The unconventional monetary policy of the European Central Bank and its impact on European bank shares

The unconventional monetary policy of the European Central Bank

and its impact on European bank shares

Ricardo M. Wagner, 11376813

July 1, 2017

University of Amsterdam, Amsterdam Business School

MSc Finance, Corporate Finance Track

Master Thesis

II

Statement of Originality

This document is written by student Ricardo Wagner who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

III

Abstract

In January 2015 the European Central Bank (ECB) announced the launch of a new unconventional monetary policy measure. Under a large-scale asset purchase programme, commonly known as quantitative easing (QE), the ECB started to purchase public and later on private sector securities. The objective of this programme was to lift up the inflation rate and to foster economic recovery in the Euro area. Applying an event study approach and a cross sectional analysis this research investigates the implications of the QE programme on banks of the Euro area. The findings show that overall Euro area banks benefited from QE. In addition the findings show differences in the magnitude of share price reactions between banks from southern and northern Europe. The cross sectional analysis confirms the previous findings and provides evidence that risk and performance indicators are influencing the share price reaction. In total, the results of this research show that all banks benefited from QE, although banks from southern Europe more compared to their northern counterparts.

1

Table of Content

2

List of Abbreviations

ECB European Central Bank

QE Quantitative Easing

BoJ Bank of Japan

JPY Japanese Yen

Fed Federal Reserve System

BoE Bank of England

USD US Dollar

GBP Great Britain Pound

GDP Gross Domestic Product

SMP Securities Markets Programme

LTRO Longer-term Refinancing Operations

APP Asset Purchase Programme

CBPP Covered Bond Purchase Programme

OMT Outright Monetary Transactions

ABSPP Asset-Backed Securities Purchase Programme

3

1. Introduction

„The primary objective of the ECB’s monetary policy is to maintain price stability. This is the best contribution monetary policy can make to economic growth.“1

Triggered by the European debt crisis during the past years Europe has been facing times of declining economic growth, investment bottleneck and the fear of an uprising deflation. In times like this, ordinary monetary policy instruments, like adjusting the short interest rate, have only limited effect on reaching the objective of the European Central bank (ECB). This limited effectiveness gets even stronger when the short-term interest rate reaches the lower bound. Keynes (1936) was among the first economists to worry about this problem. Bernanke et al. (2004) argue that the short-term interest rate cannot go below zero. This is mainly based on the fact that people can hold cash instead of keeping their money on their bank account. Limited by the zero lower bound major central banks around the world started looking for new, unconventional monetary policy measures they could use to reach their objectives of fostering economic growth and thereby to overcome the current crisis. Quantitative easing (QE) is probably the best-known one among these unconventional measures. Until today QE is used by the majority of the most important central banks. The Bank of Japan (BoJ) was the first of the major central banks to use QE in 2001. Effectively two operations have been executed. On the one hand the BoJ increased its monthly purchase of long-term Japanese government bonds till JPY 1.2 trillion between 2001 and 2004. On the other hand the central bank kept the current account balance level of commercial banks higher than the required reserves (Kobayashi et al., 2006). The next central banks that applied QE were the central banks of the United States (Fed) and the United Kingdom (BoE). The American QE programme, which was announced as a large-scale asset purchase programme, was mainly focused on mortgage backed securities and treasuries. It had a total volume of over USD 3 trillion and ran between 2008 and 2014. The QE programme of the BoE took place between 2009 and 2012 and had a total volume of GBP 375 billion. The majority of this volume was used to purchase British gilts2 (Fawley and Neely, 2013).

1 _{https://www.ecb.europa.eu/mopo/intro/html/index.en.html} 2 _{http://www.dmo.gov.uk/index.aspx?page=gilts/about_gilts}

4

Compared to the strategy of the Fed and BoE, the ECB started its QE programme relatively late, in March 2015. Before starting the implementation of a large-scale asset purchase programme the ECB used various smaller programmes (LTRO, SMP). The intention behind these programmes was to ensure money supply in the inter banking market and to ease the financial and funding risk with a special focus towards the southern European countries. With the start of the QE programme, which replaced most of the small previous programmes, the ECB aligned is strategic approach towards the American / British one (Fawley and Neely, 2013).

Every prosperous economy is based on a functional financial system. Banks as well as capital markets can offer essential services like providing funds for investment and consumption. Clayes (2016) argues that these two complementary sources of funding need to be balanced out in order to guarantee a functional financial system. Nevertheless, the exact proportion is neglectable. This balance holds for the US and British financial systems, although with a more important role for the financial markets. However, the European financial system is too bank-centric with an underdeveloped financial market. This means that financing is rather be done via bank lending than financial market related operations like issuing bonds (Fawley and Neely, 2013, Claeys, 2016). In addition to its deep entrenchment in the European financial system, the private banking system plays another important role. Besides the national central banking system, the private banking system is the mechanism whereby the ECB can influence the economy. The asset purchases under the current and previous programmes all took place in the secondary market, with the national central banks and private banks as intermediaries.

This research uses an event study approach to investigate the impact of the QE announcements on European banks shares. More precisely, separate event studies for the most important Euro area states will be conducted. This is due to the fact that although the Euro area has a common currency it is far from being an integrated economic area. Afterwards, a cross sectional regression analysis is used to look if the observed abnormal returns can be explained by specific bank characteristics.

The most recent studies of the ECB’s unconventional monetary policy have been focused on smaller programmes that took place before January 2015 and their impact on macroeconomic variables. The tenor of these literatures confirms a positive influence of unconventional monetary policy on credit and monetary variables as well as on certain parts of the financial market.

5

This research tries to extend the existing literature in two ways. First, it focuses upon the expanded asset purchase programme (APP), which was announced in January 2015 and is therefore the most recent one. Furthermore, the APP can be considered as a true QE programme. It differs significantly from the previous monetary policy measures in terms of size, purchasable assets and objectives. Second, investigating the share price reaction of European banks, this research is trying to capture the perceived perception of the QE programmes by investors and the overall financial market. Based on the exposed role of banks in the European financial system as well as the whole economy this research believes the share price reaction of banks is an indicator for the impact of the APP in different European countries. Next to that, the past revealed that European banks suffered exceptionally during the current crisis.

