same robustness checks are performed with the dataset uncorrected for outliers, the results are the}

same.

4_{A Wald test proved that removing the squared term does not harm the fit of the model for all regressions,}

(19)

19 report the results using the methodology from Niu (2012) without the squared term. I find a highly significant negative relationship between charter value and the systematic risk measure, where the GMM estimator is well specified. This is in line with the findings from Keeley (1990) and Demsetz et al. (1996). For the other risk measures I do not find any significant results and this implies that charter value is not a good predictor of bank risk.

Furthermore, I have performed the same robustness checks which I have done for the U-shaped relationship. That is firstly, excluding capital ratio and return on assets as control variables from the regression. In table A2.3 in the appendix I present the results from the regressions with the Niu (2012) specification but without the squared term, and without capital ratio and return on assets as control variables. I still find the negative significant relationship between charter value and the systematic risk measure. In addition, I now also find results for a highly significant negative

relationship for the nonperforming loans ratio and a highly significant positive relationship for the Z-score. Taking into account that the Z-score is an inverted measure of insolvency risk, I therefore also find a significant negative relationship between charter value and insolvency risk. This indicates that there is a negative linear relationship between three out of five risk measures and charter value. These results contradict the nonlinear U-shaped relationship.

Finally, I have also done the regressions with the unlagged control variables. For these regressions I have performed the same extra robustness checks by using a specification with and without capital ratio and return on assets as control variables. I report these results in tables A2.5 and A2.7 in the appendix. The findings from these robustness checks by excluding lagged control variables are almost the same as the findings from the Niu (2012) specification, with lagged variables, results. That is, I now find for both specifications highly significant negative coefficients for the systematic risk and nonperforming loans measures and a highly significant positive coefficient on the Z-score. The latter results are consistent with the linear negative relationship argued by Keeley (1990) and Demsetz et al. (1996).

5.3 Overall results

To summarize the findings mentioned in the empirical results section, the only significant nonlinear relationship I find is for the nonperforming loans ratio. This relationship is not a regular U-shaped relationship but an inverted U-U-shaped relationship. However, this result on the

(20)

20 systematic risk measure, the nonperforming loans ratio measure and the insolvency risk measure. The negative linear relationships are in line with the findings from Keeley (1990) and Demsetz et al. (1996). For the total risk and firm specific risks measures I do not find any significant results and charter value is therefore not an adequate variable to explain these risk measures.

6. Conclusion

According to Keeley (1990) is charter value an important measure to limit the risk-taking behavior of individual banks and could thereby help to preserve financial stability worldwide. This so called charter value paradigm states that if the charter value of the bank increases, banks typically will limit their risk-taking behavior in order to preserve the charter value and avoid high potential bankruptcy costs. This implies a negative relationship between charter value and bank risk. However, Boyd and De Nicolo (2006) challenge this view in their paper and argue that there is an opposing effect in the deposit and loan market. Their model argues that lower competition leads to higher default risk by borrowers as a consequence of higher demanded interest rates. As charter value is negatively related to competition, this would mean that there is a positive relationship between charter value and bank risk. Martinez-Miera and Repullo (2010) extended the model of Boyd and De Nicolo by allowing for imperfect correlation in loan defaults and find a U-shaped relationship

between competition and bank risk. Based on the findings of Martinez-Miera and Repullo (2010), Niu (2012) conducted a study to test for the U-shaped relationship between charter value and bank risk and indeed finds the U-shaped relationship for the American banking sector.

(21)

(22)

22

7. References

Arping, S., 2014. Market Power and Prudence in Dynamic Banking: The Charter Value Hypothesis Revisited. Available at SSRN 2386998.

Arellano, M., & Bover, O., 1995. Another look at the instrumental variable estimation of error-components models. Journal of Econometrics, 68, 29–51.

Beck, R., Jakubik, P., & Piloiu, A., 2013. Non-performing loans: What matters in addition to the economic cycle?. (Working Paper No. 1515). Retrieved from European Central Bank website: https://www.ecb.europa.eu/pub/pdf/scpwps/ecbwp1515.pdf

Berger, A. N., Klapper, L. F., Turk-Ariss, R., 2009. Bank competition and financial stability. Journal of Financial Services Research, 35(2), 99-118.

