The impact of the leverage and liquidity coverage ratios from Basel III on bank risk-taking in Germany

(1)

Master’s thesis for the Msc Finance

The impact of the leverage and liquidity coverage ratios

from Basel III

on bank risk-taking in Germany

Author:

Supervisor:

Jan-Willem Goudkamp

Prof. dr. L. H. Hoogduin

Student ID:

Date:

2324369

Abstract

23-05-2017

(2)

2

I. Introduction

In the aftermath of the financial crisis, the international community agreed that banks should be subject to stricter regulations. This consensus led to the post-crisis international agreement on banking stability known as Basel III. Its goal is to make sure that banks do not once again engage in the risk-taking practices that were the primary cause of the crisis. The Basel III agreement was implemented in the United States as part of the Dodd-Frank Act and in Europe as part of Directive CRD IV.

This thesis investigates whether two of the ratios that were introduced in Basel III, the leverage ratio (LR) and the liquidity coverage ratio (LCR), are indeed effective in reducing bank risk-taking. It is a highly important topic, because risk-taking by banks can have severe consequences in terms of cost to the taxpayer, as was proven by the credit crunch2_{. The sample size in this study consists of German data.}

The fundamental idea behind Basel III, and capital requirements in general, is that with the government as a de facto active monitor on behalf of deposit holders, banks have less incentive to engage in risk-taking (Dewatripont and Tirole, 1994). However, the literature is highly divided on whether risk-taking is really reduced by these types of measures.

Firstly, there have been researchers who studied the impact of some of Basel III’s requirements on the degree of bank risk-taking empirically, and their results suggest that (some of) Basel III’s ratios indeed reduce risk-taking. Maria & Georgoulea (2016) and Adesina and Mwamba (2016) for example, find a negative relationship between risk and the leverage ratio. Chalermchatvichien et al. (2014) finds a similar relationship between risk and the net stable funding rate (NSFR).

1

Forbes (2011)

2

(4)

4

A second strand of literature is skeptical with respect to the impact of capital requirements in general. They argue that bank funding costs are likely to increase as a result of (equity) capital requirements, with a potential adverse impact on bank value and on the economy in general (Cummings & Wright, 2016). In addition, capital requirements could lead to banks taking more risks in order to ensure the same profit as before the implementation of the requirements (Koehn & Santomero, 1980).

Some authors, on the other hand, are in favour of capital requirements, but are not satisfied with Basel III and doubt its effectiveness (for example: Moosa & Burns, 2013 and Blundell-Wignall & Atkinson, 2010). Firstly, because the main criticism on Basel II was that it used only a risk-weighted capital ratio, and this practice has not been abandoned fully in Basel III (Blundell-Wignall & Atkinson, 2010). A second often-mentioned criticism on Basel III is that the leverage ratio is too low to effectively reduce the probability of another systemic crisis in the future (Admati & Hellwig, 2013).

In order to investigate whether the LR and LCR from Basel III reduce risk-taking, we perform ordinary least-square regressions with the score as dependent variable. The z-score is a common and accurate indicator of bank risk-taking (Laeven & Levine, 2009 and Chiaramonte et al., 2016). It is defined by the difference between the return on assets and the capital/asset ratio divided by the standard deviation of the return on assets. Similar to Laeven & Levine (2009) we perform a robustness check with equity volatility as an alternative measure of bank risk-taking. Decreasing bank values are associated with higher risk-taking, so we perform an additional regression with Tobin’s Q, an indicator of bank value, as dependent variable.

In a fashion similar to Laeven & Levine (2009) we use the loan loss provision, total deposits, earnings and the degree of ownership concentration as independent variables. We add the LR and the LCR to these independent variables. Our results indicate that the LR should increase to around 22%. The LCR seems to have no impact on risk-taking, which is likely due to recent ECB policies aimed at improving bank liquidity.

(5)

5

II. Literature review

This section provides the theoretical background behind this study. First, this literature review provides an overview of the history of the Basel Accords. Afterwards, it discusses three relevant strands of literature. This section ends with a short conclusion on the contribution of this study to the existing literature on Basel III and capital requirements.

2.1 History of Basel III

The Basel Committee on Banking Supervision was established in 1974 by central bankers of the Group of Ten (G10).3_{Its goal is primarily to improve global banking}

regulation. As such, the committee designs guidelines on banking supervision (BIS, 2016). In the 1980s, the Basel Committee started to become concerned about capital adequacy ratios, as the Latin American debt crisis worsened (BIS, 2016). This led to the Basel Capital Accord, more commonly known as Basel I, in 1988. The main idea of this framework was to introduce a minimum capital-to-risk-weighted assets ratio of 8%. This framework was ultimately implemented in all countries with internationally operating banks (BIS, 2017).

In 2004, a new Basel Capital Accord (known as Basel II) was released to the public, such that ‘recent financial innovations would be properly adressed’ and ‘capital adequacy requirements would better reflect the underlying risks’ (BIS, 2016). It was based on three pillars (BIS, 2016):

1. minimum capital requirements, including a capital requirement for common equity over risk-weighted assets of 2%, which sought to develop and expand the standardised rules set out in the 1988 Accord;

2. supervisory review of an institution's capital adequacy and internal assessment process;

3. effective use of disclosure as a lever to strengthen market discipline and encourage sound banking practices

3_{The so-called Group of Ten, or G10, consisted of Belgium, Canada, France, Germany, Italy, Japan, Netherlands,}

(6)

6

In 2008, it turned out that Basel II had some weaknesses as its implementation had not prevented the banking crisis. This led to a new accord, known as Basel III, which will be phased in gradually until 2018. Its aim is to decrease the likelihood of systemic crisis such as the credit crunch occurring in the future (BIS, 2016). Basel III contains several adjustments to Basel II.

First of all, the capital requirement for common equity over risk-weighted assets was to be increased to 4.5% from 2%. This capital requirement is higher (at maximum 1.5% higher) for so-called Globally Systemically Important Banks (GSIBs), because these banks are so big that a bankruptcy of one of these banks would likely threaten the entire financial system. In other words, these banks are considered ‘too big to fail’. Which banks are considered GSIBs is determined annually by the Financial Stability Board on the basis of four criteria, these being: (a) size, (b) cross-jurisdiction activity, (c) complexity, and (d) substitutability (European Parliament, 2016).

Secondly, a countercyclical capital buffer was introduced in Basel III, which allows governments to require extra reserves from banks during periods of high credit growth. Thirdly, three new minimum ratios were added: the leverage ratio (LR), the liquidity coverage ratio (LCR) and the net stable funding ratio (NSFR). The LR is not based on risk-weighted assets and requires a minimum amount of capital relative to all of a bank’s assets (BIS, 2016).

The LCR requires that banks have sufficient cash to survive a 30 days’ stress period (BIS, 2013). The NSFR addresses maturity mismatches (BIS, 2013).

The leverage and liquidity coverage ratios have already (partly) been implemented, but the NFSR will be implemented in January 2018. This paper thus excluded the NSFR because of data availability reasons.

