Quantitative easing and financial intermediaries : the impact of Federal Reserve's QE announcements on US

(1)

UNIVERSITY OF AMSTERDAM AMSTERDAM BUSINESS SCHOOL

MSc Business Economics Master Specialisation Finance

Quantitative Easing and Financial Intermediaries

The impact of Federal Reserve’s QE announcements on US

financial intermediaries

Author: Fabjan Abazaj Student number: 11136731

Thesis supervisor: Dr. Razvan Vlahu Date: 07/07/2016

(2)

Abstract

As a result of the latest global financial crisis, the US Federal Reserves (Fed) started implementing unconventional monetary tools in order to stimulate the distressed economy. This research examines the impact of the three main quantitative easing (QE) announce-ments on the equity value for a sample of US financial intermediaries by conducting an event study. Additionally a cross-sectional regression on abnormal returns based on CAMEL model indicators is conducted in order to shed light if the QE program disproportionally impact financially weak banks. The results provide evidence that the initial QE announce-ment has a strong positive impact on financial intermediaries compared to the second and third QE announcement where the evidence is mixed. Additional results indicate that the QE program benefits more financially weak banks.

Keywords: Quantitative Easing, Financial Intermediaries, Event study, Federal Re-serve

(3)

Statement of Originality

This document is written by student Fabjan Abazaj who declares to take full responsi-bility 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 com-pletion of the work, not for the contents.

(4)

1 Introduction

Tregenna (2009) outlines that prior to the latest financial crisis, US banks experienced a period of high profitability. The significant increase in leveraging has aided financial in-termediaries in increasing their profits. Moreover, the credit boom fueled by US financial intermediaries planted the seeds for the most disastrous financial crisis in the history. The aftermath of the crises highlighted the need for the US Federal Reserve (Fed) to shift from conventional monetary policy tools to unconventional ones in order to save the economy from the financial turmoil. Following the collapse of Lehman Brothers, Fed embarked on such unconventional monetary path. These relatively new, non-standard tools known as Quantitative Easing (QE) refer to the creation of Federal Reserve funds in order to pur-chase financial assets from the private sector. Prior to the implementation of QE, the target federal funds rate, known as the conventional monetary policy instrument of Fed has been at the zero lower bound. Mishkin (1996) argued that when nominal interest rates reach the zero level bound, conventional monetary tools of central banks become ineffective in order to boost the outlook of the economy. Consequently, Fed faced with impotent conven-tional tools switched to unconvenconven-tional ones in order to stimulate the economy. Through an open market operation, the US central bank purchased assets such as US Treasury Bonds, debt obligations of the Government-Sponsored Entities (GSEs) and Mortgage-Backed Se-curities (MBS). The QE program began on 25th of November 2008 and consists of three main periods namely QE1, QE2 and QE3 and lasted till October 2014. Over this period, Fed purchased more than $3.5 trillion worth of financial assets from troubled financial in-termediaries, resulting in the largest intervention in the history of financial markets.1. In retrospect, the total monetary value injected in the economy accounts for roughly 20% of US 2016 GDP. Considering this unprecedented open market operation, current literature distinguishes between two main transmission channels through which QE program affects the economy. The first one is the signaling channel, where the central bank through the QE announcements about future monetary policy decisions changes and shapes market

(6)

pectations (Krishnamurthy and Vissing-Jorgensen 2011, Christensen and Rudebusch 2012). The second refers to the portfolio balance channel, where the central bank through asset purchases, reduces the supply of available assets held by the private sector. The low supply of an asset can induce economic agents to purchase other close asset substitutes, eventually bidding up their price. (Gagnon et al. 2011; Krishnamurthy and Vissing-Jorgensen 2011).

These extraordinary measures aimed to ease financial conditions, as well as putting a break to the crisis while supporting economic recovery. Several studies have examined the QE impact on the US economy and the financial market. Overall, these studies point towards a similar conclusion that the effect is significantly positive for the economy and has a considerable impact on the financial market. For instance, Gagnon et al (2010) provide evidence that the Large-Scale Asset Purchase (LSAP) conducted by the Fed has a significant impact in lowering long-term borrowing rates which eventually can boost economic activity. Krishnamurthy and Vissing-Jorgensen (2011) investigate the first two periods of QE, and conclude that both periods have a considerable effect in lowering interest rates on those assets purchases by the Fed. Furthermore, DAmico and King (2010) find that the QE program as a whole results in a downward pressure in the yield curve of roughly 50 basis points. Before the US QE program implementation, in March 2001 the Bank of Japan (BOJ) conducted similar unconventional monetary policies in attempt to spike economic activity through the purchase of Japanese Government Bonds, Kimura and Small (2006). As a result of the financial crisis, on March 2009 the Bank of England (BoE) started implementing the QE program by purchasing private and public sector financial assets but with a majority of assets purchased being UK government bonds. Joyce et al. (2010) studying the impact of BoE QE policy on government bonds, find that asset purchases from QE financing lowered medium to long term yields. The European Central Bank (ECB) pursued similar unconventional monetary policy actions as those of the Fed and BoE in order to stimulate the Eurozone economic activity. One of the interventions refers to the Security Markets Program (SMP) where the ECB purchased large amounts of government debt from the financial sector. Eser and Schwaab (2013) analyze the impact of SMP towards the yield

(7)

of assets purchased and find that the program lowers the yields on five-year maturity bonds for certain Eurozone countries.

The QE program spurred a lot of discussion since the approach is quite unprecedented in regard to its composition, size and overall intervention in the financial market which makes it an interesting topic to examine. Most of the proponents support the idea that the QE program boosted the troubled post-crisis US economy and steered it away from a deflationary spiral scenario. On the other hand, opponents support the idea that the program might lead excessive financial risk taking, since low interest rates create cheap credit which can be invested in financial assets.2. Additionally they argue that the QE impact in stimulating the real economy is minor and not evident. The majority of the studies examining the case of the US conclude a positive effect of the QE program on the economy, although the magnitudes of their estimates differ. In light of this, the impact of the QE program can be considered as significantly positive towards the financial market and the real economy. Most of the literature mentioned on this research focuses on examining the QE impact on financial assets as well as macroeconomic effects. Yet, it will be interesting to study the QE effect on the part of the financial market which triggered the crisis as well as the distress suffered during that time.

This study investigates the QE effect on financial intermediaries, specifically the impact on banks equity value by raising the question: “To what extent did the US QE announce-ment by the Federal Reserve impact financial intermediaries equity value?” Moreover, it is compelling to examine if these unconventional monetary measures have a significantly different impact on financial intermediaries conditional on their characteristics.

This research conducts an event study on the three main announcement dates of the US QE program in order to estimate the effect on financial intermediarys equity value, i.e abnormal and cumulative returns. The three main event dates refer to the announcement of the three period of QE. Financial intermediaries will comprise a sample of 51 US banks collected from Datastream. Next, this study implements a cross-sectional regression on

2

(8)

abnormal returns for the sample of banks in order to shed light on whether the QE measures have a different effect depending on banks characteristics. The indicators of the CAMEL model will be used for banks characteristics.