Berger, A. N., DeYoung, R., 1997. Problem loans and cost efficiency in commercial banks. Journal of Banking & Finance, 21(6), 849-870.

Besanko, D., Thakor, A. V., 2004. Relationship banking, deposit insurance and bank portfolio choice (No. 0411046). EconWPA.

Bliss, R. R., Flannery, M. J., 2002. Market discipline in the governance of US bank holding companies: Monitoring vs. influencing. European Finance Review, 6(3), 361-396.

Blum, J. M., 2002. Subordinated debt, market discipline, and banks' risk-taking. Journal of Banking & Finance, 26(7), 1427-1441.

Blundell, R., & Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics, 87, 115–143.

Boyd, J. H., De Nicolò, G., Jalal, A. M., 2006. Bank risk-taking and competition revisited [electronic resource]: new theory and new evidence. International Monetary Fund.

Brewer, E., Saidenberg, M. R., 1996. Franchise value, ownership structure, and risk at savings institutions (No. 9632). Federal Reserve Bank of New York.

Brunnermeier, M. K., Dong, G. N., Palia, D., 2012. Banks’ non-interest income and systemic risk. In AFA 2012 Chicago Meetings Paper.

(23)

23 Delis, M. D., Kouretas, G. P., 2011. Interest rates and bank risk-taking. Journal of Banking & Finance,

35(4), 840-855.

Demsetz, R. S., Strahan, P. E., 1997. Diversification, size, and risk at bank holding companies. Journal of money, credit, and banking, 300-313.

Demsetz, R. S., Saidenberg, M. R., Strahan, P. E., 1996. Banks with something to lose: The disciplinary role of franchise value. Economic Policy Review, 2(2).

Galloway, T. M., Lee, W. B., Roden, D. M., 1997. Banks' changing incentives and opportunities for risk taking. Journal of Banking & Finance, 21(4), 509-527.

Gan, J., 2004. Banking market structure and financial stability: Evidence from the Texas real estate crisis in the 1980s. Journal of Financial Economics, 73(3), 567-601.

Ghosh, S., 2009. Charter value and risk-taking: evidence from Indian banks. Journal of the Asia Pacific economy, 14(3), 270-286.

Hakenes, H., Schnabel, I., 2011. Bank size and risk-taking under Basel II. Journal of Banking & Finance, 35(6), 1436-1449.

Haq, M., Heaney, R., 2012. Factors determining European bank risk.Journal of International Financial Markets, Institutions and Money,22(4), 696-718.

Hellmann, T. F., Murdock, K. C., Stiglitz, J. E., 2000. Liberalization, moral hazard in banking, and prudential regulation: Are capital requirements enough?. American economic review, 147-165.

Jiménez, G., Lopez, J. A., & Saurina, J. (2013). How does competition affect bank risk-taking?. Journal of Financial Stability, 9(2), 185-195.

Jones, J. S., Miller, S. A., Yeager, T. J., 2011. Charter value, Tobin's Q and bank risk during the subprime financial crisis. Journal of Economics and Business, 63(5), 372-391.

Keeley, M. C., 1990. Deposit insurance, risk, and market power in banking. The American Economic Review, 1183-1200.

(24)

24 Marcus, A. J., 1984. Deregulation and bank financial policy. Journal of Banking & Finance, 8(4),

557-565.

Marquez, R., 2002. Competition, adverse selection, and information dispersion in the banking industry. Review of financial Studies, 15(3), 901-926.

Martinez-Miera, D., Repullo, R., 2010. Does competition reduce the risk of bank failure?. Review of Financial Studies, 23(10), 3638-3664.

Niu, J., 2012. An empirical analysis of the relation between bank charter value and risk-taking.

The Quarterly Review of Economics and Finance,52(3), 298-304.

Niu, J., 2008. Can subordinated debt constrain banks’ risk-taking?. Journal of Banking & Finance, 32(6), 1110-1119.

Park, S., 1997. Risk-taking behavior of banks under regulation. Journal of Banking & Finance, 21(4), 491-507.

Pathan, S., 2009. Strong boards, CEO power and bank risk-taking. Journal of Banking & Finance, 33(7), 1340-1350.

Repullo, R., 2004. Capital requirements, market power, and risk-taking in banking. Journal of financial Intermediation, 13(2), 156-182.