(7)

7 Table 1: Basel III requirements

Capital requirements:

 Common equity should be 4.5% relative to the total risk-weighted assets. Maximum of 1.5% higher for systemically important banks.

 At least 3% of Tier 1 capital should be held as a reserve against total exposure (leverage ratio).

Liquidity requirements:

 The liquidity coverage ratio (LCR), which requires that high-quality liquidity cash inflow should be higher than the net outflow during a 30 day stress period.

 The net stable funding ratio (NFSR), which requires that the available of amount of stable funding should exceed the required amount of stable funding over a one-year period of extended stress.

Capital buffers:

 A conservation buffer of 2.5%

 A countercyclical buffer that allows governments to require 2.5% additional capital during periods of high credit growth.

Note: Basel III has been implemented in the European Union, first under Directive 2013/36/EU (CRD

IV), then under EU Regulation 575/2013. In the United States, Basel III was implemented as part of the Dodd-Frank Act. Information for this table was retrieved from the website of the Bank for International Settlements (2013).

2.2 Theoretical idea behind Basel III

Banks inherently have an incentive to engage in risk-taking activities by taking on a high amount of leverage (Li, 2017 and Jensen & Meckling, 1976). This incentive is even larger when investment decisions are not verifiable to outsiders (Li, 2017). Large debtholders may be able to closely monitor the bank, but the regular, uninformed holders do not have the means nor an incentive to monitor the bank because their deposits are usually guaranteed up to a certain amount by governments (Dewatripont & Tirole, 1994). In order to make sure that banks do not take on too much leverage, banks should thus be subject to proper regulation, where the government acts as an active monitor on behalf of the small deposit holders (Li, 2017 and Dewatripont & Tirole, 1994).

(8)

8

financial ratios to make sure to the public that they are still liquid and solvent. However, the opinion of economists on the impact of Basel III and capital requirements in general on bank risk-taking is highly divided.

2.3 Studies in favour of Basel III

Several empirical studies have been conducted whose results seem to support the notion that Basel III would be effective in reducing bank risk-taking.

Chalermchatvichien et al. (2014), for example, find that the implementation of Basel III likely decreases bank risk-taking, measured by the z-score. The z-score is defined as the return on assets minus the capital/assets ratio over the standard deviation of the return on assets. It is an often used proxy for bank risk-taking (see for example Leaven & Levine, 2009) as it accurately measures the inverted probability of bank failure (Chiaramonte et al, 2016). Chalermchatvichien et al. (2014) estimate the ex-ante effects of Basel III in Asia, by estimating the net stable funding ratio (NSFR) had Basel III been implemented in Asia in the first decade of the 21st_{century and performing a regression analysis between this}

estimated NSFR and the z-score. Their results suggest that an improvement in capital stability (measured by the NSFR) by one standard deviation would likely decrease the amount of risk-taking (measured by the z-score) by 5.37%.

Adesina and Mwamba’s (2016) findings also support Basel III. They observe that South African banks with lower common equity capital relative to total bank value have a higher risk-appetite, also measured by the z-score (Adesina and Mwamba, 2016). These results support Basel III: the ratio between equity capital and total value is essentially the leverage ratio.

(9)

9

2.4 Arguments against capital requirements

Most criticism on capital requirements points out that an increase of these capital requirements would likely lead to higher funding costs, with all sorts of negative side-effects (for example: Cummings & Wright, 2013 and Koehn & Santomero, 1980).

An often-heard argument is that more equity financing leads to an increase in funding costs, because shareholders always require a higher risk compensation than debtholders (Admati & Hellwig, 2013). It would thus be logical to think that banks’ funding costs must inherently increase if they would have to hold on to more equity, as is the case with Basel III’s leverage ratio.

However, this reasoning is contrary to standard finance theory. The well-known Modigliani & Miller (1958) theorem states that funding costs are independent of capital structure. The total funding costs depend on the overall risk profile of a firm. It should, in the absence of taxes, not matter, in terms of funding costs, whether a firm is financed with debt or equity (Modigliani & Miller, 1958).

Admati & Hellwig (2013) provide a slightly alternative explanation for the causal link between an increase in equity capital requirements and higher funding costs. They argue that equity is relatively expensive compared to debt because of the existence of a so-called taxpayer’s subsidy on debt (Admati & Hellwig, 2013). Governments tend to bail out banks that are likely to go bankrupt and provide deposit insurance to deposit holders. According to Admati & Hellwig (2013), banks can borrow relatively cheaply because the deposit holder is sure to get his money – their bank will either get bailed out or deposit holders can apply for deposit insurance. As a result, the interest rate deposit holders demand from their banks should be lower than what it would normally be. Banks thus have a significant benefit of financing their operations with debt instead of equity, arguably because of the taxpayer’s subsidy.

Taking away part of this benefit, by means of imposing a minimum level of equity-financing, will inherently increase the costs of funding for banks.

(10)

10

likely have a detrimental impact on the general economy as a result of the fact that banks will charge the higher funding fee to customers.

Secondly, the value of banks may decrease because of the increase in funding costs. After all, the present value of future earnings goes down when the funding costs go up. Konishi & Yasuda (2004) and Keeley (1990) show that a decrease in bank value is likely to induce risk-taking in banks, because when a bank has a relatively low value, the owners have less to lose, which will make them less risk-averse.

Thirdly, Koehn & Santomero (1980) argue that the increase in funding costs may incentivise banks to take on more risks to ensure the same return as before the implementation of the capital requirement through asset substitution. As such, they argue that the probability of default increases as a result of capital requirements, which is the opposite of the regulatory intent (Koehn and Santomero, 1980).

2.5 Studies that argue that Basel III is insufficient

There are also authors who argue that the ratios from Basel III are in principle useful, but not sufficient in reducing incentives for risk-taking. This strand of literature thus agrees with the first strand (proponents of Basel III) in the sense that they believe the idea behind Basel III is good. However, while the first strand argues that Basel III is sufficient, this third strand is of the opinion that Basel III is not extensive enough to effectively counter excessive risk-taking.

Blundell-Wignall & Atkinson (2010), in their elaborate critique on Basel III, and Moosa & Burns (2013), for example, argue that the fact that the capital ratio was risk-weighted under Basel II, likely turned out to be one of the indirect causes of the credit crisis. According to them, this problem has not been addressed fully yet in Basel III.

(11)

11

which is a minimum of capital that banks should reserve against total assets instead of

risk-weighted assets. However, the Basel Committee sees this leverage ratio as a

‘supplementary tool’, not as a primary tool (Moosa & Burns, 2013).

Blundell-Wignall & Atkinson (2010) argue that ‘the leverage ratio should not be thought of as a backstop measure, given how effective (or rather ineffective) the capital weighting approach has been’. Moosa & Burns (2013) add that the leverage ratio is indeed much more objectively calculated and should therefore be used as the ‘prime’ measure of capital control.

In addition, a considerable amount of authors argue that the current level of the leverage ratio (3%) is not high enough to reduce the probability of another systemic crisis occurring in the future (Financial Times, 2010).