The rest of this thesis is structured as follows. Section II provides a brief summary of the existing literature related to the QE topic. Section III lays down the foundation and hypotheses of this research. Section IV describes the methodology and data. Section V discusses the results of the event study as well as cross-sectional estimates on bank characteristics. Section VI provides additional results and robustness checks. Lastly, Section VII concludes and states the limitations of this study.

2 Literature Review

Mishkin (1996) argues that when nominal interest rates reach the zero level bound, conven-tional monetary tools of central banks become ineffective in order to stimulate the outlook of the economy. Bernarke and Reinhart (2004) put forwards three alternative tools which aid central banks when they are faced with a nominal short-term interest rate approaching zero. First, central banks can influence interest rate expectations of the public by guaran-teeing that the future short-term interest rate will be kept low. Secondly, a central bank can expand the composition of the balance sheet by increasing its size while exceeding the required level needed to optimize short-term policy rate which is also referred as “Quanti-tative Easing” (QE). This method involves money creation by the central bank in order to swap securities held by the public while simultaneously increasing the money supply in the market. Lastly, it can adjust the composition of the balance sheet aiming to impact the relative supplies of securities available in the market.

In 1961, in the US economy nominal interest rate reached the zero level bound and the central bank started to alternate its balance sheet. This alternation is known as “Operation Twist” and marks the first large-scale asset purchase implemented by the Fed as a non-standard monetary policy tool. Through the purchase of long-term Treasury Bonds financed by the sales of short-term Treasury Bills this operation aimed to lower long-term interest

(9)

rates. In essence, this operation cannot be fully considered as Quantitative Easing since the financing do not involve money creation nor expansion of the central bank’s balance sheet. The first study examining the large-scale asset purchases effect as a monetary tool is Modigliani and Sutch (1966). Their findings pointed towards that the effect of this operation does not significantly lower long-term interest rates. Contrary to these findings a recent study by Swanson et al. (2011) concludes that the operation indeed has a significant effect on long-term Treasury yields but the impact is modest in magnitude resulting in a decrease of 15 basis points on long-term Treasury yields.

Prior to the recent financial crisis, in the period of 2001-2006 the Bank of Japan started implementing the QE program which sparked interest in the literature to examine the effects of this type of unconventional monetary policy. Ugai (2006) provides a survey of empirical evidence of the Japanese experience, while his discussion compromises of varied opinions. The author concludes that the program substantially impacts market expectations through signaling that interest rate will be kept lower in the future. On the other hand, he found mixed evidence whether the program directly affected risk premia or bond yields. Yet, Bernanke et al. (2004) analyzing the effects of the Japanese QE program do not find a considerable impact of the announcement effects but rather their evidence points towards that the Japanese yields decreases by roughly 50 basis points. Another study examining the Japanese experience of Kobayashi et al. (2006) finds positive effects of the QE program in the banking sector. By conducting an event study on the introduction and expansion of the program, the authors find that the Japanese QE results in excess returns in the banking sector, portraying a positive impact of the market reaction. This favorable effect of QE on the Japanese banking sector is also supported by Bowman et al. (2011) findings which examine the efficacy of the QE program in stimulating bank lending. By using bank level data ranging from 2000-2009 they find a positive and significant effect of bank lending indicating that the QE expansion plausibly enhanced the flow of credit. Nonetheless, the magnitude of the enhancement is likely small.

(10)

to further stimulate the distressed economy. The literature witnessed a dramatic increase on this research topic subsequently. The Federal Reserve Bank (Fed) is the first central bank to accommodate the QE program while The Bank of England (BoE) and The European Central Bank (ECB) followed suit. Most of the research in examining the QE program solely focuses on assessing the impact on financial markets as well all on a macroeconomic level. However, other studies have been conducted to assess spillover effects of the QE program. Although four central banks implemented QE program, the motivations and circumstances of the program differ from one another. For instance, Fawley and Neely (2013) argue that ECB and BoJ centered their programs towards direct lending to financial intermediaries which indicates the bank-centric structure of their financial system. On the other hand, Fed and BoE expanded their monetary base through the purchase of bonds which reflects their market-centric economy.

Referring to the US experience, almost all studies conclude that the QE program has a notable positive impact on long-term bond yields, although the magnitude of the estimates differs. For instance, Gagnon et al. (2010) examine the Large-Scale Asset Purchase Program (LSAP) of QE through an event study analysis of Fed communications in order to shed light regarding the effects of LSAP in changes in asset prices: 2 and 10 year U.S. T-bills, agency MBS yield, 10-year treasury premium, 10-year swap rate and the Baa corporate index yield. They examine changes in interest rates around the official communications while taking cumulative changes as a measure of the overall effect. They found that the $300 billion purchase of US Treasuries led to a decrease of roughly 90 basis points in US log-term Treasuries. Applying the same methodology, Krishnamurthy and Vissing-Jorgensen (2011) find that the LSAP program of QE1 and QE2 has a significant impact on lowering the nominal interest rates on Mortgage-Backed Securities (MBS), Treasuries, Agencies related securities and Corporate Bonds. However, the magnitude of these effects varies for different QE periods, for different types of securities purchased and for different maturities. For instance, during QE1 where a large amount of Agency MBS and Agency Debt are purchased, the initial stage of the program led to a large decrease in MBS rates.

(11)

In contrast, QE2 which involves the purchase of long-term Treasuries the effect is relatively contained. Likewise, D’Amico and King (2010) investigate the effects of QE1 purchase of $300 billion of U.S T-bills conducted by the Fed. The authors use a panel of daily CUSIP level data in order to estimate the impact of LSAP on price elasticities and substitutability effect of US Treasuries and eventually the reduction on yields. Consistent with the previous studies examining the US experience they find that the program resulted in a decline of 3.5 basis points on the days when the program occurred. However, the overall effect of the program caused an enduring downward shift in the yield curve to roughly 50 basis points.

Neely (2010) conducts an event study in which the researchs focus shifts towards exam-ining the impact on international bond yields and exchange rate. The author finds that the Fed announcement of the LSAP has a notable impact on foreign bond yields and exchange rates. Stroebel and Taylor (2009) analyze time series data and mainly focusing on the quantitative purchase of Mortgage-backed securities conducted by the Fed. They find a rel-atively small effect on the spread between swap yields and MBS yields. Similarly, Hancock and Passmore (2011) examine the Fed purchase of MBS but they focus on assessing the impact on mortgage rates. Overall, they find that the purchase affected mortgages rates by significantly putting downward pressure.

Gambacorta et al. (2012) asses the effectiveness of the unconventional Monetary policy while conducting a cross country analysis. By estimating a panel VAR from eight developed economies, the study evaluates the macroeconomic effects of the unconventional monetary policies. The authors find that an expansion of the central bank balance sheet while the nominal interest rate approached to the zero bound resulted in a non-permanent increase in economic activity and consumer price. Yet, the authors conclude that this expansion in the central bank balance sheet does not infer a positive macroeconomic effect.

