The impact of uconventional monetary policy on banking and bovereign default risk

(1)

The Impact of Unconventional Monetary Policy

on Banking and Sovereign Default Risk

Master Thesis

Student Name Sascha Bühler

Thesis Supervisor Dr. Tanju Yorulmazer

Faculty of Economics and Business MSc Finance: Banking and Regulation Academic Year 2016-2017

(2)

I

Statement of Originality

This document is written by Student Sascha Bühler 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.

(3)

II

Acknowledgement

The process of writing my Master Thesis would have been much more difficult without the profes-sional and personal support and advice of many people. I would like to take this opportunity to ex-press my gratitude to them.

First and foremost, I would like to express my sincere gratitude to my supervisor, Dr. Tanju Yorulmazer for his great support of my studies and the many valuable comments and discussions we shared. He has not only been a great academic advisor to me but also enriched my personal develop-ment. I particularly enjoyed his Banking lecture and our time working together on my thesis.

I would like to thank my colleagues of the Data and Reporting Advisory team at Deloitte who created a collaborative and stimulating work environment, notably Stijn Roersch, Daan van den Boogaard, Dirk Boersma, Pieter Jaap den Haan and many more.

I have also been blessed with good friendships I made during my time in Amsterdam that greatly facilitated my studies and Master Thesis. I keep the “German-Swiss in Amsterdam” in very good memories.

My deepest gratitude to my parents and brother, for their continuous backing of my studies and for offering me the freedom to pursue my interests throughout my life.

(4)

III Abstract

This study examines the effect of unconventional monetary policy by the Federal Reserve and Euro-pean Central Bank on bank and sovereign credit default swaps in the aftermath of the financial crisis. The paper explains the fundamentals of monetary policy and discusses the transmission channels through which the real economy is affected. Using the event study methodology, I find results which indicate that the communication of unconventional monetary policy by central banks contain valua-ble information with substantial effects around announcement dates. The results suggest that the Fed-eral Reserve and European Central Bank communications significantly reduce bank and sovereign default risk measured by credit default swap spreads. Nevertheless, contrary results are also observed when announcements reveal a bad economic state and outlook. Furthermore, the market responds stronger and clearer to European Central Bank compared to Federal Reserve communications.

Keywords: central bank communication, unconventional monetary policy, transmission channels, default risk, credit default swaps, event study analysis

(5)

IV

List of Abbreviations

ABS Asset-backed securities ASC Abnormal Spread Change

CAASC Cumulative Average Abnormal Spread Change CASC Cumulative Abnormal Spread Change

CBPP Covered Bonds Purchase Program CDS Credit Default Swap

ECB European Central Bank EST Eastern Standard Time Fed Federal Reserve

GDP Gross Domestic Product LSAP Large Scale Asset Purchase MBS Mortgage-backed securities OLS Ordinary Least Squares

OMT Outright Monetary Transactions OTC Over-The-Counter

PIIGS Portugal, Italy, Ireland, Greece, Spain PSPP Public Sector Purchase Program QE Quantitative Easing

(7)

1

1. Introduction

The financial crisis began in 2007, infected the entire financial system and ended in an economic bust across advanced economies. Financial institutions lost confidence among each other which led to a credit squeeze for companies. This initiated a negative spillover from the financial sector to the real economy. Governments, henceforth referred to as sovereigns and central banks stepped in as a safeguard for fi-nancial stability and economic activity and conducted unprecedented monetary and fiscal actions to evade recession. The interventions inflated the balance sheets of central banks and sovereigns, and hence higher debt ratios, which raised concerns about sovereigns’ solvency. Consequently, credit ratings de-teriorated and funding costs increased for several countries. Sovereign risk had further adverse effects on financial markets. The fragile banking sector therefore got affected since it relies on governmental guarantees to a certain level. Additionally, banks were exposed to the sovereign sector which further stressed banks’ solvency (Haldane et al. 2016, Panetta et al. 2011). Ang and Longstaff (2013) found increasing default intensities for sovereigns during the financial crisis. Reasons include the increasing liabilities through supporting failing institutions, the weakening of the economic situation and the insta-bility of the financial sector.