There are some common problems, which occur while assessing the effectiveness of unconventional monetary policy measures. Joyce et al. (2011 B) note that unconventional monetary policy is lacking a proven theoretical framework, which is due to the fact that this policy is a practical response to the current crisis. Furthermore, the various unconventional measures have individual primary objectives, although they have a common secondary goal. The APP from January 2015 aims at lowering the interest rates and lifting asset prices to stimulate the real economy. In contrast, the LTRO (2011) was introduced to mitigate the liquidity and funding risk in the European banking sector (Darracq-Paries and De Santis 2015). These different objectives can be analysed, but it is difficult to compare them to each other and identify their contribution to the economic recovery. Another problem is that it is unknown how the economy would have developed without any monetary policy intervention. So a large fraction of research is based on a counterfactual approach. Finally, it is important to remember that the Eurozone is not a single integrated economic area. There are independent governments, which follow their own strategic agenda to ensure the best possible economical result for their own country.

The reminder of this thesis is structured as follows. Chapter 2 provides theoretical background about unconventional monetary policy. In chapter 3 the hypotheses are derived, while chapter 4 explains the methodology of this research .The empirical results are displayed in chapter 5. Finally, chapter 6 concludes and provides a brief discussion about the findings.

6

2. Literature Review

The literature review chapter of this thesis is structured as follows. First, a brief description about the conventional monetary policy and its limitations is provided. Next, an overview about unconventional monetary policy with a clear focus on QE is presented. Subsequently the potential transmission channels of QE are shown. The chapter concludes with giving examples about unconventional monetary policy in Japan, United Kingdom, USA and Europe.

2.1 Conventional monetary policy and its limitations

Current literature states that in ordinary times central banks execute their monetary policy by controlling the short-term nominal interest rate. This is done by buying and selling short-term debt securities (Fawley and Neeley, 2013, Kozicki et al., 2011). This is in line with Bernanke and Reinhart (2004), who mention that the main part of conventional monetary policy is to determine prices and yields of financial assets. These kinds of operations allow central banks to influence the short-term nominal interest rate as well as the monetary base. Based on the delayed incorporation of inflation expectations, central banks are even able to influence the short-term real interest rates. Real interest rates, in turn, determine asset prices and thereby affect investment and consumption decisions of companies as well as of individuals.

Joyce et al. (2011 B) identify two important limitations of this conventional monetary policy. The first problem is linked to short-term nominal interest rate at the zero lower bound. With a rate close to zero, central banks do not have any longer the possibility to lower the interest rate because the nominal rate cannot go below zero. This is mainly based on the fact that people can hold cash, which pays a nominal interest rate of zero instead of keeping their money on their bank account (Fawley and Neeley, 2013, Bernanke and Reinhart, 2004). First concerns about this obstacle of conventional monetary policy can be backdated to Keynes (1936).

The second problem is a dysfunctional financial system/market. In times of a crisis the functionality of the interbank market as well as the whole credit market cannot be guaranteed. The solvency of financial institutions (interbank market) and their customers is unsafe. Consequently, the relationship between nominal and real interest rates and does not sustain. Furthermore, banks could hold back private or industrial sector lending, which could stimulate the economy, in order to minimize their risk exposure. So, with a dysfunctional financial system and nominal interest rates close to zero central banks lose their ability to affect the economic

7

development in an effective way (Joyce et al., 2011 B). This implies an increasing probability for a prolonged period of declining economic growth or even stagnation, investment bottleneck and in the worst scenario even a deflation.

2.2 Unconventional monetary policy

To avoid the mentioned threats central banks, need to apply unconventional monetary policy measures. Bernanke and Reinhart (2004) provide three different types of monetary policy when the nominal short-term interest rate is close to zero. First, central banks can try to influence public expectations about future interest rates. Therefore, central banks must convey credibly that in the long run the short-term rate will remain under the current market expectations. Second, central banks can recompose the structure of their balance sheet and thereby change the relative supply and demand of certain securities available in the market. Lastly, central banks can simply extend the size of their balance beyond the amount that is needed to maintain the short-term interest rate close to zero. Such an expansion of the balance sheet, which aims at shaping quantities instead of prices, is commonly known as quantitative easing or QE (Bernanke and Reinhart, 2004, Falagiarda and Reitz, 2015).

A huge body of literature like Bernanke et al. (2004), Joyce et al. (2011 B) and Krishnamurthy and Vissing-Jorgensen (2011) investigate potential transmission channels of quantitative easing. Their findings highlight three main channels – the portfolio rebalancing channel, the signalling channel and the market function/ liquidity premia channel.

The portfolio rebalancing channel is probably the most researched transmission channel of quantitative easing. The theoretical background of this channel is about imperfect asset substitutability, portfolio substitution effects and investor behaviour (Tobin, 1969). Large-scale asset purchases of long term and relatively riskless assets by central banks lead to increasing asset prices, decreasing yields as well as decreasing market supply of the purchased assets. The excess money, which investors get for selling their assets to the central banks, needs to be reinvested. Thus, investors seek to adjust their portfolios and thereby raising prices and reducing yields on alternative relatively more risky assets. With overall declining yields on long term assets and at the same time rising prices, financing becomes cheaper, which in turn will help to stimulate economic growth (Bernanke and Reinhart 2004, Beirne et al., 2011 and Joyce et al. 2011).

8

Christensen and Gillan (2013) argue that the signalling channel is the most straight forward transmission channel, as quantitative easing operations act as a signal and thereby change the market expectations about the future path of monetary policy. As mentioned by Bernanke and Reinhart (2004) unconventional monetary policy can also include sending trustworthy signals about the future path for monetary policy. When central banks announce their upcoming guidance for the monetary policy, the market reacts immediately to this signal. Direct market operations by central banks, like QE, are an even more credible commitment for the future policy. Expanding their balance sheet by purchasing long-term duration assets, central banks can send a strong signal towards the market. Mainly due the fact that when central banks raise short-term interest rates, while at the same time holding a large amount of long duration assets, they will realize large losses on their investments (Clouse et al. 2003, Beirne et al., 2011).

The market function/liquidity premia channel refers to observable effect on decreasing liquidity premia after the market entry of central banks. In times of a crisis markets for certain assets can dry out. This means that these markets suffer from low liquidity and corresponding high liquidity risk premia. Under quantitative easing policy central banks can participate in such markets. This leads to higher liquidity and lower liquidity risk premia or in short, improved market functionality. A reason for this is, that central banks are considered as buyers with deep pockets. Therefore, investors are more likely to participate, as they know that they can sell to the central bank in case of divesture. Potentially, this effect can spill over to other assets classes through investors, which are seeking assets with higher yields compared to the ones purchased by the central banks. Subsequently, rising asset prices increase the wealth of their owners and their ability to invest or consume (Falagiarda and Reitz, 2015, Christensen and Gillan, 2013 and Beirne et al., 2011). Worth noticing is that the described effect can be considered temporary as it is depending on the acting of the central bank (Joyce et al., 2011 B).