Roodman, D., 2006. How to do xtabond2: An introduction to difference and system GMM in Stata. Center for Global Development working paper, (103).

Salas, V., Saurina, J., 2003. Deregulation, market power and risk behaviour in Spanish banks. European Economic Review, 47(6), 1061-1075.

Saunders, A., Wilson, B., 2001. An analysis of bank charter value and its risk-constraining incentives. Journal of Financial Services Research, 19(2-3), 185-195.

Saunders, A., Strock, E., Travlos, N. G., 1990. Ownership structure, deregulation, and bank risk-taking. the Journal of Finance, 45(2), 643-654.

Stiglitz, J. E., & Weiss, A., 1981. Credit rationing in markets with imperfect information. The American economic review, 393-410.

(25)

25 Suarez, J., 1994. Closure rules, market power and risk-taking in a dynamic model of bank behavior.

LSE Financial markets group. 123-172.

(26)

26

8. Appendix

8.1 Appendix of the Data section Table A1.1

Number of banks per year

Year Number of banks

2006 2007 36 45 2008 48 2009 52 2010 54 2011 57 2012 59 2013 57 2014 54

Number of banks which have a respective year included in the dataset, from 2006 to 2014.

The difference between the number of banks per year is caused by the availability of bank level data in Bankscope and the data criteria. The data criteria is that the data of a bank is only included in the dataset if all independent variables are available to examine one year of the specification. Furthermore, year 2006 only functions as a source of information to perform the analysis for year 2007. This is because the methodology includes variables which consists of data from the preceding year, or t-1 variables (see methodology section). Hence, there should be at least two connecting years to be included in the sample. Table A1.1 reports the number of banks which have for a certain year the required data. For example, 54 out of 60 banks have data available over the year 2014.

Table A1.2

Number of banks per number of years available Number of years available Number of Banks

3 3 4 2 5 4 6 5 7 5 8 9 9 32

(27)

27 In table A1.2 I report how many banks in the dataset have a certain amount of years

available. The difference in data availability leads to a different number of samples per bank. For example, one bank might only have 2012 and 2013 data, while another might have data available for all nine years. In table A1.2 I report, for example, that 32 banks have all nine years available. The difference in the number of years available implies that an unbalanced dataset is used to perform the analysis.

8.2 Appendix of the Empirical results section Table A2.1

The relation between charter value and bank risk without the squared term.

(0.100) (0.143) (0.101) (0.123) (0.171) Charter value 0.006 -1.085** -0.005 -0.112 9.114 (0.023) (0.519) (0.009) (0.086) (5.861) Size 0.004* 0.200*** 0.001 0.023*** 1.123 (0.002) (0.076) (0.002) (0.008) (1.002) Capital ratio -0.145*** 1.370* -0.072*** -0.224** -35.904 (0.038) (0.703) (0.023) (0.098) (21.977) Return on assets -0.006 0.444 -0.106*** -0.601** 98.669*** (0.048) (0.990) (0.037) (0.275) (20.441)

Noninterest income ratio -0.002 -0.033 -0.006* 0.015 0.344

(0.003) (0.128) (0.003) (0.013) (0.691)

Cost income ratio -0.001 -0.071 -0.007 0.024** 0.586

(0.005) (0.109) (0.005) (0.012) (0.917) Subdebt -0.035 4.664 -0.104 0.662** 30.999 0.130 (6.903) (0.077) (0.265) (19.540) GDP-growth -0.123** -1.063 -0.076** 0.023 -10.556 (0.048) (1.250) (0.037) (0.228) (11.164) Population -0.005 0.052 0.003 -0.0188** 1.000 (0.004) (0.089) (0.004) (0.010) (0.907)

AR(1) test (p-value) 0.000 0.004 0.001 0.140 0.003

AR(2) test (p-value) 0.161 0.1384 0.062 0.851 0.987

(28)

28 Table A2.2

The relation between charter value and bank risk with the squared term included, but without Capital Ratio and Return on Assets as independent variables.