A majority of the studies investigating the UK experience with the QE program fo-cuses on the impact on the financial market especially on the financial assets and overall macroeconomic effect. In similar fashion as the Fed asset purchases, the Bank of England (BoE) purchased private and public sector financial assets but with a majority of assets

(12)

purchased being government securities also known as UK “gilts”. Joyce et al. (2010) assess the impact of the Bank of England QE policy on asset prices focusing on the effect of the gilt purchases by conducting an event study analysis and survey data. They find that asset purchased from QE financing may have lowered medium to long term gilts. Additionally, the authors argue that the QE impact in encouraging the macro economy is hard to esti-mate since the transmission mechanism is subject to long lags and hard to disentangle the specific contribution of asset purchases. Likewise, Meaning and Zhu (2011) examine the fruitfulness of the Fed and BoE asset purchases on the financial market. Consistent with the results of previous studies, the authors find that the asset purchases have a significant positive effect on financial markets: bond yields significantly decreased while the price of some risky assets appreciated. However the magnitude of the effect decreased in tandem with subsequent extensions of the asset purchase program. Kapetanios et al.(2012) assess the macroeconomic impact of the initial round of QE conducted by the BOE. The author’s results portray that GDP and annual CPI inflation reacted as a consequence of the QE program. Most importantly, they conclude that in the absence of the program real GDP would have experienced a higher decrease and inflation would have been lower or even worse, negative. Notwithstanding, Lyonnet and Werner (2012) assessing the effectiveness of QE program on nominal GDP growth conclude that the program has no apparent impact on the UK economy.

As a consequence of the recent financial crisis of late 2007, the European Central Bank (ECB) embarked in similar unconventional and non-standard program as the Fed and BoE. One of the programs refers to the 3-year Long-Term Refinancing Operations (LTROs) which aimed supporting bank lending and liquidity in the euro area. Darracq-Paries and De Santis (2015) examine the macroeconomic impact of LTROs and argue that the main transmission channel of this unconventional policy is lubricated through the reduction in liquidity and funding risks in the Eurozone banking sector. Their results suggest that the LTROs program increases GDP, loan volume to non-financial corporations while simultaneously increasing goods and prices, eventually leading to credit-crunch prevention. Moreover, unconventional

(13)

tools introduced in the Euro zone reinforced market functioning and boost macroeconomic activity and employment although the increase is modest but nevertheless significant, Lenz at al. (2012).

Another unconventional program refers to the Securities Markets Program (SMP) which reflects the asset purchased from ECB. Eser and Schwaab (2013) examine the implications of SMP towards the yield impact of asset purchased in five euro area sovereign bonds. They concluded that the SMP decreased the yields on five-year maturity bonds of Spain, Italy, Ireland, Portugal and Greece also known as PIIGS euro countries. Focusing on the same countries Falagiarda and Reitz (2015) examine the impact of ECB announcements regarding unconventional monetary policy tools on the sovereign spreads relative to their counterparty, Germany for the period between 2009 2012. Their estimates suggest that ECB announcements have a notable impact in reducing long-term government bond yield spreads counterparts to Germany. However, Greek bond yield spreads are immune towards the announcement effect.

3 Hypotheses

This section discusses the main hypotheses of the research which are built on the existing literature.

3.1 Hypothesis 1

“The QE announcements of the Fed will create positive abnormal returns for US financial intermediaries”

Most of the above mentioned literature lays the empirical groundwork that indicates a positive impact of QE on the financial markets as well as on the overall economy (Bowman et al. 2011; Gagnon et al. 2011; Joyce et al. 2010; Meaning and Zhu 2011; Kapetanios et al. 2012). Therefore, referring to the aforementioned evidence, this research expects a favorable impact on financial intermediary’s equity: translated in positive abnormal returns. Moreover, the hypothesis can be supported by the findings of Kobayashi et al. (2006) which

(14)

find significant positive effects for every announcement of Japanese QE towards bank equity values.

3.2 Hypothesis 2

“The first QE program will have stronger effects compared to subsequent stages”

The studies of (Krishnamurthy and Vissing-Jorgensen 2011; Meaning and Zhu 2011) indicate that the impact of QE on the financial assets, macro economy and overall financial market is more intense during the first round of the program. For instance Meaning and Zhu (2011) argue that the initial degree of effectiveness for the first stages of the QE program is hard to achieve for subsequent terms since the surprise element eventually disappears. While Krishnamurthy and Vissing-Jorngense (2011) state that the magnitude of the QE program effect differs with respect to the asset purchased, over maturities and across QE1 and QE2.

Thus, this study expects similar magnitude for the initial round of the US QE since the surprise element of the program disseminates for subsequent rounds and the effect of the program changes in regard to the composition, maturities and timing of each initiative. The change in magnitude can be reflected in higher abnormal returns for the first round compared to subsequent rounds.

3.3 Hypothesis 3

“Weaker financial intermediaries will benefit more from the QE program”

Kobayashi et al. (2006) - outlined in addition to their findings - that the Japanese QE resulted in excess returns in the banking sector; they also provide cross-sectional evidence that the program will unequally aid financially weaker banks. In their research, they con-sider bank-specific characteristics and indicators which relate mainly to profitability and liquidity. Referring to their findings, this research may corroborate a similar conclusion that the QE has a major impact on financially weaker banks compared to stronger ones.

(15)

4 Methodology & Data

This section describes the methodology and data used throughout this study. The upcoming subsection provides information about the data used in this research and data sources. Followed by the subsections which explain the event study as well as the cross-sectional regression on bank characteristics based on the CAMEL model. The latter subsections outline the descriptive statistics for the dependent and independent variables used in the cross-sectional regression, the description of the variables and a correlation matrix.

4.1 Data description

The timeline for the program starts from 25th November 2008 referring to the initial Fed announcement and ends on the 29th of October 2014 with the announcement to conclude the asset purchase program. This research uses 3 main event dates from Fed communications which refer to the introduction of the three QE periods. Appendix A presents a list of the QE program key decisions announced by the Fed. The event dates are first-handedly gathered by the Fed website and their press releases. A sample of US financial intermediaries is collected through Datastream which includes historical stock prices needed to calculate actual returns, abnormal returns and cumulative abnormal returns. Appendix B provides the list of financial intermediaries examined for the event study. Returns for the S&P 500 index and NYSE stock exchange are calculated for the period consistent with the QE program. Table 1 presents summary statistics for the index returns and the market returns. Balance sheet data for bank characteristics used in the cross-sectional regression are accessed at Compustat-Capital IQ3. By using this database, this research compensates for the inability to access balance sheet data for the full sample of banks used for the event study but only 27 banks out of 51. This data limitation impacts the significance of the

3

While using Compustat-Capital IQ this research could access balance sheet data for a higher number of US banks compared to using Datastream where most of the data needed was missing for the sample of banks. The most appropriate database would be Bankscope, however this research could not get access to it.