9

2.3 Examples of quantitative easing

In the early 2000s, Japan was suffering from a busted bubble, deflationary tendencies and lethargic economic growth. To overcome this situation the BoJ used unconventional monetary policy measures. These measures have been summarized under the phrase quantitative easing (Joyce et al., 2011 b). The outcome of this policy has been assessed in various researches. Ugai (2007) provides a survey about the empirical analyses of the Japanese QE programme. He states that the market has been mostly affected by commitment of the Bank of Japan to keep the short-term interest low in the long run (Signalling channel). Next to that, he presents only weak results in favour of a significant impact of neither the expansion nor an altered composition of central bank’s balance sheet. The author concludes that, although the current research possibly not being able to clearly identify the transmission channel and their relative importance, QE had a positive impact on the financial environment and thereby even on the Japanese economy. Girardin and Moussa (2011) support the findings. By applying a Markov-switching VAR approach they show that the QE programme prevented the economy from further recession and deflations threat. Furthermore, the output of the Japanese economy was also stimulated, mainly based on the commitment to keep the interest rates low. Bernanke, Reinhart and Sack (2004) conducted another extensive study about quantitative easing and its economic impact in Japan. Their event study results provide only limited evidence of the success of quantitative easing. They use an estimation model to predict the future long-term yields and show that long-term yields might decrease more than expected during the quantitative easing period. Kobayashi et al. (2006) conducts another event study, which is focussed on the reaction of Japanese bank. The results of this study show positive excess returns for Japanese banks shares on QE announcements, which are linked to a flatter future yield curve. Worth noticing is that financially weak banks experienced a higher abnormal return.

In the aftermath of the financial crisis 2007 the BoE used unconventional monetary policy measures to stabilize the financial sector and prevent the British economy from a recession. The QE programme started in 2009 and preliminary ended 2012, with a total amount of bought asset of £ 375 billion equal to 24% of the annual British GDP. The majority of the assets that have been bought under QE were medium and long-term government bonds (Claeys, 2014). Kapetanios et al. (2012) use a multiple counter factual approach to investigate the impact of the British QE programme on the broad economy. Their findings suggest that the intervention of the

10

BoE mitigated the fall of the British GDP and prevented the inflation level to reach the lower bound. More specific results can be found by Joyce et al. (2011 B), which give an overview on the reaction of different asset classes to QE. Although, the authors mention considerable uncertainty about the precision of asset price reaction, based on potential lagged reactions or unknown influential factors, the results of their event study indicate a decrease of 100 basis points for British government bonds. A similar, though weaker reaction can be observed for corporate bonds, while equity prices did not react in a clear way to QE. Furthermore, the authors report an increased bond and equity issuing, which is a proxy for rising financing and investment activity. Hence, the results of this research suggest that QE had a positive impact on the British economy (Joyce et al.,2011 B).

From December 2008 until October 2014 the FED conducted three consecutive QE programmes. The total amount of bought assets under the large-scale asset purchase programme is about USD 3.5 trillion or round 21% of the annual American GDP (Claeys, 2014). The outcome of these programmes has been investigated various times. For instance, Krishnamurthy and Vissing-Jorgensen (2011) use multiple event studies to analyse the reaction of a bunch of interest rates on the QE policy. They observe a larger and significant drop in nominal interest rates for treasuries, mortgage-backed securities and high rated corporate bonds. Besides, the authors mention only small spill over effects for lower rated corporate bonds. Finally, evidence is provided that the rates of securities which are bought under the QE programme, are affected the most. In another research Gangno et al. (2011) also want to quantify the empirical implications of QE by using an event study approach and a time series analysis. They confirm the findings by Krishnamurthy and Vissing-Jorgensen (2011) and reveal that the QE programme had significant and long lasting influence on various types of securities, e.g. corporate bonds, mortgaged backs securities and treasuries. Their estimates suggest a yield drop between 30 and 100 basis points depending on the type of security. Additional findings show that beneath corporate financing also private financing has become cheaper resulting in a stimulus for the economy. Fratzscher et al. (2012) find evidence for reduced bond yields and rising equity markets, which are mainly explained by the portfolio-rebalancing channel. The importance of this channel is underpinned by the fact that the executed operations of the central bank had more influence compared to their initial announcements. It contrasts to the work of Bernanke et al. (2004), who were only able to quantify the outcomes of the signalling channel. They use an event study approach to quantify

11

the importance of shaping the market expectations on the future interest rate policy and thereby successfully reducing long-term interest yields. However, the authors were not able to provide significant evidence for other potential transmission channels of unconventional monetary policy measures. This could be based on the available data at the time when the study was conducted. Regardless, in another research Bernanke and Reinhart (2004) propose that lower yields, caused by unconventional monetary policies, lead to more economic growth due to more credit lending. This is in line with the Chung et al. (2012) who project an increase of the American GDP by around 2% based of the first American QE programme. This increase is mainly based on low long-term interest rates and higher stock prices.

In line with its foreign counterparts the ECB used unconventional monetary policy measures to overcome the recent crisis. Over the past years, the financial crisis evolved into a sovereign debt crisis, which in turn affected the entire European banking sector. In contrast to the programmes of the FED and BoE the ECB started relatively late, in spring 2015, with its large-scale asset purchase programme. Before that, the released programmes were mainly targeting to ease the tense situation of the southern European states on the financial market. Consequently, the existing literature is mostly focused on the impact of other unconventional monetary programmes that have been conducted between 2008 and 2015 3.

Falagiarda and Reitz (2015) use an event study to determine the reaction of European sovereign bond spreads on several unconventional monetary policy announcements of the ECB between 2008 and 2012. More precisely, they investigate spread movements of so-called PIIGS states bonds compared the German bonds. Their results indicate that the announcement alone could reduce the spreads for nearly all stressed states and so the perceived risk of investors. In addition, the findings highlight the spread reaction varies in magnitude and significance depending on the objective of the announced programme.

Darracq-Paries and De Santis (2015) study the impact of another unconventional programme, namely the ECB’s 3-year long-term refinancing operation (LTRO). The aim of this programme was to support the liquidity situation of European banks and thereby ensure that banks can grant more loans to economy. By using a vector autoregressive framework based on data from a special ad hoc questionnaire (February 2012) and from the euro area bank lending survey (April

3 _{See Appendix, table A for a comprehensive overview about the unconventional monetary programmes of the} European Central bank

12

2012) the authors try to estimate the influence of the LTRO. The findings suggest the main transmission channel of the LTRO is the liquidity premia or market function channel. Thus, the injected extra liquidity led to relaxed bank lending conditions. Consequently, the authors argue that the expansionary effect of the LTRO programme can stimulate the economy via increased loan volume to non-finance corporations and declining lending rates, which will finally result in an increasing gross domestic product and a rising price level.