Systemic risk would arguably be much lower if banks had to adhere to a higher leverage ratio. Borrowing, as explained by Admati & Hellwig (2013), magnifies risk. If you borrow for a house of 200.000 euros with 30.000 euros ‘equity’, a decrease of 15% (15%*200.000=30.000) in value would mean that you lose all your equity. Banks, under Basel III, only have to keep 3% in equity. A value drop in total assets of only 3% would thus mean that banks effectively lose all their equity. Admati & Hellwig (2013) therefore argue that the leverage ratio should be much higher. They propose a leverage ratio of about 20-30% at minimum.

(12)

12

It is interesting to note that in 2010, when the details of Basel III were being negotiated, several prominent economists4_{signed a letter that was sent to the Financial}

Times calling for a higher leverage ratio of at least 15% (Financial Times, 2010).

While most criticism on the effectiveness of Basel III focuses on the leverage ratio and the risk-weighted assets, there is also criticism on other parts of the Accord. Blundell-Wignall & Atkinson (2010), for example, are highly skeptical of the newly introduced liquidity coverage and net stable funding ratios. They argue that the liquidity ratios are unnecessary, because the liquidity shortage during the credit crunch was actually caused by solvency issues. Blundell-Wignall & Atkinson (2010) state that: ‘The cause of the crisis was a solvency problem, after which uncertainty arose as to banks’ ability to pay which, in turn, led to a buyers strike affecting short-dated funding.’

Furthermore, Blundell-Wignall & Atkinson (2010) argue that as long as banks are solvent they should be left to deal with liquidity issues alone. Managing maturities is an essential part of the banking business, and banks should not be considered ‘naïve at their own business’ (Blundell-Wignall & Atkinson, 2010). If large liquidity issues arise in the market, it is the task of the central bank to guarantee the stability in the system (Blundell-Wignall & Atkinson, 2010).

2.6 Contribution of this paper to the literature

From the above, it becomes clear that there are main three strands of literature relevant to the discussion on the impact of Basel III.

Firstly, there are the proponents of Basel III, who believe that banks are inherently drawn to high leverage which in turn leads to higher risks (Dewatriponte & Tirole, 1994). Basel III puts a maximum to this leverage and should thus be effective in reducing risk-taking. Chalermchatvichien et al (2014), for example, find a positive relationship between the z-score and the NSFR.

Secondly, there are critics of capital requirements in general. They believe that capital requirements will lead to increased funding costs of banks. An increase in funding costs has several drawbacks. For example, bank value may go down which can potentially lead to more risk-taking (Keeley, 1990). In addition, increased funding costs may lead to a

4

(13)

13

reduction in total lending with an adverse impact on the economy (Cummings & Wright, 2016). Furthermore, the increased funding costs may incentivize banks to take more risks in order to ensure the same earnings as before the implementation of the capital requirement (Koehn and Santomero, 1980).

Thirdly, there are authors that are highly in favour of capital requirements, but who are not satisfied with Basel III. Firstly because Basel III still incorporates ratios on risk-weighted assets, but also because the leverage ratio is too low in their view (Blundell-Wignall & Atkinson, 2010). Furthermore, Blundell-(Blundell-Wignall & Atkinson (2010) argue that the liquidity ratios included in Basel III are unnecessary with an active central bank.

The main aim of this study is to extend this discussion by investigating the impact of the leverage and liquidity coverage ratios on risk-taking in German banks in 2014-2015. We will check whether these ratios have a positive, an insignificant or a negative (statistical and economic) relationship with the z-score in order to see which strand of literature relates best to our observations.

(14)

14

III. Data, methodology and descriptive statistics

This section first discusses the sample used in this study. Afterward, it will provide a first investigation into the relation between the most important variables.

3.1 Sample

The sample consists of annual cross-sectional data of 768 German banks from the years 2014 and 2015. This data was obtained from Orbis Banks Focus.

This study focuses on Germany because this country has the highest concentration of banks in Europe and is the largest economic power in the Eurozone.

Germany’s banking system can be characterized as a so-called ‘Three-Pillar’ system (Brunner et al, 2004). This means that the country has three types of banks: public sector banks (the Landesbanken and the Sparkassen), co-operatives and regular commercial banks (Brunner et al, 2004). It could be argued that public sector banks should be left out of the sample because they are less susceptible to taking excessive risks, as their main goal is not necessarily to make profit. However, several public sector banks needed to be bailed out during the credit crisis.5_{Therefore, this study still includes these banks, also because it}

was found that German bailouts led to an increased z-score in, on average, all German banks (Dam & Koetter, 2009). Basel III was implemented in order to reduce exactly this type of bank behavior, which is why this study includes all German banks in the sample.

5

(15)

15

3.2 Methodology and description of variables

In order to assess whether Basel III has the intended impact on the degree of risk-taking by the banks in the sample, this paper follows the methodology of Laeven & Levine (2009) and Chalermchatvichien et al. (2014).

Hence, the following equation is estimated:

ln (𝑧 − 𝑠𝑐𝑜𝑟𝑒) = 𝛼 + 𝛽 (𝐿𝑜𝑎𝑛 𝑙𝑜𝑠𝑠 𝑝𝑟𝑜𝑣𝑖𝑠𝑖𝑜𝑛) + 𝛿(𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠/𝑎𝑠𝑠𝑒𝑡𝑠) + 𝜀(𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠) + 𝜖 (𝐿𝐶𝑅) + 𝜃 (𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 𝑟𝑎𝑡𝑖𝑜) + 𝜗 (𝑂𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝 𝑐𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛) (1)

Equation (1) estimates the natural logarithm of the z-score as a function of several controlling factors. Aside from the z-score, an alternative measure of banking risk is the volatility of equity (Laeven & Levine, 2009). The volatility of equity will thus serve as a robustness check. Tobin’s Q will also be used as a proxy for banking risk, since a low bank value is associated with greater risk (For example: Keeley, 1990 and Jones et al, 2011).

As was mentioned earlier, Laeven & Levine (2009) and Chalermchatvichien et al. (2014) use a similar equation to assess the effect of certain factors on banking risk; however, there are a couple of differences.

The main difference is that this study is the first study that includes both the LCR and the LR as independent variables in the equation in order to assess its effect on the z-score.

Another key difference is that this thesis uses Tobin’s Q as a dependent variable whereas Laeven & Levine (2009) as well as Chalermchatvichien et al. (2014), use Tobin’s Q solely as an independent variable. The reason for this is that, among others, Jones et al. (2013) found that a lower bank value is associated with higher bank risk. To see how the LR and the LCR interact with bank value could thus be valuable to this study, even more so because capital requirements likely to negatively affect bank value as they may increase the total cost of funding for a bank (Cummings & Wright, 2015).

(16)

16

ratio instead of average earnings growth. The reason for this is that the period over which the data for this study was gathered is only limited due to the fact that Basel III was implemented only in 2013.