(16)

results used in the cross-sectional estimates as data for 27 banks does not provide enough robustness in explanatory power.

4.2 Event Study

In order to estimate the effect of QE announcements on US financial intermediaries, an event study is used. Referring to Mackinlay (1997), abnormal returns are estimated through em-ploying the market model approach and two day event window (0;+1). The benchmark market model used to calculate daily expected returns for the sample of financial interme-diaries refers to:

Rit = αi+ βiRmt+ εit (1)

(εi= 0) var (εit) = σ2εit

Where Rit and Rmt are the period-t returns on security i and the market portfolio, respectively, while t is the error term assumed to have zero mean. αi and βi and εit refer to the parameters of the market model. ˆαi and ˆβi are estimated through an Ordinary Least Squares (OLS) regression using a T=200 trading day estimation window. S&P 500 Index is used to estimate the market model parameters. The difference between actual returns and estimated returns of OLS results in abnormal returns which looks as follows:

ARit= Rit− ˆαi + ˆβi+ Rmt (2)

In order to determine if abnormal returns are statistically different than zero, the fol-lowing test statistics is conducted:

t − test = ((Σar) /n) / (ar sd/sqrt (n)) (3)

Where ar refers to abnormal returns, ar sd the standard deviation of the abnormal returns for each financial intermediary and n refers to the number of days in the event

(17)

window. If the t − test absolute value is equal to or higher than 1.96, then the abnormal return for the relevant financial intermediary’s stock is statistically significant at a 5% confidence interval. If the t − test in absolute terms is higher or equal than 1.68, then the abnormal returns for that financial intermediarys stock is statistically significant at a 10% confidence interval. Finally, if the t − test is higher or equal to 2.58 in absolute terms, the abnormal return is statistically significant at a 1% confidence interval.

The impact of US QE announcement can be positive, negative or zero. Nevertheless, referring to the existing literature this study expects positive abnormal returns for the sample of US financial intermediaries. It is also possible that the QE announcement may result in negative or zero abnormal returns. For instance, this can occur when the content and the intervention of a QE period was not deemed effective from the financial market and additional stimulus packages were needed to boost the US economy.

4.3 Cross sectional regression on bank characteristics

After reviewing the results of abnormal returns on the sample of US financial intermediaries, this section examines differences in abnormal returns. In order to find if the QE program supports financially weaker intermediaries I include six bank specific indicators where the CAMEL model is used as a baseline which is based on: capital adequacy, asset quality, earnings, liquidity and size. In addition, this research includes another variable in order to provide supplementary estimates for a financially weak bank. The variable is named fragility and is calculated as the ratio of non-performing loans to total assets. These ratio attempts to capture the fragility of a bank based on its core assets: loans. This indicator is similar to asset quality since both give somewhat of an indication on the quality of a banks asset. Referring to Hypothesis 3, if the QE program unequally benefited financially weaker banks, this research expects indicators that reflect a weak financial condition to get a positive sign. On the other hand, a negative sign is expected on those indicators which demonstrate a strong financial condition.

(18)

specific indicators as dependent ones. Using an OLS regression model, abnormal returns equal bank characteristics consistent with the CAMEL model.4

The regression model looks like:

ARit = β0+ β1CapitalAdequacyit + β2AssetQualityit+ β3Earningsit +β4Liquidityit + β5Sizeit + β6Fragilityit+ εit (4)

Following the theory of Pasiouras et al. (2006) the indicators for the CAMEL model are chosen which are described in the upcoming section.

4.4 Description of the variables

Capital Adequacy provides an indication of a banks capital soundness. Additionally, capital serves as the first line buffer in absorbing a banks risks: liquidity, credit and interest rate risk. In their research Pasiouras et al., (2006) define capital adequacy as a ratio of total equity to total assets. However, they argue that Tier 1 Capital Risk-Adjusted Capital ratio will serve as a more appropriate measure but their research was faced with a considerable number of missing values in the dataset. According to (Basel Committee, 2004), this ratio is defined as common shareholder’s equity plus minority interests, minus the percentage of preferred stock and goodwill as a portion of risk-weighted assets. In order to provide a better indication of a bank’s capital strength this study opts for Tier 1 ratio in defining capital adequacy. The variable CapitalAdequacyit measures the capital strength of bank i at time t. The higher this ratio, the “healthier” a bank’s financial state and the higher the buffer to absorb potential losses. Hence, this research assumes a negative relationship between Capital Adequacy and ARit (β1 less than 0).

In most cases, loans comprise the majority of a bank’s assets. It follows that Asset Quality can be measured by the quality of a bank’s loans. Loan loss provision covers the losses a bank incurs from non-performing loans while net interest income refers to the

4

Management Quality indicator of the CAMEL model is not included in this research because data to estimate this variable could not be accessed

(19)

difference among the income arising from bank’s interest earning asset and the charges related to paying out liabilities. This ratio provides an indication among the provisions of bad loans of a bank and the net interest income earned. A high loan loss provision demonstrates a high probability of non-performing loans which entails a low asset quality for a bank. The variable AssetQualityit describes asset quality of bank i at time t where a high value for this variable outlines a weak financial condition for a bank. Consequently, this research assumes a positive relationship with ARit (β2 higher than 0).

Profitability of a bank is usually measured by the ratio Return on Asset (ROA) where earnings before interest, taxes, depreciation and amortization are divided by total assets. Among others, the studies of Pasiouras et al., (2006), Wheelcock and Wilson (2000), Poon and Firth (2005) use the same ratio to measure profitability. The variable P rof ititcalculates the profitability of bank i at time t. the higher the return on assets the more justified the financial status of a bank. Therefore a negative relationship with ARit is expected (β3 less than 0).

Liquidity is as a crucial component for a bank which can serve to satisfy its liabilities. In their work Pasiouras et al. (2006) define liquidity as liquid assets to customer and short term funding. This research could not access data for customer and short term funding and instead used a ratio of liquid assets to total assets. Liquid assets comprise those assets which can be easily be converted into cash like treasury bills, deposits with central banks, cash itself and demand balances with other banks. The variable Liquidityit shows the liquidity of bank i at time t. A bank with a high liquidity ratio, a bank can be considered financially “strong”. Therefore, I expect a negative relationship with the dependent variable ARit (β4 less than 0).

The size of a bank measured as the logarithm of total assets of bank i at time t, in many studies is considered as an indicator in explaining the financial condition. For instance, Lennox (1999), found that larger banks compared to smaller ones are less probable to face credit restrictions Moreover Pasiouras et al. (2006) argue that bank size can be considered as an essential default determinant which can be used as an estimate for a bank’s financial

(20)

condition. The higher the value of a bank’s assets defined as Sizeit the more sound its financial condition. Consequently, a negative correlation with ARit is expected (β5 less than 0).