In September 2012, the outright monetary transactions (OMT) programme replaced LTRO. Altavilla et al. (2014) study the influence of this programme, comparing the government bond yields of different euro area states. For the PIIGS states Spain and Italy they observe a significant drop of 200 basis points on their government yields, while Non-PIIGS states yields remained unchanged. Subsequently their counter factual analysis suggests a significant effect on the credit volume and economic growth in affected countries. The results of Altavilla et al. (2014) testify earlier findings by Eser and Schwaab (2013), which find that the yields of all five PIIGS states dropped due the announcement of the ECB’s Securities Market Programme (SMP).

Most of the studies about unconventional monetary policy measures, like QE, find a positive impact on the real economy. In detail, QE programmes lifted asset prices, lowered bond yields and improved the functionality of the financial system. Two transmission channels mainly, but not only, explain the observed results. First, the commitment of the central bank to keep interest rates low in the future (Signalling channel). Second, by influencing the composition of institutional investors’ portfolios, which lead to lower yields, rising asset prices and thereby to more attractive financing and investment opportunities (Portfolio Rebalancing channel).

Nevertheless, it is indispensable to note that the results of the mentioned literature should be interpreted with caution. Currently, a clear theoretical framework of unconventional monetary policy is missing. The term “unconventional monetary policy” itself is used as an umbrella term for several measures with all of them have individual objectives (Kozicki et al., 2011). In addition, there is little or only limited empirical evidence of the effect of unconventional monetary policy measures, more precisely the effect of QE. This is mainly based on the novelty of this type of policy. Bernanke and Reihart (2004) mention that it is difficult to attribute the overall results of unconventional monetary policy back to the single measures like QE or low interest rates. Next to that a bunch of researches itself (Joyce et al., 2011, Gagnon et al., 2010) mention that their results are exposed to a degree of uncertainty, which can be linked to a lack of

13

reliable and long term data or to methodologies that are mainly based on counter factual approaches.

3. Hypotheses

The objective of this thesis is to examine the effect of the ECB’s QE programme on European bank shares. Based on the literature review in chapter two, three hypotheses are derived in this chapter.

Hypothesis No. 1

“European bank shares exhibit a positive abnormal return after the QE announcements of the ECB.”

Literature like Krishnamurthy and Vissing-Jorgensen (2011), or Fratzscher et al. (2012) point out that QE positively influences real economy by using different transmission channels. More important, all these channels can be linked to business opportunities of banks, which underpin the strong position of banks in the European financial system (Claeys, 2016). Next to that, an improved functionality of the financial market should also affect banks positively. First support for this hypothesis comes from Kobayashi et al. (2006) who find a positive reaction of banks on the Japanese QE programme.

Hypothesis No. 2

“The second and third QE announcement led to lower abnormal returns compared to the initial one.”

Meaning and Zhu (2011) find significant differences in the market reactions for the QE announcements in UK and in US. They observe the greatest market response for the first QE announcement of each programme, while the following ones triggered significant weaker market reactions. A possible reason could be the fact that only the initial announcement can be considered as truly unexpected, while the financial market has already incorporated the impact of the subsequent ones. Therefore, it might be difficult to exceed the markets expectations and see a strong reaction. This is in line with Krishnamurthy and Vissing-Jorgensen (2011), who argue that asset price reaction depends on the change of expectations around QE announcements. Next to

14

that the authors show that market reaction is also determined by factors like eligible assets, objectives and volume of the announced programme.

Hypothesis No. 3

“Banks in a weak financial condition benefit more from the QE programme and therefor exhibit higher abnormal returns”4.

Esser and Schwaab (2013), Altavilla et al. (2014) and Falagiarda and Reitz (2015) show that in general PIIGS states benefit more from unconventional monetary policy. The importance of the PIIGS states is also highlighted by the fact that most of the previous ECB programmes were executed in order to support the southern European states. By the unique and close linkage of domestic banks with the domestic economy it seems reasonable that these banks will benefit more than banks from non-PIIGS states. Support for Hypothesis 3 is again coming from Kobayashi et al. (2006), who provides cross sectional evidence that financially weak banks generate higher abnormal returns compared to their counterparts.

4 _{This research uses six bank specific indicators in order to define the financial condition of a bank. Next to that a}

15

4. Methodology

This chapter describes the methodology and the data used in this research. First, the event study methodolgy is explained. Thereafter an overview about the cross-sectional regression approach and a description of the used variables is provided. Finally, this chapter concludes with descriptive statistics for the bank characteristics.

4.1 Event study

To measure the effect of the QE announcements on share prices of European bank shares an event study is used. This research conducts separate event study for all sample banks and two cross-country event studies. The latter ones divide the sample into PIIGS and Non-PIIGS banks. The theoretical framework of this method is based on the efficient market theory, which suggest that new information is immediately incorporated in the stock market (Fama, 1970).

The methodology of this research is based on MacKinlay (1997) and de Jong and Goeij (2011). The intention behind an event study is to compare realized returns with estimated normal returns over a defined time. The observed difference is the so-called abnormal return.

In the course of this research the market model is used to measure the normal return. This model links the return of a specific asset with the return of a market portfolio and is given by following equation (MacKinlay, 1997).

𝑅_!" = 𝛼_!+ 𝛽_!𝑅_!" + 𝜀_!"

E (𝜀!") = 0 Var (𝜀!") =𝜎!!!

With

𝑅_!" = Return of asset price 𝑖 at time 𝑡

𝑅_!" = Return of the market portfolio at time 𝑡 𝜀_!" = Zero mean disturbance term

𝛼_! = Intercept term

16

The estimated daily abnormal return of share 𝑖 is then derived

𝐴𝑅_!" = 𝑅_!"− (𝛼_!+ 𝛽_!𝑅_!")

The coefficients 𝛼!& 𝛽! are estimated trough an ordinary least squares regression of 𝑅!" on 𝑅!". An estimation window (𝑇) of 150 trading days ending 5 days before the event date is used to estimate the coefficients. By choosing the end of the estimation window 5 days before the beginning of the event window this research assures that the estimation window and event window do not overlap. If this would be the case then the estimate normal returns would reflect the influence of the event and thereby biasing the abnormal returns (MacKinlay, 1997).