(0.116) (0.139) (0.130) (0.134) (0.080)

Charter value 0.229 -3.337 -0.066 0.231 22.653

(0.216) (4.472) (0.174) (0.522) (78.222)

Charter value squared -0.097 1.062 0.022 -0.187 -3.169

(0.087) (1.883) (0.070) (0.209) (31.604)

Size 0.003 0.191** 0.001 0.027*** 0.965

(0.003) (0.084) (0.002) (0.009) (0.736)

Noninterest income ratio -0.003 -0.059 -0.006 0.017 -0.827

(0.005) (0.137) (0.004) (0.015) (0.630)

Cost income ratio -0.003 -0.089 -0.008 0.026* -0.529

(0.007) (0.113) (0.007) (0.015) (0.641) Subdebt -0.039 4.186 -0.095 0.594** 53.421** (0.158) (6.419) (0.080) (0.302) (24.305) GDP-growth -0.118*** -1.306 -0.090** -0.170 -10.695 (0.037) (1.140) (0.043) (0.255) (12.967) Population -0.006 0.081 0.004 -0.020* 1.069 (0.006) (0.090) (0.005) (0.011) (0.747)

AR(1) test (p-value) 0.001 0.004 0.003 0.104 0.002

AR(2) test (p-value) 0.003 0.185 0.089 0.080 0.242

(29)

29 Table A2.3

The relation between charter value and bank risk without Capital Ratio and Return on Assets as independent variables and without the squared term.

(0.115) (0.141) (0.121) (0.126) (0.071)

Charter value 0.004 -1.035** -0.011 -0.195*** 15.643**

(0.035) (0.478) (0.016) (0.068) (7.089)

Size 0.003 0.197** 0.001 0.027*** 0.860

(0.003) (0.083) (0.002) (0.009) (0.692)

Noninterest income ratio -0.002 -0.045 -0.006 0.016 -0.813

(0.005) (0.135) (0.004) (0.015) (0.627)

Cost income ratio -0.003 -0.074 -0.008 0.027* -0.610

(0.008) (0.111) (0.007) (0.016) (0.644) Subdebt -0.054 4.121 -0.100 0.548** 50.584** (0.145) (6.583) (0.074) (0.248) (23.583) GDP-growth -0.112*** -1.305 -0.098** -0.169 -13.103 (0.037) (1.139) (0.040) (0.257) (12.887) Population -0.006 0.082 0.004 -0.020* 1.137 (0.007) (0.089) (0.005) (0.011) (0.779)

AR(1) test (p-value) 0.001 0.004 0.003 0.093 0.001

AR(2) test (p-value) 0.006 0.197 0.093 0.067 0.223

(30)

30 Table A2.4

The relation between charter value and bank risk with the squared term included, but without lagged control variables.

Lagged dependent variable 0.352** 0.054 0.240** 0.893*** -0.082

(0.149) (0.144) (0.117) (0.141) (0.064)

Charter value 0.100 -5.540 -0.124 0.499 19.057

(0.217) (4.671) (0.178) (0.512) (37.241)

(0.086) (1.978) (0.072) (0.204) (16.296) Size 0.006 0.222*** 0.003** 0.018** 0.019 (0.005) (0.084) (0.001) (0.007) (0.222) Capital ratio -0.044 -0.509 -0.012 -0.192 78.248*** (0.035) (1.531) (0.031) (0.169) (8.241) Return on assets 0.074 0.214 -0.073 -0.450 44.862* (0.076) (1.531) (0.060) (0.291) (26.260)

Noninterest income ratio -0.003 -0.030 -0.007 0.013 0.734*

(0.004) (0.087) (0.005) (0.012) (0.414)

Cost income ratio -0.003 -0.072 -0.008 0.023* 0.529

(0.006) (0.089) (0.007) (0.012) (0.377) Subdebt 0.107 4.388 0.017 -0.578* 26.711 (0.125) (3.119) (0.070) (0.301) (18.604) GDP-growth -0.150*** -1.267 -0.091*** 0.001 -16.101** (0.050) (1.437) (0.035) (0.203) (8.188) Population -0.009 0.087 0.001 -0.014 -0.076 (0.011) (0.093) (0.002) (0.010) (0.200)

AR(1) test (p-value) 0.010 0.016 0.002 0.141 0.000

AR(2) test (p-value) 0.015 0.283 0.085 0.299 0.384

(31)

31 Table A2.5

The relation between charter value and bank risk without lagged control variables and without the squared term.