Finally, this study uses a dummy variable (which takes into account if the bank has an ultimate shareholder or not) to control for ownership concentration, whereas the other two studies use the amount of cashflows to major shareholders for that purpose. This study chose to make use of information on ultimate shareholder ownership instead, because only limited data was available on cashflows to major shareholders.

Subsections 2.1 until 2.6 provide an extensive explanation of each variable used in the analysis in this thesis.

1) Z-score

This study uses the z-score as a measure of banking risk because the z-score has proven to be an accurate predictor of bank failure (Chiaramonte et al., 2016 and Lepetit & Strobel, 2015). Laeven and Levine (2009) calculate the z-score as follows:

𝑧 = (𝑅𝑜𝐴+𝐶𝐴𝑅)

𝜎(𝑅𝑜𝐴) (2)

Where ROA is the return on assets, CAR is the capital-to-assets ratio and σ(RoA) is the standard deviation of the return on assets. As the z-score essentially estimates the inverted probability of bankruptcy of a bank, a higher value indicates that a bank is less likely to fail (Chiaramonte et al., 2016). Similar to Chalermchatvitchien et al. (2014) we use the natural logarithm of the z-score because using normal z-scores would lead to highly skewed results. Skewed results would make regression analysis virtually impossible since one of the underlying assumptions of ordinary least-square regressions is that all the variables need to be normally distributed (Brooks, 2003).

2) Tobin’s Q `

(17)

17

of liabilities and dividing this by the sum of total assets (Lemmens, 2012 or De Jonghe et al., 2007). Keeley (1990) found that decreasing values (measured by Tobin’s Q) in the 60s and 70s may have provided an incentive for the increase in risk-taking behaviour by banks in the 80s. Furthermore, both Konishi & Yasuda (2004) and Weisbrod et al. (1992) associate lower values with higher degrees of risk-taking in Japanese banks prior to the financial crisis in Japan during the 90s. A link between higher degrees of risk-taking and lower values was also observed by Jones et al. (2011) during the period before the credit crisis in 2008.

According to Laeven & Levine (2009) it should be calculated as:

𝑇𝑜𝑏𝑖𝑛′𝑠 𝑄 = (𝐸𝑄 + 𝐿)/𝐴 (3)

EQ equals the market value of equity, L the value of liabilities and A the value of assets. Tobin’s Q was solely used as an independent variable in Laeven & Levine (2009) and in Chalermchatvichien et al. (2014). The observation that lower bank value is associated with higher bank risk-taking implies that Tobin’s Q might be used as a possible proxy for the degree of bank risk-taking. This makes gaining an insight into the interaction between the LR, the LCR and Tobin’s Q highly valuable.

3) Volatility of equity

According to Laeven & Levine (2009), volatility of equity can be used as a measurement of banking risk. They calculate daily stock volatility, but this data was unavailable for a large number of banks in the sample in this study. Therefore, we calculate the volatility of equity as follows:

𝑣𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑒𝑞𝑢𝑖𝑡𝑦 = 𝜎 (𝑒𝑞𝑢𝑖𝑡𝑦 𝑣𝑎𝑙𝑢𝑒), 𝑡 − 1, 𝑡 − 2 (4)

(18)

18

similar relationship should exist since equity volatility is also a measure of bank risk-taking (see for example: Laeven & Levine, 2009 and Chalermchatvichien et al., 2014).

However, as yearly equity data is only available since 2013, and because we use equity volatility over the past three years, we were only able to perform the robustness check for the year 2015.

4) Basel III: the LCR and LR

This study uses the LCR and LR as Basel III variables. The LCR measures how capable a bank is of fulfilling its short-term obligations, relative to its income. It is calculated as follows:

𝐿𝐶𝑅 = 𝐻𝑖𝑔ℎ 𝑞𝑢𝑎𝑙𝑖𝑡𝑦 𝑙𝑖𝑞𝑢𝑖𝑑 𝑎𝑠𝑠𝑒𝑡𝑠

𝑇𝑜𝑡𝑎𝑙 𝑛𝑒𝑡 𝑐𝑎𝑠ℎ 𝑜𝑢𝑡𝑓𝑙𝑜𝑤𝑠 𝑜𝑣𝑒𝑟 𝑡ℎ𝑒 𝑛𝑒𝑥𝑡 30 𝑑𝑎𝑦𝑠> 100% (5)

The LCR thus requires that the high-quality liquidity cash inflow exceeds the net cash outflow during a 30 days stress period. Practically, this means that a bank needs sufficient liquidity to survive for up to 30 calendar days in a stress scenario similar to the events during the financial crisis in 2008 (BIS, 2013). It is expected that a higher LCR should reduce banking risk since a higher ability to fulfill short-term obligations ought to reduce the probability of distress.

The LR is a minimum amount of core capital that banks should hold over total exposure:

𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒 𝑟𝑎𝑡𝑖𝑜 = 𝑇𝑖𝑒𝑟 1 𝐶𝑎𝑝𝑖𝑡𝑎𝑙

𝑇𝑜𝑡𝑎𝑙 𝐸𝑥𝑝𝑜𝑠𝑢𝑟𝑒 > 3 % (6)

(19)

19

5) Ownership Concentration

Laeven & Levine (2009) were the first to find a highly significant relation between bank risk-taking and ownership concentration. They found that the greater the cashflow rights of the largest owner, the higher the bank’s z-score, and thus the risk. It would thus make sense to include ownership concentration as a controlling variable. As a proxy for ownership concentration this study will use data from Orbis Banks’ Focus on ultimate shareholder ownership. According to this database 50 German banks from the sample in our study have such an ultimate owner who owns at least 50.01% of the shares.

It is noteworthy that the public banks are not all fully owned by the state. The word

public purely means that they serve a public good, but not that the state fully owns them

(VÖB, 2017). The ‘ultimate owner’ variable is thus not a simple proxy for state-ownership. In order to control for ownership concentration this study will introduce a dummy variable, with value 1 for banks that have an ultimate owner and 0 for other banks. A full list of these banks can be found in the appendix.

6) Other controlling variables

(20)

20

3.3 Descriptive statistics

Table 2: Mean, median and standard deviation for key variables

Variable Mean Median Standard deviation Observations

Z-score 2.02 1.81 1.41 1468

Tobin’s Q 88 89 3 1468

Volatility of equity 14 8 66 760

Earnings over assets 0.50 0.50 0.41 1468

Loan loss provision 0.80 3.6 20.6 1478

Deposits 5.53 5.59 0.65 1478

Leverage ratio 8,2 7,8 2,3 759

Liquidity coverage ratio 275.4 258 136 1158

Note: The z-score per bank is calculated as the logarithm of the sum of the return on average assets and the

capital-to-assets ratio divided by the standard deviation of the return on average assets over the last couple of years. Tobin’s Q is the sum of equities at market value and liabilities divided by the total assets. The volatility of equity is the standard deviations of equity for the three years prior to 2015. The number of observations is thus smaller than for the other variables. The earnings over assets are the earnings divided by the total assets. The loan loss provision is the loan loss provision divided by net interest revenue. Deposits are the logarithm of the total amount of deposits held by the banks’ debtors in the relevant year. The leverage ratio is the amount of core capital over the total exposure. The leverage ratio was only available for 759 datapoints over 2015, hence the fewer amount of observations. The liquidity ratio is calculated by dividing the high-quality liquid assets by the net cash outflows.