First, a single day event window is used. The short time period is based on the assumption that newly available information is immediately incorporated. In addition, Brown and Warner (1985) mention that the power of the test statistic is decreasing with a wider event window.

Afterwards the event window is extended in order to check the robustness of the single day event window results. In that case, the individual 𝐴𝑅_!" of each bank is summed up over the event window from 𝑡 = 𝑡_! to 𝑡_! where 𝑡 = 0 defines the event day and 𝑡_! ≤ 0 ≤ 𝑡_!.

𝐶𝐴𝑅_! = 𝐴𝑅_!" !!

!!!!

As a robustness check and for the purpose of investigating potential differences between banks from PIIGS and Non-PIIGS the sample is divided into PIIGS and Non-PIIGS banks across all states. Moreover the Euro Stoxx 50 is used as the market index. The individual abnormal return is then unweighted aggregated over the banks in each sample.

𝐴𝐴𝑅_! = 1

𝑁 𝐴𝑅!" !

!!!

17

McWilliams and McWilliams (2011) present a commonly used test statistic in order to test if the abnormal returns are statistically significant different from zero 𝐻!: 𝐸 𝐴𝑅!" = 0 . The statistic, that is based on Dodd and Warner (1983) assumes that the abnormal returns are independent and identically distributed and uses standardized abnormal returns. The standardized abnormal return is computed by dividing the abnormal return through its adjusted standard deviation

𝑆𝐴𝑅_!" = 𝐴𝑅_!" 𝑆_!" ~ 𝑡 (𝑇 − 2) With 𝑆_!" = 𝑆_!!_{∗ 1 + 1 𝑇 (𝑅} !" − 𝑅!)! (𝑅!"− 𝑅!)! ! !!! !.!

Where 𝑆_!! _{is the variance of company 𝑖 ‘s regression residual of the market model,}

𝑇 is the length of the estimation window, 𝑅_!" is the market return on day t, while 𝑅_! is the average market return over the estimation window.

The test statistic for the standardized cumulative abnormal return 𝑆𝐶𝐴𝑅!" over a wider event window is defined as the following

𝑆𝐶𝐴𝑅_!" = 𝑆𝐴𝑅_!" ∗ 1 𝑡_!− 𝑡_!+ 1 !! !!!! ~ 𝑡 (𝑇 − 2) (Dodd and Warner, 1983).

18

The significance test of 𝐴𝐴𝑅! for the PIIGS and Non-PIIGS sample uses the following test statistics. 𝑇𝑆 =𝐴𝐴𝑅! 𝑠 ~ 𝑁 (0,1) With 𝑠 = 1 (𝑇 − 1) ∗ (𝐴𝑅𝑅_!− 𝐴𝑅∗₎! ! !!! !.! Where 𝐴𝑅∗ ₌ 1 𝑇 𝐴𝐴𝑅! ! !!!

This test statistic is known as crude dependence adjustment method and accounts for potential event clustering, which can lead to cross sectional dependence. The variance of the average abnormal returns is derived from the times series observations of the average abnormal returns in the estimation window (de Jong and de Goeij, 2011, Brown and Warner 1980).

4.2 Cross sectional analyses

After calculating abnormal returns a cross sectional analysis is used to examine up to what extent bank specific characters are influencing the observed excess returns. This research focusses on six bank specific indicators for a bank’s financial condition. According to the third hypothesis this research expects financially weak banks disproportionately benefit from QE and therefore generate a higher abnormal return.

“Banks in a weak financial condition will benefit more from the QE programme and thereby generate higher abnormal returns.”

19

𝐴𝑅!" = 𝛽!+ 𝛽!𝐿𝑇𝐶𝑅!"+ 𝛽!𝑁𝑃𝑇𝐿!"+ 𝛽!𝑇1𝐶𝑅!" + 𝛽!𝑅𝑂𝐴!"+ 𝛽!𝐷𝐼𝑉!"+ 𝛽!𝑇𝐴!" + 𝛽_!𝑃𝐼𝐼𝐺𝑆_!"+ 𝜀_!"

𝐿𝑇𝐶𝑅 = Long Term Credit Rating

𝑁𝑃𝑇𝐿 = Non Performing Loans to Total Loans Ratio 𝑇1𝐶𝑅 = Tier 1 Capital Ratio

𝑅𝑂𝐴 = Return on Assets 𝐷𝐼𝑉 = Dividend Payer 𝑇𝐴 = Total Assets

𝑃𝐼𝐼𝐺𝑆 = Indicator if a banks is based in a PIIGS state

Capital ratios have been used to estimate the financial condition of a bank since years (Estrella et al., 2000). Nevertheless, the Tier 1 Capital Ratio, which is expressed through T1CR, has probably become the best-known capital ratio in the recent past. While tier 1 capital is defined as common shareholders’ equity and qualified preferred stocks, less goodwill and other adjustments, the ratio itself is the tier 1 capital as percentage of total risk-weighted assets. As a high tier 1 capital ratio signals a sound financial condition of a bank, a negative relation between

T1CR and AR is expected.

The fraction of non-performing loans of a bank is a widely used measure for bank’s asset quality (Shrieves and Dahl 1992). In this research, the ratio of Non-Performing Loans to Total Loans (NPTL) is applied as an indicator for bank’s asset quality. One of the main tasks of banks is to provide loans for investment and consumption. If these loans become non-performing banks suffer losses, which decrease profitability and at the same time increase the financial risk. So, a high non-performing loans to total loans ratio is a sign that banks face serious problems regarding their asset quality. Subsequently, this research supposes a positive relationship between NPTL and AR.

One common way to measure profitability is to look at the Return on Assets (ROA) (Bourke, 1989, Molyneux and Thornton, 1992). According to Athanasoglou et al. (2008) return on assets is a key profitability measure for banks, as it takes the risk associated with high leverage into account. A high ROA can be linked to more profitable banks, which in turn are in a better financial condition. Based on that, a negative influence of ROA on AR is expected.

The logarithmic value of Total Assets (TA) is a measure for firm size and an indicator for profitability and financial risk (Boyd and Runkle 1993, Lepetit et al., 2008). As big banks tend to

20

be more profitable and less risky they should generate lower abnormal returns. Consequently, this research anticipates a negative relation of the TA and AR.

Dividend (DIV) is a dummy variable, which is 1 if the bank paid a dividend in the corresponding

fiscal year and 0 otherwise. Fama and French (2001) state that relatively more profitable firms are more likely to pay a dividend. Thus, a negative impact of DIV on AR is expected

The variable Long-Term Credit Rating (LTCR) represents the current long-term debt rating of a bank. The original (qualitative / non-numeric) rating is derived from Moody’s and converted into a numeric metric. In this metric, the highest investment grade Aaa equals 22, while the lowest grade, C, equals 2 and the threshold for investment/non-investment grade is Baa3 or 13. Credit ratings provide a signal for the risk of default of a specific security or for the whole bank. A higher value of the credit rating implies a lower financial risk. Hence, a negative relation between LTCR and AR is expected.