Berger et al (2008) state that an increase in non-performing loans will directly impact a banks risk exposure. As a result, a high number of delinquent loans will adversely affect a banks financial stability and lead to a weaker financial state. The variable F ragilityit describes bank’s i financial condition at time t measured as the ratio between non-performing loans to total assets. Hence, a positive relationship with the dependent variable is expected ARit (β6 higher than 0).

4.5 Descriptive statistics

Table 1 provides summary statistics on the returns for S&P 500 Index and NYSE stock exchange for the period consistent with QE timeline, 2008-2014. The highest return for S&P 500 and NYSE is 10% while the mean for both is negative and low. During this period, the US financial market experienced a lot of fluctuations as a consequence of the financial crisis which explains average negative returns.

Table 2 provides the summary statistics for the cross sectional regression on the first QE announcement based on six bank specific indicators.5 The dependent variable is abnormal returns the sample of banks while the independent ones are: Capital Adequacy, Asset Quality, Earnings, Liquidity, Size, and Fragility.

On average abnormal returns are positive at 4.6% while the range varies from -12% the minimum to the highest at 49% reflecting a considerable gap. Overall, the QE announce-ment resulted in positive abnormal returns. Capital Adequacy measured as Tier 1 Capital ratio has a mean of 10.7% and ranges from 4.95% to 15.1%. It seems that all of the banks in the sample met Basel III capital requirements i.e Tier 1 >4.5%. Liquidity measured as the ratio of liquid assets total assets have a positive mean of 0.046 and a maximum value

5_{As mentioned in the previous section, the dependent variable will consist the highest abnormal returns}

(21)

of 0.40 which entails that on average most of the banks have a satisfying value of liquid assets as a percentage of total assets. Asset Quality estimated as loan loss provision to net interest income indicates that loan provisions cover on average almost 47% of net interest income. Earnings measured as dividing earnings before income, taxes and deprecation over total assets on average is positive at 1.1% while the minimum a negative 2.2% and the maximum 3.2%. Size measured as the logarithm of total assets on average is fairly positive. Lastly, fragility measured as non-performing loans to total assets on average is positive at 10.9% indicating a moderate value of non-performing loans.

4.6 Correlation matrix

Table 3 reports the correlation between the dependent variables and independent ones explained in the previous section. Overall, the correlation of the variables seems to support Hypothesis 3, that the QE aided more financially weaker banks. For instance, there is a positive correlation between asset adequacy and abnormal returns which is expected although with a smaller magnitude of 15%. Likewise, liquidity and earnings have a negative correlation with the dependent variable with magnitudes of -2% and -19% respectively, hence supporting Hypothesis 3. Nevertheless, the magnitude of liquidity is quite low since a bank with a high level of liquidity is considered to have a strong financial condition. As expected, the variable, fragility displays a positive correlation with abnormal returns in spite of a low magnitude of 9%. Lastly, the variables size and asset quality do not have the expected direction but a positive correlation with the dependent variable.

Referring to the correlation between the dependent variables, to be noted is the highest negative correlation between earnings and asset quality of -81%. Considering the fact that asset quality is estimated by the ratio of provision for loan losses to net interest income the directions is economically expected since a high value for provision losses will result in less earnings for a bank. To be also noted is the negative correlation of -72.8% between the variable fragility and earnings. A bank with a high number of performing loans as a frac-tion intuitively cannot experience high earnings. Hence, this correlafrac-tion can be supported

(22)

economically. Surprisingly enough, the size of a bank has a negative correlation with the earrings although the magnitude is quite small, -6.2%.

5 Empirical Results

This section presents a description of the key results. To begin, the estimates of the event study of the Fed announcements on the sample of banks is presented. Followed by, the results of the cross-sectional regression based on six bank specific indicators.

5.1 Event study results: The impact of Fed announcement on financial

intermediaries

Following the theory of Mackinlay (1997), this research estimates abnormal returns and cu-mulative abnormal returns by using a market model approach and a two day event window (0;+1). The event dates used for the event study refer to the three main announcements of the US QE periods: QE1, QE2 and QE3. On November 25th 2008 (QE1), Fed announced for the first time the QE program by stating that $500 billion of Agency Mortgage-Backed-Securities (MBS) issued by Government Sponsored Enterprises (GSEs) and $100 billion of Agency Debt will be purchased from the private sector.6 _{Almost two years later after the} initial QE announcement, on October 15th 2010 (QE2) Fed chairman Ben Bernanke stated that additional QE purchases are needed in order to further stimulate the outlook of the economy. In his speech he said the following: “We will continue to monitor economic de-velopments closely and to evaluate whether additional monetary easing would be beneficial. In particular, the Committee is prepared to provide additional monetary accommodation through unconventional measures if it proves necessary, especially if the outlook were to de-teriorate significantly.”7Following this announcement, Fed began purchasing $600 billion of US long-term treasuries. On the 22nd of August 2012, the Federal Open Market Committee (FOMC) announced QE3 by specifying that “additional monetary accommodation would be

6_{https://www.federalreserve.gov/newsevents/press/monetary/20081125b.htm} 7_{https://www.federalreserve.gov/newsevents/speech/bernanke20100827a.htm}

(23)

likely warranted soon.”8

The results of the event study i.e abnormal returns and cumulative abnormal returns are presented in Table 4, Table 4A and Table 5, Table 5A respectively. In addition, actual returns for the sample of banks are presented in Table 6 The first column of Table 4 and 4A reports abnormal returns on the announcement date of QE1 where thirty-eight finan-cial intermediaries experienced positive abnormal returns while only eleven negative ones. Out of the sample of financial intermediaries, thirtyone banks have statistically significant positive abnormal returns at a 1%, 5% and 10% confidence interval. However this is not the case for negative abnormal returns since the majority is statistically insignificant. Ab-normal returns for the QE1 announcement, range from the lowest -12% (First Bancorp) to the three highest of 49.1% (Comerica Bank), 40.1% (Fannie Mae), 29.2% (Keycorp) and 20.5% (Freddie Mac). It seems that the first announcement of QE which mostly comprised of MBS purchases resulted in significantly higher abnormal returns for the Government Sponsored Enterprises. These results indicate that both Fannie Mae and Freddie Mac are two of the intermediaries who benefit the most out of the initial QE announcement. Even though some banks experienced negative abnormal returns; overall it seems that the ini-tial announcement of QE has a strong positive impact on banks equity value. Comparing abnormal returns with actual returns on the date of QE1 announcement, one can observe positive actual returns for the majority of the banks (only six banks have negative actual returns). Since the majority of the banks have positive actual returns, one might expect positive abnormal returns for the first announcement. The results of QE1 announcement, are consistent with Joyce et al (2011) and Koabayashi et al. (2006) stating that the market predicted a positive impact of the QE program. Additionally these results support Hypoth-esis 1 since the majority of abnormal returns resulted positive however this is not the case for the announcement of QE2 and QE3. Table 7 reports the significance test across all banks for the QE1 announcement. Across all banks abnormal returns and cumulative abnormal returns following QE1 are significant at 10% and 1% level respectively. Overall, the QE1

8

(24)

announcement creates 2.8% abnormal returns and 3.6% cumulative abnormal returns across all banks in the sample. Therefore, the first QE announcement supportsHypothesis 1 since both abnormal and cumulative abnormal returns are positive and statistically significant.