The average z-score is somewhat lower than in Chalermchatvichien et al. (2014) and in Laeven and Levine (2009). Laeven & Levine (2009) use a global sample, whereas the sample of Chalermchatvichien et al. (2014) is limited to Asian banks. Chalermchatvichien et al. (2014) find that Asian banks have on average a higher z-score than the global average found by Laeven and Levine (2009). The average z-score in Germany seems to be slightly lower than the global average. This result can reasonably be expected if one takes into account the fact that German banks are relatively higher leveraged than in other countries (New York Times, 2008).

The fact that profits are somewhat lower in Germany could also be the reason that Tobin’s Q seems to be somewhat lower in German banks than in the Asian sample used by Chalermchatvichien (2014). After all, Tobin’s Q is an indicator of bank value, and when bank value is seen as the value of the bank’s future profits, bank value is likely to be lower when the cost of funding increases (Cummings & Wright, 2016).

(21)

21

100%. On average the LCR is 276,3%, indicating that German banks are highly liquid indeed.

Table 3: Correlation matrix for key variables

Variable 1 2 3 4 5 6 7 8 9 1 Z – score 1 2 Equity volatility -0.12*** 1 3 Tobin’s Q 0.01 -0.07** 1 4 Loanloss 0.00 -0.21*** 0.00 1 5 Deposits 0.13*** 0.02 0.11*** -0.02 1 6 Earnings/Assets -0.01 -0.17*** -0.14*** 0.69*** -0.08*** 1 7 LCR -0.09 0.00 0.03 0.00 -0.04*** 0.00 1 8 Leverage ratio 0.04*** -0.10*** -0.43*** 0.10*** -0.19*** 0.25*** 0.11*** 1 9 Ownership dummy -0.15*** 0.07** -0.16*** 0.04 -0.43*** -0.02 -0.03 -0.11** 1

Note: * denotes significance at 10%, ** denotes significance at 5%, *** denotes significance at 1%. Variables 1 to 6

incorporate historical data going back at least 2 years. The leverage ratio is available for 2015. The ownership concentration dummy is assumed for 2015. Equity volatility is calculated for 2015 with data going back to 2013.

Table 3 gives an overview of the correlations between each of the key variables and the significance of each of these correlations, to give a first impression of how variables are related to each other. The three most important conclusions that can be drawn from this correlation table are:

Firstly, that it seems that there are no indications for multicollinearity as none of the variables appear to be highly correlated.

Secondly, there seems to be a significant negative correlation of 0.12 between the z-score and equity volatility. The interpretation of this result is that for each percentage point that the equity volatility goes up, the z-score goes down by 0.12. This result is to be expected since higher equity volatility implies higher risk-taking, whereas a lower z-score is also an indicator of higher risk-taking. It seems that adding equity volatility as a dependent variable constitutes a proper robustness check.

Additionally, it seems that the dummy variable for ownership concentration is significantly negatively correlated with both the z-score and Tobin’s Q, while it is significantly positively correlated with equity volatility. This could imply that indeed a higher degree of ownership concentration increases riskiness in banks, similar to Laeven & Levine, 2009).

(22)

22

riskiness of banks. The correlation between equity volatility and the leverage ratio suggests the same: there is a negative relation between the leverage ratio and equity volatility, which implies that the leverage ratio could decrease volatility and thus risk.

However, it seems that there is a negative relation between the leverage ratio and bank value measured by Tobin’s Q. This could confirm the theory from Cummings & Wright (2013) that an increased leverage ratio will increase the cost of funding which may lead to a lower valuation of future earnings. A low bank value increases the likelihood of excessive risk-taking among banks, which implies that the significant positive correlation with the z-score does not necessarily mean that the ratios decrease banking risk (Keeley, 1990).

There seems to be no significant correlation between on the one hand the liquidity ratio and on the other hand the z-score, Tobin’s Q and equity volatility.

(23)

23

IV. Main Findings

This section gives an overview of the main findings of this study. First, we shortly discuss the impact of each independent variable on the z-score. Afterwards, this study will look at the relationship between these independent variables and the volatility of equity as part of a robustness test. The relationship between value (Tobin’s Q) and the independent variables will also be investigated. Section 4.3 will be concerned with the implications of our three main findings.

4.1 The impact of the independent variables on the z-score

Table 4 shows the results of a regression analysis estimating the impact of the independent variables on the z-score. In each model between (1) and (6) an independent variable is added, in order to come up with an equation that tests the impact of the LCR and the leverage ratio amongst other variables on the z-score.

Table 4: The impact of the independent variables on the z-score

Firstly, it should be noted that the explanatory power of the model, denoted by the adjusted R2_{seems to be fairly low. It is slightly lower than similar models estimated by}

Laeven & Levine (2009) and Chalermchatvichien (2014), both of which have an R2 _of

Dependent variable (1) ‘ Z-score (2) ‘ Z-score (3) Z-score (4) Z-score (5) Z-score (6) Z-score Constant 0.44 0.43 0.45 0.36 -0.13 -0.09 (0.16) (0.16) (0.17) (0.37) (0.80) (0.88) Deposits 0.29*** 0.29*** 0.29*** 0.33*** 0.46*** 0.47*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Loan loss provision 0.001 (0.75) 0.001 (0.70) 0.001 (0.82) 0.003 (0.5) 0.003 (0.5) Earnings/Assets -0.03 (0.82) 0.02 (0.9) -0.42** (0.05) -0.44** (0.05) Liquidity coverage ratio -0.001*** (0.00) -0.001 (0.15) -0.00 (0.20) Leverage ratio 0.07*** (0.00) 0.08*** (0.00) Ownership concentration -0.82*** (0.00) Observations 1465 1458 1456 1158 560 550 Adj. R2 _0.02 _0.02 _0.02 _0.03 _0.07 _0.09

Note: This table shows the outcome of six regressions. Regression (1) to (4) consist of data from

(24)

24

around 0.2. This lower value is likely due to the fact that the sample in this study concerns a much shorter period than the aforementioned studies, due to the data availability constraints of this study.

Despite the relatively low predictive value of the model, this study still provides some useful insights into the relationship between the independent variables and the degree of bank risk-taking.

Firstly, the results from table 4 suggest a significant positive relation between deposits and the z-score. Since deposits are used as a proxy of size, it seems as though larger banks carry less risk. This is in line with Kasman (2016) who finds that larger (Turkish) banks carry less risk.

A possible explanation for this phenomenon would be that larger banks are more able to diversify risks and are therefore less likely to default . Nevertheless, there are also authors who find an opposite relationship. For example Rahman et al. (2015) find that larger banks in Bangladesh tend to take larger risks than relatively smaller banks. To further investigate the relationship between size and bank risk-taking goes beyond the scope of this paper, but would be a good suggestion for further research.