PIIGS is a dummy variable that is equal to 1 if a bank is base in a PIIGS country and 0

21

4.3 Data

This paragraph provides an overview about the data used in this research. The event dates are manually obtained from the ECB website, respective their press releases. The three identified events took place between January 2015 and December 2016. This period constitutes the framework for research. The hand-collected sample consists of public listed banks of the Euro area and is mainly constructed according to the capital key of the ECB. The capital key distribution is a proxy for a country’s share in European population and the GDP of the Euro area. Furthermore, the sample is constructed according to the availability and reliability of the needed data. The total sample includes 35 banks from 10 European states. A list of the sample banks is provided in the appendix (table B). Daily share and indices prices for the event studies are collected via DataStream. Accounting data which will be used in the cross-sectional analysis, is collected by Compustat-Capital IQ, while missing data is manually gathered through annual reports. The credit ratings are derived from Moody’s Investor Services.

4.4 Summary Statistics

Table 1 provides an overview on the realized returns of the sample banks on the three event dates. The average realized return for the first, second and third event was 2.28%, 2.24% and 3.05%. Referring to the first event date, the highest return is 7.90% while the lowest is -2.77%. For the second (third) event the highest return is 12.52% (9.30%) and the lowest is -2.93% (-2.12%). Table 2 gives an overview on the independent variables that are used in the cross-sectional regression. Panel A and B divide the sample according the fiscal years and present number of observations, mean, standard deviation, minimum and maximum. Panel C and D divide the sample into PIIGS and Non-PIIGS banks and provide the differences of the means of both groups.

On average, Moody’s has rated the banks with a Baa3, which is on the lower bound of the investment grade range. The lowest rating is Ca, which is described as „Highly speculative and

likely in, or very near, default, with some prospect of recovery of principal and interest.” 5. The

highest rating is A1 that is defined as „upper medium grade and subject to low credit risk.” 6 _The non-performance loans to total loans ratio has an average of 18.65% in 2015 and a slightly lower

5 _{https://www.moodys.com/researchdocumentcontentpage.aspx?docid=PBC_79004} 6 _{https://www.moodys.com/researchdocumentcontentpage.aspx?docid=PBC_79004}

22

average of 18.00% in 2016, indicating that banks are trying to reduce their bad loans. The bandwidth of this variable reaches from 1.80% (0.90%) to 53.6% (56.3%) for 2015 (2016) clearly highlighting the difference between the individual banks. Tier 1 capital ratio remains stable around 13.3% for both years, whereby the overall minimum and maximum is 8.17% (2016) and 19.9% (2016). The profitability of the sample banks, expressed by the return on assets reaches from min. -2.3% (-2.40%) to max 1.3% (1.40%) and has an average of 0.01% (0.00%) for 2015 (2016) The variable log value of total assets, which measures the size of the sample banks remains stable, while the minimum and maximum indicate a decent growth among the banks. Finally, in 2015 (2016) 22 (20) of 35 banks paid a dividend.

Referring to Panel C and D it is revealed that Non-PIIGS banks are rated 5 ratings higher than PIIGS banks. Next to that, the fraction of non-performing to total loans is roughly 14% lower for Non-PIIGS banks compared to PIIGS banks. Furthermore Non-PIIGS banks are on average bigger than their PIIGS counterpart. All three reported differences are significant.

23

5. Empirical Results

This chapter displays the empirical results of the event study and the cross-sectional analysis. In the first part, the individual abnormal returns are presented. The second part of this chapter deals with the results and implications of the cross-sectional regression analysis.

5.1 Event Study Results

Table 3 displays the results of the event study for each bank on each of the three announcement dates. Each column is linked to one announcement date.

Column 1 presents the results for the first event, which took place on 22nd January 2015. On this day the ECB announced the launch of the APP. This programme enables the ECB to buy bonds issued by Euro area central governments, agencies and other European institutions with a total monthly volume of EUR 60 billion7. Commonly this announcement highlights the beginning of European QE. Across the whole sample 6 banks generated a significant abnormal return within the 1% to 10% level. In total 17 banks experienced a positive abnormal return, while 16 banks realized a negative one. Among the 17 positive abnormal returns 9 are from PIIGS banks and 8 from Non-PIIGS banks. Referring to the group of 16 banks that experienced negative abnormal returns 13 banks are located in a PIIGS state, remaining 3 are not. The highest positive return of 4.34% and the highest negative return of -4.17% can be linked to Greek banks, namely Alpha Bank and Piraeus Bank. Worth noticing is that a large group of Italian and Spanish banks experienced a negative abnormal return. Contrary to this, banks from Greece, which is also among the PIIGS states, experienced a slightly positive abnormal return of 0.17%. The share price movements triggered by the first event do not meet the expectations in different ways. First, the fact that abnormal returns are nearly balanced between positive and negative is against the hypothesis No. 1, which predicts an overall positive share price reaction. Second, most of negative abnormal returns come from PIIGS banks, which seems counterintuitive and not in line with hypothesis No. 3. The most possible explanation for the observed results is that the expectations of the financial market or more precisely the expectations of market participants did not meet with the content of the announced QE programme. For instance, the volume of the monthly purchases or the eligibility of the asset might mismatch the expectations. These

24

individual non-matched expectations could explain the non-homogeneous price movements and is in line with findings of Krishnamurthy and Vissing-Jorgensen (2011). The authors observe different asset price reactions in dependence of the fulfilment or non-fulfilment of market expectations. Another possible explanation is that there is a certain level of uncertainty of the effectiveness as well as the consequences of the announced QE programme. Before the announcement of QE, the main interest rate stayed around the lower bound for a long-time period and various alternative programmes have been introduced to overcome the crisis. Through the introduction of QE, the ECB could have given the impression that these measures did not achieve the desired effect. Consequently, the announced programme could also be interpreted as another attempt with uncertain outcome. Thus, the non-uniform price reaction of the bank shares can be a sign that the financial market is not sure how to assess the QE programme and its consequences.