Lagged dependent variable 0.339** 0.067 0.255** 0.908*** -0.082

(0.142) (0.146) (0.114) (0.137) (0.063) Charter value -0.001 -1.418*** -0.009 -0.160** 13.365*** (0.029) (0.456) (0.015) (0.080) (4.211) Size 0.006 0.220*** 0.003** 0.017 0.009 (0.005) (0.082) (0.001) (0.007) (0.228) Capital ratio -0.044 -0.564 -0.012 -0.190 78.647*** (0.035) (1.213) (0.031) (0.169) (8.413) Return on assets 0.080 0.137 -0.076 -0.421 44.204* (0.074) (1.598) (0.060) (0.292) (26.910)

Noninterest income ratio -0.003 -0.030 -0.007 0.013 0.765*

(0.004) (0.092) (0.005) (0.011) (0.414)

Cost income ratio -0.003 -0.070 -0.009 0.026** 0.584

(0.006) (0.088) (0.008) (0.013) (0.396) Subdebt 0.098 4.665 0.021 -0.634** 26.920 (0.121) (3.134) (0.076) (0.306) (18.236) GDP-growth -0.149*** -1.222 -0.090*** -0.004 -15.480** (0.050) (1.453) (0.033) (0.192) (7.902) Population -0.009 0.092 0.002 -0.015 -0.087 (0.011) (0.092) (0.002) (0.009) (0.207)

AR(1) test (p-value) 0.010 0.014 0.002 0.143 0.000

AR(2) test (p-value) 0.019 0.326 0.085 0.365 0.387

(32)

32 Table A2.6

The relation between charter value and bank risk with the squared term included, but without lagged control variables and without capital ratio and return on assets as independent variables.

Lagged dependent variable 0.409** 0.055 0.223* 0.918*** 0.037

(0.161) (0.143) (0.127) (0.128) (0.072)

Charter value 0.120 -6.435 -0.182 -0.080 57.587

(0.201) (4.501) (0.165) (0.599) (60.108)

(0.080) (1.901) (0.067) (0.242) (24.129)

Size 0.005 0.218*** 0.003** 0.019** 0.728

(0.004) (0.074) (0.001) (0.008) (0.612)

Noninterest income ratio -0.003 -0.019 -0.005 0.015* -0.671

(0.006) (0.094) (0.004) (0.009) (0.832)

Cost income ratio -0.004 -0.066 -0.007 0.029** -0.716

(0.006) (0.084) (0.007) (0.014) (0.609) Subdebt 0.100 4.584 0.032 -0.169 24.506 (0.113) (3.263) (0.082) (0.452) (47.456) GDP-growth -0.142*** -1.302 -0.106*** -0.228 -6.845 (0.045) (1.233) (0.037) (0.278) (13.864) Population -0.008 0.077 0.001 -0.017 0.713 (0.008) (0.090) (0.002) (0.011) (0.778)

AR(1) test (p-value) 0.007 0.014 0.005 0.089 0.002

AR(2) test (p-value) 0.005 0.278 0.099 0.081 0.245

(33)

33 Table A2.7

The relation between charter value and bank risk without lagged control variables, without capital ratio and return on assets as independent variables and without the squared term.

(0.149) (0.144) (0.121) (0.123) (0.071)

Charter value -0.002 -1.477*** -0.017 -0.218*** 17.991***

(0.032) (0.447) (0.017) (0.067) (6.304)

Size 0.005 0.212*** 0.003** 0.019** 0.686

(0.004) (0.072) (0.001) (0.008) (0.606)

Noninterest income ratio -0.003 -0.017 -0.005 0.016* -0.696

(0.005) (0.100) (0.004) (0.009) (0.767)

Cost income ratio -0.004 -0.064 -0.007 0.029** -0.739

(0.007) (0.088) (0.007) (0.014) (0.650) Subdebt 0.079 4.887 0.041 -0.175 22.814 (0.106) (3.312) (0.083) (0.413) (48.671) GDP-growth -0.135*** -1.334 -0.108*** -0.220 -8.292 (0.047) (1.223) (0.036) (0.273) (14.161) Population -0.008 0.082 0.002 -0.017 0.743 (0.007) (0.089) (0.002) (0.011) (0.780)

AR(1) test (p-value) 0.006 0.012 0.005 0.088 0.002

AR(2) test (p-value) 0.006 0.332 0.106 0.088 0.225