Secondly, there seems to be no statistically and economically significant effect of the loan loss provision on the z-score. This finding is in line with Laeven & Levine (2009), but not with Chalermchatvichien et al. (2014). A possible explanation for this finding could be that the loan loss provision should be lagged by one period, since having a reserve for unexpected losses in the previous year can only result in lower probability of default one period later. Therefore, this study performed a regression analysis (which can be found in the appendix) with the lagged loan loss provision, but the result did not become more significant. To further assess the impact of the loan loss provision on the z-score, however, goes beyond the scope of this paper as well.

(25)

25

It seems as though the LCR does not have a statistically or economically significant impact on the z-score. This result was not expected. After all, one would expect that a higher LCR should reduce banking risk since a higher ability to fulfill short-term obligations ought to reduce the probability of distress. This finding could support the opinion of Blundell-Wignall & Atkinson (2010), who argue that the introduction of the LCR is superfluous, because ensuring the liquidity of the financial system is already the primary task of central banks.

Contrary to the LCR, the LR has a highly significant correlation of 0.07 with the z-score. This implies that an increase of one percentage point in the leverage ratio increases the z-score with 0.07, which suggests that indeed the LR decreases risk-taking by banks, as is expected by Yan et al., 2012 and found by Maria & Georgoulea (2016) for Greek banks.

While it seems that the LR has a statistically significant negative impact on the z-score, the question remains whether this result is economically significant enough. A 1 % increase in the leverage ratio relates to an increase in the z-score of 0.07. Is this enough? By how much does the z-score need to decrease to decrease the probability of another systemic crisis such as the credit crunch? We will return to this fundamental question in the implications paragraph.

(26)

26

4.2 Robustness check with equity volatility and Tobin’s Q as alternative dependent

variables

Table 3 depicts the results of the regression analyses with equity volatility and Tobin’s Q as alternative measures of risk instead of the z-score.

Table 5: Robustness check with equity volatility and with Tobin’s Q Dependent variable (1) Equity volatility (2) Equity volatility (3) Equity volatility (1) Tobin’s Q (2) Tobin’s Q (3) Tobin’s Q Deposits 3.73 (0.40) 2.96 (0.46) 4.24 (0.35) 0.39*** (0.00) -0.25 (0.15) -0.49** (0.01) Loan loss provision -0.65*** (0.00) -1.04*** (0.00) -1.07*** (0.00) 0.03*** (0.00) 0.02** (0.03) 0.02** (0.04) Earnings/Assets 1.91 (0.84) 23.7** (0.04) 25.31** (0.03) -1.75*** (0.00) -0.34 (0.43) -0.05 (0.90) Liquidity Coverage ratio -0.01 (0.88) -0.01 (0.93) -0.01 (0.95) 0.00** (0.01) 0.00*** (0.00) 0.00* (0.09) Leverage Ratio -3.53** (0.01) -3.50** (0.01) -0.71*** (0.00) -0.89*** (0.00) Ownership concentration 21.8 (0.11) -1.92*** (0.00) Observations 555 552 552 1165 564 550 Adj. R2 _0.04 _0.06 _0.07 _0.04 _0.27 _0.33

Note: This table shows the outcome of six regressions. Regression (1) to (3) show the relationship between

equity volatility and the independent variables. All these regressions are on the year 2015, given data availability constraints. Regression (4) is on 2 years, hence the larger amount of observations.

(27)

27

In regressions (4) to (6) in table 5, we use Tobin’s Q as an alternative dependent variable, because relatively low bank value is generally associated with higher risk-taking, since owners have less to lose when bank value is relatively low (Konishi and Yasuda, 2004). The result for the liquidity coverage ratio still holds but the leverage ratio seems to have a significant negative relation with Tobin’s Q. Our results are thus not robust for Tobin’s Q. A 1% increase in the leverage ratio leads to approximately a 0.7% decrease in Tobin’s Q, where we would have expected an increase, since higher bank value is associated with lower risk (Keeley, 1990).

Finally, the result for ownership concentration remains consistent with all variables, adding value to the findings of Laeven and Levine (2009) and Chalermchatvichien et al. (2014).

4.3 Implications of these findings

The main findings of our study are:

 The LCR seems to have no effect on the z-score, the equity volatility and on Tobin’s Q;

 The LR has a statistically significant positive relationship with the z-score, a negative relation with the equity volatility and a statistically significant negative relation with Tobin’s Q;

 Ownership concentration seems to be negatively related to the z-score, positively related to equity volatility and negatively related to Tobin’s Q.

4.3.1 The liquidity coverage ratio – why does there seem to be no impact on risk?

Increased uncertainty during the credit crunch led to several bankruns, which in turn led to liquidity problems for a large number of banks. Basel III contains two liquidity ratios, the NSFR and the LCR, which are aimed at making sure that such liquidity issues do not arise once more in the future (Blundell-Wignall & Atkinson (2010)).

(28)

28

ECB already does ‘whatever it takes’ to stimulate lending after the crisis several liquidity support programmes for banks.

What has the ECB done after the crisis in order to improve bank liquidity? Firstly, ever since November 2008, the ECB has maintained a policy of ‘full allotment’. According to Benoit Coeuré, member of the Executive Board of the ECB, this policy is defined as ‘unlimited liquidity to banks at a predictable cost against an expanded set of eligible collateral’ (ECB, 2013). In practice this means that ever since the onset of the financial crisis, banks have had access to unlimited liquidity against barely any collateral (Financial Times, 2011). In a sense, banks in the Eurozone are thus infinitely liquid.

In addition, the ECB started large liquidity support programmes for banks during the aftermath of the credit crunch. The aim of these programmes was to increase bank liquidity and to decrease the rates on sovereign debt during the banking crisis in the eurozone (ECB, 2017). These programmes were called LTROs (Long-term refinancing operations) at first. Essentially these LTROs were loans from the ECB to banks against a reduced interest rate (ECB, 2017). Banks could use these loans to invest in sovereign debt (against a higher interest rate – making profit from the difference). These loans, from the ECB to banks, could be paid back in three years (ECB, 2017).

Its successor, from 2014 onwards is called TLTRO (Targeted Long-Term Refinancing Operations). TLTROs are essentially LTROs targeted at businesses and households (with the exception of mortgages). They are largely similar to LTRO’s, with a couple of differences. The most important difference with LTROs is that the aim of TLTROs is to boost the amount of lending from banks to the real economy instead of to governments (ECB, 2017).

The data in our sample is from 2014 and 2015. In these years, Eurozone banks borrowed a total of 329 billion euros against very cheap rates from the ECB under the TLTRO programme (ECB, 2017). While the ability for banks to obtain new loans under the LTRO programme had terminated in 2013, these already issued loans had a duration of three years. It is therefore highly likely that both the LTRO and TLTRO programmes have affected the liquidity levels of German banks during our sample period.