Column 2 displays the results of the second event. On the 10th March 2016 the ECB announced extension and adjustments on the current APP. First, the monthly volume of the purchases was increased from EUR 60 billion to EUR 80 billion. Second, the corporate sector purchase programme (CSPP) was added to the current APP. With this additional programme the ECB added investment-grade euro denominated non-bank corporation bonds to the universe of eligible assets8. Contrary to the first announcement, the second one had a clear positive impact on the Euro area banks. In total, 33 of 35 banks realized a positive abnormal return, of which 9 were significant within the 5% to 10% level. Divided into PIIGS and Non-PIIGS banks each group contains one bank that generated a negative abnormal return. As it has been the case for the first event, the highest positive and negative abnormal return came from the PIIGS banks. The Irish Allied Irish Bank set the maximum with 9.62% while the Spanish Banco BMP set the minimum with -0.55%. The overall results of the second event work in favour for the first and third hypothesis. Nearly all banks experienced a positive abnormal return while PIIGS banks experienced the majority of high positive returns. The share price reactions clearly indicate a positive reception of the second announcement and can be a sign for rising confidence in the financial market or a sign for investors’ rising sentiment.

Nevertheless, the observed results are contradicting with the second hypothesis that following QE announcements will have a lower impact than the initial one. However, Krishnamurthy and

25

Vissing-Jorgensen (2011) argue that the fulfilment or non-fulfilment of the market’s beliefs is the most important factor, which determines share price reactions. Assuming, that the first announcement may have disappointed the investors in terms of size or eligible assets it seems reasonable that the second announcement had a relatively great impact. Meaning and Zhu (2011) mention that the unexpected, or in other words, the surprising part of the announcement triggers the share price reaction. Arguing that the financial market or the investors did not anticipate the increase in the monthly purchase or the adding of the CSPP the exceptional share price reaction is reasonable.

Finally, column 3 exhibits the results for the third event on the 8th December 2016. The announcement states that the ECB will reduce the volume of the monthly purchases from EUR 80 billion to EUR 60 billion beginning in April 20179. Again, the overall reaction was positive with 25 banks experiencing a positive abnormal return and 10 a negative one. In total, 8 observed abnormal returns were significant within the 1% to 10% level. Grouped by PIIGS banks and Non-PIIGS banks, one can mention that shares of 14 PIIGS and 11 Non PIIGS banks reacted positively, while at the same time 8 PIIGS and 2 Non-PIIGS banks reacted negatively. The abnormal return of 6.45% generated by the Italian Banco BPM is the highest positive observed return and even significant at the 1% level. Simultaneously, the Greek Piraeus Banks obtained the highest negative abnormal return with -5.59%, significant at the 10% level. Worth noticing is that 3 out of 4 Greek banks realized huge negative abnormal returns. However, all Spanish banks and most of the Italian banks captured a positive abnormal return. The results of the third announcement are in line with the expectations and hypotheses. A large group of banks reacted positively to the announcement and most o the PIIGS banks did as well. Regarding the magnitude of the abnormal returns, it is revealed that the third announcement had less influence compared to the second one, which strengthens the second hypotheses. Meaning and Zhu (2011) provide a possible explanation. They argue that every announcement effect after the initial one suffers from general anticipation effects of the financial market. Another explanation for the smaller abnormal returns can be found in the content of the third announcement. There was no surprising element like adjusting the range of eligible assets or increasing the volume of the monthly purchases. Instead there was an announced reduction of the monthly purchase in the nearer future. Nevertheless, the forward guidance given in this announcement mentions that, if

26

necessary, the ECB can again adjust the programme with regards to lifetime, size and eligible asset classes. Through that commitment the ECB sent a clear and trustworthy signal to the financial market.

All three announcements have resulted in abnormal returns with at least some of them being significant. For the initial announcement, the results suggest that the financial market was unsure how to assess the effectiveness of the announcement QE programme. This led to non-uniform abnormal returns, where nearly half of the sample banks experienced negative abnormal returns. Notwithstanding, the second event triggered positive abnormal returns for nearly all banks. This reaction clearly proofs that the financial market was convinced from the second announcement. Contrary to the first event, PIIGS banks experienced higher abnormal returns. Finally there can be seen similar patterns in the third and second event, except for the overall number of positive abnormal returns slightly declined as well as the overall magnitude of the abnormal returns. These non-uniform but overall positive share price reaction have been also overserved in the related literature. Joyce et al. (2011) find non that share prices in the U.K. did not move in a consistent way according to the QE announcements of the BoE, while Kobayashi et al. (2006) find the a similar pattern for the QE announcements of the BoJ. In addition, Fratscher et al. (2012) find rising share prices around the QE announcements of the Fed.

5.2 Cross Sectional Regression Results

After showing that the QE announcements led to abnormal returns for European banks the next step is to see if these observed returns are determined by the financial condition of the sample banks. Therefor an ordinary least squares regression with abnormal return as dependent and six financial indicators as independent variables are used. The financial condition of a bank is described by the following six indicators: Long Term Credit Rating, Non-Performance to Loans

Ratio, Tier 1 Capital Ratio, Return on Assets, Dividend and Total Assets. A dummy variable, PIIGS, specifies if a bank is based in a PIIGS state or not. This dummy joins in the second

specification of each regression.

In line with the third hypothesis this research expects negative coefficients for Long Term Credit

Rating, Tier 1 Capital Ratio, Return on Assets, Dividend, Total Assets and a positive one for Non-Performance to Loans Ratio. Table 4 shows the results of the cross-sectional analysis.

27

However, it is important to notice that the explanatory power of the regression is limited by the low sample size. Regardless, some coefficients enter at the 5% and 10% significance level. Across all three events only two variables enter with a constant coefficient sign. On the one hand, the Long-term Credit Rating variable maintains its unexpected positive sign, for the regression referring to the first event even significant at the 5 % level. This result shows that the financial market believes that banks with a higher credit rating compared to lower rank ones, are going to benefit more the QE. As a higher credit rating signals that a bank has a low risk of default this finding contradicts with the third hypothesis.

On the other hand, the Dividend Payer variable enters all three events with the expected negative sign, for the first event even with a significance level of 10%, revealing that dividend paying banks are expected to generate lower abnormal returns then non dividend paying banks. Assuming that only banks that are in a good financial condition pay dividends this finding is in line with the third hypothesis.

The Non-Performing to Total Loan Ratio variable enters the regressions for the first and second event with the expected positive coefficient sign, whereas it enters the third regression with a negative one. More important, adding the PIIGS dummy leads to an increasing significance level for the first regression. While the variable was only significant at the 10% level before adding the PIIGS dummy its significance level increases to the 5% level. Together with the expected positive sign, indicating that banks with a larger fraction of non-performing loans are going to benefit more from the announced QE programme this finding supports the third hypothesis.