Referring to column two and three of Table 4 and 4A different results in magnitude and direction are presented for the introduction of QE2 and QE3 where the majority of abnormal returns are negative and the magnitudes are quite low compared with QE1. Moreover, less significant results are reported in these two columns. The same pattern of returns is observed for bank’s actual returns on the announcement of QE2 and QE3 where the majority of the returns are negative. These results are consistent with Hypothesis 2 stating that higher abnormal returns are expected for the initial stage of the US QE compared to subsequent ones. Meaning and Zhu (2011) argue that the surprise element of the program disappears for subsequent periods which might lead to lower and negative abnormal returns for the sample financial intermediaries. Moreover, during the period of QE2 and QE3 the US financial market had gone through considerable volatility which may affect the negative direction of actual and abnormal returns. Additionally, Krishnamurthy and Vissing-Jorgensen (2011) assert that the magnitude of the QE effect differed for different periods depending on the composition of asset purchased and over maturities. For instance, the composition of asset purchased during QE1 compared to the composition of QE2 and QE3 have a higher impact on financial intermediaries which is reflected in strong positive abnormal returns. The results of QE2 and QE3 contradict the findings of Joyce et al (2011), Krishnamurthy and Vissing-Jorgensen (2011) and Eser and Schwab (2013), stating that the LSAP program will result in positive abnormal returns.

Referring to cumulative abnormal returns, the sample of US banks experienced similar results in magnitude and directions as abnormal returns. For example, during the announce-ment of QE1 most of the banks witnessed positive cumulative abnormal returns (36 out of 48 banks). For the announcement of QE2 and QE3, a higher number of banks resulted in negative cumulative abnormal returns compared with the initial announcement. Also to be highlighted is the fact that the announcement of QE1 has higher cumulative results

(25)

com-pared with the subsequent announcements which reflect the positive reaction of the market. During the announcement of QE1, the highest cumulative abnormal returns refer to 77% (Freddie Mac), 45% (MGIC Investments) and 38% (Fannie Mae), significant at 1% and 5% confidence interval. Again the Government Sponsored Enterprises seem to benefit the most. However, the actual returns of Freddie Mac and Fannie Mae on the announcement date of QE1 are the highest out of the sample banks; 17.78% and 38.24% respectively.

Altogether the event study results indicate that the QE announcement on all three periods has a disproportional impact manifested in some banks having higher, positive and negative abnormal returns. The next section, by conducting a cross-sectional regression on bank characteristics will shed light on this disproportional impact

5.1.1 Cross-sectional estimates

Table 8 presents the results of the cross-sectional regression based on the six bank spe-cific indicators. As mentioned in the previous section, the dependent variable consists of abnormal returns on the 25th of November 2008 (QE1).

As outlined in the data description, the independent variables fail to be significant due to a low number of observations. Nevertheless, it is important to provide an economic interpretation. The estimators of capital adequacy, asset quality, profitability and size display the expected direction. It seems that, earnings have the highest negative coefficient impacting abnormal returns. While for asset quality and profitability this is not the case, considering their low coefficients. As expected, asset quality measured as the ratio between loan losses provisions to net interest income, has a positive relation with the dependent variable. The positive correlation indicates that financial intermediaries with a high ratio of loan loss provision to net interest income benefit more from the program.

Moreover, results suggest that well capitalized banks and banks with higher liquidity and higher larger banks earn less abnormal returns during the first QE announcement. Conventional wisdom in banking suggests that the higher the liquidity, profitability and size of a bank the more financially strong a bank is. Hence, these estimators tend to support

(26)

Hypothesis 3. The study of Wheelock and Wilson (2000) provide evidence that financial intermediaries with lower asset quality are more prone to fail. Additionally, they find statistically significant results that profitability is negatively correlated with the likelihood of a bank failure. On the other hand Pasiouras et al. (2006) argue that profitability of a bank is a key indicator of the longterm success or failure of the bank, emphasizing the positive relation with financial stability and strength. Their arguments are supported by Poon and Firth (2005) who find that profitability is positively correlated with higher ratings, implying a financially strong bank.

Considering the size indicator, Falkenstein et al. (2000) find that banks size is an important component of a bank failure and financial stability. Consequently, a smaller bank can be less diversified compared to a bigger one. It follows that less diversification leads to higher chances a bank suffers from idiosyncratic shocks. Lennox (1999), argues that size of a bank matters in accessing external financing. In doing so, large banks are less likely to fail since they face less credit constraint in case of financial distress.

Different signs to what is expected resulted for the estimators of liquidity and fragility. For instance, liquidity is expected to have a negative relationship with abnormal returns since a higher ratio of liquid assets to total assets entails a bank less exposed to liquidity risk and less likely to fail. Therefore the higher the value of liquid assets a bank has, the more robust its financial state. The positive sign on liquidity suggest that more liquid banks are expected to benefit more from the QE program which does not support Hypothesis 3. However, Kobayashi et al. (2006) find similar results where liquidity is positively correlated with abnormal returns, although it is not expected as in this research.

Similarly the variable of fragility estimated through the ratio of non-performing loans to total assets result in a negative sign. As mentioned in the previous paragraph, this variable is expected to have a positive relation with the dependent variable since a high value of non-performing loans to total assets portrays a financially weak bank. Hence, in order to support Hypothesis 3, a bank having a high ratio of non-performing loans is expected to benefit more from the QE program which is not the case. The study of Barr and Siems

(27)

(1994) found that a high level of non-performing loans plays a crucial role in determining a failing bank. Their findings are supported by Demirguc-Kunt (1989) who also find that the quality of a banks loans is an indicator for insolvency.

Altogether the estimators of capital adequacy, asset quality, profitability and size are adequate estimators for the dependent variable since their sing is supporting Hypothesis 3. However, the estimators of financial state and liquidity fail to support that the QE program benefited financially weaker banks.

6 Robustness Checks

This section comprises the robustness checks for the event study where a three day event window (-1;+1) is going to be utilized.

6.1 6.2 Three-day event window (-1;+1)

Joyce et al. (2011) in estimating the Bank of England QE program effect on UK gilts uses a three day (-1;+1) event window as a robustness check. This approach is supported by MacKinlay (1997) who proposes a three day event window (-1;+1) as the most precise window for the reason that it enables any indirect impact to take place during the days around the event as well it does not lose the regression effectiveness. Hence this research uses a three day interval as a robustness check. Additionally, a three day event window tests how persistent abnormal and cumulative abnormal returns in comparison with a two-day event window used in the previous section.