(29)

29

other hand providing unlimited liquidity looks quite contradictory. Imposing a minimum on liquidity levels is not likely to make sense when banks are already highly liquid. Our results seem to be in line with criticism of Blundell-Wignall & Atkinson (2010), who argue that the LCR looks like an unnecessary addition to the Basel Accord, because the Eurozone already has a highly active central bank.

However, it remains unclear whether the LCR has an impact in the absence of such a highly active central bank. Hence, it would be premature to conclude that the LCR altogether is a superfluous addition to Basel III. Therefore, a suggestion for future research would be to check whether the LCR still has no impact after the ECB stopped its unconventional policy of providing unlimited liquidity.

4.3.2 The impact of the leverage ratio

Our results indicate that the leverage ratio has a statistically significant negative impact on the z-score and an equally significant positive impact on equity volatility. These results point out that the leverage ratio may indeed decrease risk-taking among banks in Germany.

As was discussed earlier (in the literature review, section 2), there are three strands of literature on capital requirements and Basel III. The first strand of literature finds that Basel III reduces risk-taking among banks (studies such as Chalermchatvichien et al., 2014 and Adesina & Mwamba, 2016). They conclude that Basel III should thus have a positive impact. The second strand of literature is highly skeptical of capital requirements and believes that they might increase the funding costs of banks which may have an adverse impact on the real economy (Cummings and Wright, 2016). The third strand argues that Basel III may not be sufficient in decreasing risk-taking (Blundell-Wignall & Atkinson, 2010 and Admati & Hellwig, 2013).

(30)

30

The question as to whether this impact of 0.21 is enough critically depends on how high we want the z-score to be. As was mentioned earlier, the z-score indicates the (inverted) probability of bankruptcy for a bank. Masoom (2013) provides an overview of z-scores and relates them to the likelihood of bankruptcy. He states that a z-score of below 1.8 means that bankruptcy is likely, whereas bankruptcy is unlikely with a z-score of 3. The area between 1.8 and 3, Masoom (2013) argues, is ‘a grey area where a high degree of caution is recommended’. It seems logical to assume that, on average, it is desirable to minimise the probability that banks go bankrupt. After all, the credit crunch came at very high costs for the taxpayer.

However, the average z-score in our sample is 2.02. This is far below the threshold of 3.0 where bankruptcy becomes highly unlikely. The average leverage ratio among German banks in our sample period is 8%. Our results therefore indicate that if we want the average z-score to increase to 3.0, this would require an increase in the leverage ratio of (3-2.02)/0.07 = 14%.

Our results thus imply that the leverage ratio should increase to 8 + 14 = 22%. Such a leverage ratio may seem high, but our results relate closely to the previously mentioned letter sent to the Financial Times in 2010 by several prominent economists. In that letter, the economists propose a leverage ratio of at least 15% (Financial Times, 2010). Admati & Hellwig (2013) even argue that the leverage ratio should be increased to 20 – 30%. Note that our study purely looks at minimizing the probability of bankruptcy and does not propose that banks should not engage in any risk-taking at all. Such a proposal would be folly. After all, lending money to another party inherently comes with some risk that the borrower will not pay back the loan.

Nevertheless, it should be noted that our results also show that the impact of the leverage ratio on bank value seems to be negative. A 1% increase in the leverage ratio according to our results leads to a 0.7% decrease in Tobin’s Q. This may not seem economically significant on a sample average of Tobin’s Q of 88, but if we were to increase the leverage ratio by 14%, this implies a quite significant reduction in bank value of (14*0.7) 9.8%.

(31)

31

because equity is always relatively more expensive compared to debt since shareholders require a higher risk compensation than debtholders (Cummings & Wright, 2013). As was mentioned in the literature review, such a statement is not in line with standard finance theory (Modigliani & Miller, 1958). Total funding costs should not increase if a larger part of the firm is financed by equity, because the demanded return on equity is not static but depends on the debt/equity ratio (Modigliani & Miller, 1958).

Tax regimes can distort this mechanism that is sometimes called the ‘irrelevance of capital structure’. As was mentioned earlier, deposit holders demand a lower interest rate than they would normally do as a result of government policies such as deposit insurance and bailouts for debt (Admati & Hellwig, 2013). After all, deposit guarantees and the fact that banks are often bailed out when they are likely to default, create certainty that deposit holders will not lose their money. Banks can thus borrow for a premium at the expense of

the taxpayer, which makes debt more attractive to them than equity (Admati & Hellwig,

2013).

In the presence of a taxpayer subsidy regime, total funding costs will thus likely increase if the leverage ratio is increased, because banks can benefit less from this taxpayer subsidy. In Germany several banks were bailed out during the crisis (Der Spiegel, 2008). In addition, there is a deposit insurance scheme; ever since 2008, there is a 100% state guarantee on deposits (Bundesbank, 2015). There thus seems to be a taxpayer subsidy regime in Germany.

The fact that banks will benefit less from this regime as a result of a higher leverage ratio is, in our view, likely the reason why this study finds a negative relationship between the leverage ratio and bank value.

This study argues that a regime with taxpayer subsidy is not sustainable in the long run because banks cannot be allowed to take on more and more leverage, and hence more and more risk, at the expense of taxpayers. In fact, it is argued that this mechanism creates an incredibly unstable financial system (Admati & Hellwig, 2013).

(32)

32

ratio and both the z-score and equity volatility. An argument could thus be made that if the leverage ratio would be increased, the resulting decrease in bank value would not likely cause an increase in degree of risk-taking in banks. After all, the decrease in bank value would be caused by the increase in the leverage ratio, which is shown to already have a negative impact on two other parameters of risk-taking.

Another often-heard argument against increasing the leverage ratio is that it would likely reduce the amount of lending by banks with adverse effects for economic growth (Cummings & Wright, 2013). However, when the leverage ratio is left at its current level, the probability of a crisis such as the credit crunch occurring in the future would remain substantial. This is illustrated by the fact that, in the current regime, a reduction of 3% in the value of total assets would in principle be enough to bankrupt a bank (Admati & Hellwig, 2013). The credit crunch led to high costs for taxpayers around the world (BBC, 2009). The argument that the leverage ratio should not be increased because it would slow economic growth thus seems to be invalid. Keeping the leverage ratio at the current level seems much riskier for the economy than increasing the leverage ratio such that the probability of another credit crunch is minimized.

Therefore, this study recommends policymakers that the leverage ratio should increase to around 22%. Such an increase likely reduces the probability of banks going bankrupt to a minimum. The increase may come at the expense of bank value, but this reduction is likely justified by a decrease in the taxpayer’s subsidy and unlikely to increase the degree of risk-taking in the way suggested by Keeley (1990). The increased leverage ratio is also likely to come at the expense of economic growth, but the risk of keeping the leverage ratio at the current level is potentially much higher.

4.3.3 Ownership concentration and risk-taking

(33)

33

managers) and that large owners have the power to influence managers to engage in risk-taking behaviour.

(34)

34

V. Conclusion

The main aim of this study is to find whether Basel III’s liquidity coverage and leverage ratios have an impact on the degree of risk-taking among German banks in the years 2014 and 2015.