Tier 1 Capital Ratio, which is expected to enter with a negative sign, does so only for the

regressions of the first event. For the remaining two dates the Tier 1 Capital Ratio variable has a positive coefficient sign, suggesting that banks with a higher Tier 1 Capital Ratio are expected to benefit more from QE. Across all events the variable fails to enter significantly. Nevertheless, as a high Tier 1 Capital Ratio is linked to banks in a good financial condition, the variable provides in sum ambiguous evidence for the question if banks in a bad financial condition benefit more from announced QE programmes.

Referring to Return on Assets variable, the regression results show that it only enters the regressions for the third event significant by the 5% level and with the anticipated negative sign. This confirms the assumption, that low profitable banks derive greater benefits from the QE programme and therefore is in line with the third hypothesis. Notwithstanding the regressions of

28

the remaining two event dates the Return on Asset variable enters without any significance and unexpected positive coefficient sign.

With reference to the Total Assets variable, one can see that it enters for the first and second event date with the expected negative coefficient, while it has a positive one when entering the regression for the third event. Across all three events the variable fails to enter significantly. As it has been the case with the Tier 1 Capital Ratio variable, the Total Assets variable allows no clear interpretation for or against the third hypothesis.

Including the PIIGS dummy variable in the second specification of each regression has no significant impact on the regression results, except for the first event. After including the PIIGS dummy variable, the Long-term Credit Rating variable as well as the Dividend Payer variable do not enter with their prior significance level of 5% and 10%, respectively. Instead both variables now enter insignificantly. In contrast, the significance level of the Non-Performing to Total Loan

Ratio variable rises from 10% to 5% level after the PIIGS dummy variable is added.

All in all it can be concluded that the cross sectional analysis provides mixed evidence with regard to the third hypothesis. Nevertheless, there is slightly more significant evidence, which works for the third hypothesis. Despite the fact that one variable enters significantly between the 10% to 5% level but the unexpected positive coefficient three other variables also enter within the 10% to 5% significant level and with the anticipated coefficient sign.

The variable for the fraction of non-performing loans to total loans (Non-Performing to Total

Loan Ratio) enters the first regression with the anticipated positive coefficient sign and at the

same time significant at the 5% level. Kobayashi et al. (2006) find for a similarly constructed measure of bank’s asset quality the same results. In addition, the variable that measures profitability through the return on assets (Return on Assets) enters the third regression also within the 5% significant level. Supporting the aforementioned variables the dividend payer variable (Dividend Payer) enters all regressions with the anticipated negative sign, once even significant at the 10% level. In conclusion, it looks that financially weak banks benefited more from the QE programme of the ECB.

29

5.3 Additional Results and Robustness

In this section two different approaches are used to check the robustness of the obtained results. First, a wider event window is used to take potential anticipation effects into account as well as delayed reactions. Based on that, this research runs additional regressions, using three-day cumulative abnormal return as the dependent variable, while the independent ones remain the same. Second, a cross-country event study is conducted, where the sample is divided into an PIIGS and Non-PIIGS banks portfolio. In this event study, the Euro Stoxx 50 is used as the proxy for the market index instead of the country specific indices.

Brown and Warner (1985) propose a small event window based on the fact that the power of the test statistic is decreasing with using a wider event window. Even more precise are McWilliams and Siegel (1997), who suggest that a single event day window (0) is the optimal length to avoid influence of other confounding effects. However, MacKinlay (1997) brings forward the argument that a three-day event window (-1/0/+1) window can capture confounding effects, like a possible anticipation effects or delayed reaction effects. Following MacKinlay (1997) and under consideration of Brown and Warner (1985) an additional event study with a three-day event window is conducted to verify the robustness of the previous findings.

Table 5 reports the event study results for the three events using a three-day event window. Referring to the first event, a three-day event window confirms the non-uniform movement of the financial market that represents the uncertainty of the effectiveness of the announced QE programme. First of all, the number of significant returns rises from 6 to 11. More crucial, 29 out of 33 banks generating a negative abnormal return. This points towards a more negative market reaction compared to the one-day event study. Simultaneously, the magnitude of the negative individual abnormal returns is rising. Among the group of banks that experienced exceptional high negative abnormal returns are mostly Spanish banks, where 5 of 7 abnormal returns were significant. The three-day event window abnormal returns of the second event appear in the second column. Verifying the one day event window results, 33 banks obtained positive abnormal returns, while 2 banks remain with a negative abnormal return. As already observed before, the magnitude of the abnormal return is increasing, contrary to the number of significant returns that remains constant. Finally the third column presents the results of the final event. In contrast to the second event a wider event window does not endorse the previous

30

results. The number of positive abnormal returns is declining from 25 to 15 just as the number of significant returns, which declines from 8 to 3. Compared to the one day event window the share price reaction is more balanced out across all banks. This is also true for the group of PIIGS banks, which is now equally divided into positive and negative abnormal returns.

Taking all these three events as a whole on can state that the three-day event study fully verifies the results from the second event and partly the findings from the third event. Referring to the first event, the results provide evidence which support the possible explanation of the results presented in the preceding chapter. However, when reading these results one should always be aware of that longer event windows do not only incorporate possible anticipation or delayed reaction effects but also other share price influencing information. Especially bank shares, which are influenced by a wide range of different types of information a wider event window, can bias the results.

Table 6 shows regression results using the three-day abnormal return as dependent variable and bank characteristics as independent ones. While the biggest difference is observed for the Return

on Assets variable a smaller one can be seen looking at the the remaining variables. The Return on Assets variable enters now significant at the 5% level across all regression and two times with

expected negative coefficient sign, strengthening the findings of table 4. Unlike before The

Long-term Credit Rating variable enters now across all regressions with no significance but two

times with the expected negative coefficient sign. While the Non Performing to Total Loan Ratio variable entered significant at 10% and with the expected coefficient sign before, it enters now significant at the 5% level but with the unanticipated negative sign. This contradicts not only the previous finding; even more it provides evidence against the third hypothesis. In favour of the third hypothesis and contrary to the results in table 4, the Tier 1 Capital Ratio variable enters now significant at 10% and with the predicted coefficient sign. Concerning the Dividend Payer variable it can be mentioned, that it maintains its anticipated coefficient sign across all regressions, but unlike before fails to enter significantly even once. Nonetheless, this finding provides weak evidence for hypothesis No. 3. Lastly, the Total Assets variable enters the regression referring to the first event significantly within the 10% level. Across all regressions the coefficient of the variable has an unexpected positive sign. Taking the significant entry and the constant unexpected sign together the Total Assets variable provides evidence against the