Table 9, 9A and Table 10, 10A presents abnormal returns and cumulative abnormal returns using a three day interval. Comparing the event study results while using different windows, it seems that using a wider interval results in lower estimates for abnormal returns. For instance, referring to the three QE announcements the majority of the banks (26 out of 51) included in the sample experienced a decrease in abnormal returns compared with the abnormal returns when using a two day window. Even when banks experienced negative abnormal returns, using a wider window resulted in a decrease in magnitude. Nevertheless

(28)

the decrease is not that pronounced and does not lead to a huge difference in magnitude. One must also consider that only a few banks experience a change in direction for abnor-mal returns i.e from positive to negative. This change occurs when abnorabnor-mal returns are positively low and change to negative when applying a wider window. Again the change is not that significant.

Referring to the first QE1 announcement 29 banks have a positive abnormal return which is almost the same number when compared with the results of a two-day event window.9 This is not the case for QE2 and QE3 announcement where most of the banks have negative abnormal returns which are quite similar with the previous estimates.

Similar changes are observed for cumulative abnormal returns for the three QE an-nouncements. For example, the QE1 announcement results in the highest positive cumu-lative abnormal returns compared with subsequent announcement but if contrasted with a two-day event window the magnitude is lower and in this case the decrease is more pro-nounced.

The overall effect of using a wider window results in a decrease for both abnormal returns and cumulative abnormal returns. However, the decline is not that noticeable for abnormal returns but for cumulative abnormal returns more notable. Nevertheless, the estimates of a three-day event window can be considered consistent with the two-event window estimates since the changes are not that significant for abnormal returns. Generally, increasing the event window leads additional factors, other than the QE announcement to “pollute” the event study results. For example, other events may occur which may impact the returns on the sample of banks. The occurrence of other events can be an explanation for the reduction in the estimates when using a wider window. Moreover, Joyce et. al (2010) found a similar reduction in results of the UK gilts when applying a wider event window.

(29)

7 Conclusion and Discussion

This research examines the impact of the US QE announcements on the equity value for a sample of financial intermediaries by conducting an event study. To address this topic, abnormal and cumulative abnormal returns are investigated for the sample of US banks around the three main announcements: QE1, QE2 and QE3. Subsequently, a cross-sectional regression based on the CAMEL model bank indicators is executed in order to shed light if the QE program disproportionally impacts financially weaker banks.

The event study provides evidence that the first QE announcement results in significant positive abnormal and cumulative abnormal returns for the majority of the banks in the sample. Moreover, across all banks in the sample abnormal and cumulative abnormal returns are statistically significant at 1% and 10% respectively. The QE1 announcement creates 2.8% abnormal returns and 3.6% cumulative abnormal returns across all banks. Therefore, the first QE announcement supports Hypothesis 1 since both abnormal and cumulative abnormal returns are positive and statistically significant.

However, for the QE2 and QE3 announcements the majority of banks experienced neg-ative abnormal returns. An explanation for these negneg-ative estimates might be that the financial market is skeptical about the effectiveness of the first QE round and additional stimulus packages are needed to boost the economy. To add, the US financial market during the QE period experienced considerable fluctuations which can explain negative abnormal returns. The results of QE2 and QE3 do supportHypothesis 2 since the first QE impact is stronger compared to subsequent stages. Meaning and Zhu (2011) argue that the surprise element of the program disappears for subsequent periods which might lead to lower and negative abnormal returns for the sample of financial intermediaries. Additionally, Krish-namurthy and Vissing-Jorgensen (2011) assert that the magnitude of the QE effect differed for different periods depending on the composition of asset purchased and over maturities. For instance, the composition of asset purchased during QE1 compared to the composition of QE2 and QE3 has a higher impact on financial intermediaries which is reflected in strong positive abnormal returns. The first QE program is composed mostly of MBS purchases

(30)

issued by Government Sponsored Entities (GSEs) such as Fannie Mae and Freddie Mac. Abnormal returns for QE1 announcement range from the lowest -12% (First Bancorp) to highest of 49.1% (Comerica Bank), 40.1% (Fannie Mae), 29.2% (Keycorp) and 20.5% (Fred-die Mac). These results show that the first announcement has a positive significant impact on the two GSEs and are two of the intermediaries who benefited the most from the initial QE announcement.

Overall, the event study results indicate that initial QE announcement has a positive impact on banks equity value. These results are consistent with the findings of Joyce et al. (2011) and Koabayashi et al. (2006) stating that the market predicted a positive impact of the QE program. Yet, the results of QE2 and QE3 announcements are not in line with these findings since most of the banks experienced negative returns.

Referring to the cross-sectional evidence, only the indicators of asset quality, capital adequacy, profitability and size tend to support Hypothesis 3. The sign of these indicators support the preposition that that the QE program benefits most financially weak banks. For instance, capital adequacy estimated as Tier 1 capital ratio results in a negative correlation with the dependent variable, abnormal returns. Likewise, profitability and size result in a negative correlation with abnormal returns. A well-capitalized and a more profitable bank is in a better financial condition, hence a negative relationship with the dependent variable is expected in order to support Hypothesis 3. Asset quality measured as the ratio of loan loss provision to net interest income results in a positive correlation with abnormal returns which is expected. A bank with a high ratio of loan loss provision to net interest income portrays a weak financial state and thus, is expected to benefit more for the QE program. The above mentioned indicators suggest that well capitalized, more profitable and larger banks earn less abnormal returns during the first QE announcement.

Liquidity and fragility, the last two indicators of the cross-sectional regression do not have the expected sign in order to support Hypothesis 3. For example, liquidity measured as the ratio of liquid assets to total assets is expected to have a negative relationship with the dependent variable since a bank with a high level of liquid asset reflects a strong

(31)

financial position. Nevertheless, the study of Kobayashi et al. (2006) finds similar results for liquidity where it is positively correlated with abnormal returns, which is not expected. This research includes fragility as an extra variable in the cross-sectional regression in order to give supplementary evidence for a financially weak bank. Fragility is estimated as the ratio of non-performing loans to total assets and is intended to capture the weakness of a bank arising from its core assets. Since a high number of non-performing loans indicates a poor financial condition for a bank, the variable is expected to be positively correlated with abnormal returns, which is not the case.

Overall, the cross-sectional regression suffers from data restriction which affects the significance of the indicators as well as the expected sign. Out of fifty-one US banks included in the sample, this research could access balance-sheet data for only twenty-seven. This low number of observation does not have enough explanatory power and significance level in order to provide remarkable results. Moreover, different ratios for capital adequacy, liquidity, asset quality and profitability as the one chosen in this research might have more explanatory power and may lead to more significant results. Nonetheless, the cross-sectional results of this research provide considerable evidence to support preposition that the QE program has a disproportional impact on US banks.

As a concluding paragraph, the event study results showed that the impact of the initial QE announcement has a higher positive impact on the sample of banks, compared to subsequent announcements. On the one hand, these results demonstrate that the first QE announcement is deemed effective from the financial market since it creates positive cumulative and abnormal returns. On the other hand, the financial market is skeptical about the effectiveness of subsequent QE periods and additional stimulus packages are needed to stimulate the economy. Lastly, the cross-sectional indicators of asset quality, capital adequacy, profitability and size do support the hypothesis that the US QE benefited more financially weak banks. However, this indication is not fully explanatory since the significance level is missing.