5.1 Summary of the main findings

From our results, it seems that the LCR has no impact on the degree of risk-taking. This is likely due to recent policies of the ECB. Ever since November 2008, banks can apply for unlimited liquidity support against basically any collateral. In addition, by means of two liquidity programmes (LTROs and TLRTOs) the ECB has tried to improve bank liquidity in the hope that banks would increase their lending to the public in order to boost the economy.

A minimum requirement on liquidity, such as the LCR, in a highly liquid market is not likely to have any impact. Indeed, the Romans would call the introduction of the LCR ‘Fundere aquas in mare’, Latin for: ‘(as useful as) carrying water to the sea’. Our findings on the LCR are in line with Blundell-Wignal & Atkinson (2010), who argue that any minimum liquidity requirement is unnecessary in the presence of a central bank that already guards the stability in the market.

However, there seems to be an impact of the leverage ratio on the degree of risk-taking. This impact looks fairly low: an increase in the leverage ratio of 1% leads to a 0.07 increase in the z-score, while the average z-score in Germany is 2.02. When one takes into account the fact that a safe z-score (where bankruptcy is highly unlikely) is considered to be 3.0, our result seems to suggest that the leverage ratio needs to increase significantly (Masoom, 2013). The average leverage ratio in the banks in our sample is 8%. This study argues that the leverage ratio should be increased to 22% (an increase of (3-2.02)/0.07 = 14% higher than the current level) in order for the German banking sector to reach an average z-score of 3.0.

(35)

35

in 2010 in a letter sent to the Financial Times that the leverage ratio should at least be 15% to ‘minimise the probability of another systemic crisis’ (Financial Times, 2010).

This study also investigates the relationship between bank value and both ratios. Our results suggest that there seems to be no economically significant relationship between Tobin’s Q and the LCR. However, when the LR would be increased to 22%, bank value would drop by about 10%. This drop in value is likely due to an increase in banks’ funding costs. An often-heard explanation for an increase in funding costs resulting from capital requirements is simply that shareholders always demand a higher return than lenders because of risk considerations (Cummings & Wright, 2013 and Admati & Hellwig, 2013). Increasing the amount of equity would thus inherently lead to higher funding costs.

However, this argument is not in line with standard finance theory, because the cost of funding should depend on total firm risk. It should be indifferent from capital structure, in the absence of taxes (Modigliani & Miller, 1958).

A more likely explanation for the increase in funding costs, and the resulting drop in bank value could be the decrease in the so-called taxpayer’s subsidy (Admati & Hellwig, 2013). German taxpayers indirectly subsidise debt-financing, as a result of government policies such as deposit insurance and bailouts. These policies have a negative effect on the interest rate that deposit holders demand. If banks were to hold on to more equity, the bank’s benefit from this subsidy would decrease.

This study argues that such a regime creates an unfair imbalance in the financial system where taxpayers seem to pay for the risks that banks take. The drop in bank value is likely justified by the decrease in the taxpayer’s subsidy.

An argument against increasing in the leverage ratio would be that, as such an increase seems to decreases bank value, the incentive for banks to engage in risk-taking behaviour would be increased. Konishi & Yasuda (2004) and Keeley (1990), for example, find that a decrease in bank value can increase bank risk-taking, because when a bank decreases in value, owners feel as though they have ‘less to lose’, which will make them less risk-averse.

Our results seem to indicate that a decrease in bank value as a result of an increase

(36)

36

leverage ratio in itself seems to have a detrimental impact on the degree of risk-taking already, as measured by the z-score and equity volatility.

Another argument against increasing the Basel III leverage ratio would be that such an increase may come at the expense of economic growth (Cummings & Wright, 2013). However, keeping the leverage ratio at the current level increases the probability of another financial crisis, as banks can very quickly lose all their equity – after all, their total asset value only has to decrease by 3% (Admati & Hellwig, 2013). In our view, keeping the leverage ratio at its current level could therefore potentially be more costly than raising it, as the risks to the financial system would remain very high.

Finally, this study uses a dummy variable for ownership concentration as a control variable, as is recommended by Laeven and Levine (2009), who state that ‘any analysis of bank risk-taking without using ownership concentration should not be taken seriously.’ Indeed, we find that ownership concentration is strongly associated with bank risk-taking. A policy recommendation of this finding may be to address the issue of ownership concentration in legislation that aims to reduce risk-taking in banks. Discussing specifically what kind of provisions on ownership concentration ought be included in such legislation goes beyond the scope of this paper and is fruit for future research.

5.2 Limitations

There are several limitations to this study that should be taken into account.

The first limitation is that the study only contains data of two years. Data on the LR and on ownership concentration is only from 2015, whereas data on the LCR is limited to the 2 years 2014 and 2015. With data available over multiple years, it would be possible to adjust the results for time effects, which could render a more accurate estimation.

(37)

37

study, especially our result on the leverage ratio. After all, a relatively highly leveraged banking sector may react more extremely to a cap that essentially maximises leverage than relatively less leveraged banking sectors.

A third limitation of this study is a limitation that should be taken into account with any study that investigates the economic impact of a certain policy: the Lucas critique (1976). This study suggests a gradual increase of the leverage ratio to 22% in order to increase the average z-score of German banks to 3 such that the probability of default decreases to practically 0. A question that arises is whether the Lucas (1976) critique would be applicable to such a recommendation. According to Lucas (1976), it is naïve to predict the impact of economic policies on the basis of historical data.

A real-life example of the Lucas critique is the practical application of the Philips curve. In the 1970s a negative relationship (this relationship is known as the Philips curve) was found between unemployment and inflation. As a result of this finding, governments started to increase inflation in order to lower the unemployment rate. However, firms started to adjust their price expectations to the higher inflation, which caused unemployment to increase again. The relationship between unemployment and inflation changed because of changed government policies (Lucas, 1976). It turned out that, in the words of Lucas (1976), governments were indeed naïve to predict the impact of increasing inflation on the unemployment rate on the basis of historical data.

(38)

38

5.2 Future research

Aside from the impact of an increase in the leverage ratio on the degree of asset substitution, there are various other questions that remain unanswered after this study. For instance, this study finds no impact of the LCR on risk-taking because of policies by the ECB. The ECB, as was mentioned before pursues a policy of boosting bank liquidity. However, the Federal Reserve, in the United States (where the LCR is implemented as part of the Dodd-Frank Act), has pursued different policies. It would be interesting to see whether, under a different central bank policy than in Europe, the LCR would have an impact on bank risk-taking.

Another opportunity for future research would be to study the impact of the net stable funding rate (NSFR) on risk-taking. As was mentioned earlier, the NSFR is another new liquidity ratio that was introduced in Basel III. It could be valuable to see if the NSFR has the same impact as the LCR. The NSFR was not implemented during our sample period, which is why we did not include it.

The impact of the leverage and liquidity coverage ratios from Basel III on bank risk-taking in Germany

Master’s thesis for the Msc Finance