(32)

References

Basel Committee. (2004). International convergence of capital measurement and capital standards: A revised framework. Basel Committee Publications, No. 107.

Barr, R., Siems, T. (1994), Predicting Bank Failure using DEA to quantify management quality. Federal Reserve Bank of Dallas, Financial industry Studies Working papers N94. Berger, A. N., Klapper, L. F., & Turk-Ariss, R. (2009). Bank competition and financial stability. Journal of Financial Services Research, 35(2), 99-118.

Bernanke, B., & Reinhart, R. (2004). Conducting monetary policy at very low short-term interest rates. American Economic Review, 94(2), 85-90.

Bernanke, Ben, Vincent Reinhart, and Brian Sack. ”Monetary policy alternatives at the zero bound: An empirical assessment.”Brookings papers on economic activity, 2004.2 (2004): 1-100.

Bowman, D., Cai, F., Davies, S., & Kamin, S. (2011). Quantitative easing and bank lending: evidence from Japan.International Financial Discussion Paper, n. 1018.

Christensen, J.H.E. and Rudebusch, G.D. (2012). The response of government yields to central bank purchases of long-term bonds. Economic Journal, vol. 122(564), pp. F385414. D’Amico, S., and King, T. (2010), Flow and Stock Effects of Large-scale Treasury Purchases, Finance and Economics Discussion Series 201052, 141.

Darracq-Paries, M., & De Santis, R. A. (2013). A non-standard monetary policy shock: the ECBs 3-year LTROs and the shift in credit supply,ECB Working Paper Series, n. 1508. De Jonghe, O. (2010). Back to the basics in banking? a micro-analysis of banking system stability. Journal of financial intermediation,19(3), 387-417.

Demirg-Kunt, A. (1989), Deposit-institution failures: a review of empirical literature. Eco-nomic Review, Federal Reserve Bank of Cleveland, 4, 2-18.

(33)

Eser, F., & Schwaab, B. (2013). Assessing asset purchases within the ECBs securities markets programme, ECB Working Paper Series, n. 1587.

Falagiarda, M., & Reitz, S. (2015). Announcements of ECB unconventional programs: Im-plications for the sovereign spreads of stressed euro area countries. Journal of International Money and Finance,53, 276-295.

Falkenstein E, Boral A, Carty LV (2000) RiskCalc for private companies: Modeling method-ology. MoodysKMV Company

Fawley, B. W., & Neely, C. J. (2013). Four stories of quantitative easing.Federal Reserve Bank of St. Louis Review, 95, (January/February 2013).

Gagnon, J., Ruskin, M., Remake, J., & Sack, B. (2011). The financial market effects of the Federal Reserves large-scale asset purchases. International Journal of Central Banking, 7(1), 3-43.

Gambacorta, L., Hofmann, B., & Peersman, G. (2014). The Effectiveness of Unconventional Monetary Policy at the Zero Lower Bound: A CrossCountry Analysis.. Journal of Money, Credit and Banking,46(4), 615-642.

Giannone, D., Lenza, M., Pill, H., & Reichlin, L. (2012). The ECB and the Interbank Market*.. The Economic Journal,122 (564), F467-F486.

Hancock, D., and Passmore, W. (2011), Did the Federal Reserves MBS Purchase Program LowerMortgage Rates?, . Finance and Economics Discussion Series 2011-01.

Joyce, M., Lasaosa, A., Stevens, I., & Tong, M. (2011). The financial market impact of quantitative easing in the United Kingdom. International Journal of Central Banking, 7(3), 113-161.

Kapetanios, G., Mumtaz, H., Stevens, I., & Theodoridis, K. (2012). Assessing the econo-mywide effects of quantitative easing*. The Economic Journal, 122 (564), F316-F347.

(34)

Kothari, S. P., & Warner, J. B. (2004). The econometrics of event studies.Available at . SSRN 608601.

Kobayashi, T., Spiegel, M. M., & Yamori, N. (2006). Quantitative easing and Japanese bank equity values. . Journal of the Japanese and International Economies, 20(4), 699-721. Krishnamurthy, A., & Vissing-Jorgensen, A. (2011). The effects of quantitative easing on interest rates: channels and implications for policy. National Bureau of Economic Research, No. w17555

Lennox, C. (1999). Identifying failing companies: a re-evaluation of the logit, probit and DA approaches. Journal of Economics and Business, 51(4), 347-364.

Lyonnet, V., & Werner, R. (2012). Lessons from the Bank of England on quantitative easing and other unconventionalmonetary policies. . International Review of Financial Analysis, 25, 94-105.

Malkiel, B. G., & Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. . The journal of Finance, 25(2), 383-417.

Meaning, Jack, and Feng Zhu. The impact of recent central bank asset purchase pro-grammes. . BIS Quarterly Review, December (2011).

Mishkin, F. S. (1996). The channels of monetary transmission: lessons for monetary policy. . National Bureau of Economic Research, No. w5464.

Modigliani, Franco, and Richard Sutch. ”Innovations in interest rate policy.” . The Amer-ican Economic Review 56.1/2 (1966): 178-197.

Neely, C. (2010), The Large-scale Asset Purchases Had Large International Effects, Federal Reserve Bank of St Louis Working Paper 2010018, 145.

Pasiouras, F., Gaganis, C., & Zopounidis, C. (2006). The impact of bank regulations, supervision, market structure, and bank characteristics on individual bank ratings: A cross-country analysis. . Review of Quantitative Finance and Accounting, 27(4), 403-438.

(35)

Poon WPH, Firth M (2005) Are unsolicited credit ratings lower? International evidence from bank ratings. . Journal of Business Finance and Accounting. 32(9&10):17411771 Stroebel, J. C., & Taylor, J. B. (2009). Estimated impact of the Feds mortgage-backed securities purchase program. . National Bureau of Economic Research, No. w15626 Swanson, E. T., Reichlin, L., & Wright, J. H. (2011). Let’s Twist Again: A High-Frequency Event-Study Analysis of Operation Twist and Its Implications for QE2 [with Comments and Discussion].. Brookings Papers on Economic Activity, 151-207.

Tregenna, F. (2009). The fat years: the structure and profitability of the US banking sector in the pre-crisis period.. Cambridge Journal of Economics,33(4), 609-632.

Ugai, H. (2007), Effects of the Quantitative Easing Policy: A Survey of Empirical Analyses, . Monetary and Economic Studies, 25(1), Institute for Monetary and Economic Studies, Bank of Japan, 148.

Wheelock, D. C., & Wilson, P. W. (2000). Why do banks disappear? The determinants of US bank failures and acquisitions. . Review of Economics and Statistics, 82(1), 127-138.

(36)

(37)

(38)

(39)

(40)

(41)

(42)

(43)

(44)

(45)

(46)

(47)

(48)

Quantitative easing and financial intermediaries : the impact of Federal Reserve's QE announcements on US