The main conventional instrument central banks have at their disposal is the adjustment of interest rates. Central banks raise interest rates when the economy overheats during a boom and lower interest rates when they want to stimulate the economy in time of a bust. By reaching the limit of interest rates modifications during the financial crisis, central banks moved forward to unconventional monetary pol-icy often cited as quantitative easing (QE)1. The main goals were price stability, to ease the financial sector and to stimulate economic growth. Central banks tried to calm markets through communications about intended monetary operations and revelation of future expectations. The Federal Reserve (Fed) introduced QE in 2008, when interest rates reached the effective lower bound (Haldane et al. 2016). The European Central Bank (ECB) realized the limitations of interest rate measures in 2009, started to con-sider unconventional monetary policy instruments and gradually expanded its QE (Andrade et al. 2016).

Research was conducted on unconventional monetary policy by the Federal Reserve Bank, the Bank of England and the European Central Bank after the financial crisis. The event study of Gagnon et al. (2011) explained the effects of Fed announcements on various domestic securities. They found reductions in long term yields, even for securities which were not purchased. Neely (2015) and Krish-namurthy and Vissing-Jorgensen (2011) confirmed these results and observed a cut in interest rates of long term assets around Fed communications. Joyce et al. (2011) identified similar results for Bank of England announcements on UK asset prices. Falagiarda and Reitz (2015) found reduced spreads be-tween stressed European countries relative to Germany around ECB communications. Furthermore,

(8)

2

studies by Ang and Longstaff (2013) and Fender et al. (2011) discovered spillover effects of central bank communications to other economies. In academia, a broad and affirmative consensus exists on the effectiveness of QE and its supporting effect to the economy. Despite endorsing findings, there is still an ongoing public discussion about the effectiveness and opaqueness of QE, especially in Europe2. For the Euro zone, the discussion may be more complex and advanced since it is a currency union with strong economic dependencies, obligations and diversity among member countries. On one side, there were supporting effects for financial markets, the economy and price stability. On the other hand, con-cerns rose that especially stressed governments could cheaply refinance rather than reform themselves. While this may have temporarily shifted problems, it is likely to cause trouble and to increase default risk in the medium or long term due to excess indebtedness and potentially higher future interest rates. This raises the first research question:

What is unconventional monetary policy and how does it affect default risk of banks and sovereigns?

Most studies are focused on domestic effects of monetary policy and on the most common secu-rities such as stocks, government and corporate bonds. I will analyze a broader and more interactive effect of Fed and ECB communications on US and European entities from the beginning of the crisis till present. The market turmoil and rising distrust about creditworthiness combined with the predeter-mined aims of central banks, serve as the rationale behind this research. My study evaluates the rela-tionship of unconventional monetary policy communications of central banks with bank and sovereign default risk, synthesized by credit default swaps (CDS). A credit default swap is a derivative instrument that allows to hedge against default risk of a reference entity, such as sovereigns or financial institutions. Since the banking and sovereign sector is highly interconnected, I focus on default risk of both due to adjustments of risk perception by market participants. This leads to the second research question: What is the effect of unconventional monetary policy communications by the Federal Reserve and

Eu-ropean Central Bank on bank and sovereign default risk?

In order to answer these questions, I analyze existing literature and follow the event study ap-proach to show that central bank communications have significant impact on CDS spreads of several US and European samples3. First, I test individual events for each group of sovereign and bank entities. Second, I test the groups for aggregated Fed and ECB announcements. Additionally, I regress on abnor-mal spread changes with several independent market variables to explain the abnorabnor-mal reaction around central bank communications.

2_{Fortune (2017), Welt (2017) and Tagesspiegel (2017)}

3_{Banks: US, PIIGS, core Euro or non-PIIGS respectively and non-Euro; Sovereigns: US, PIIGS and core Euro}

or non-PIIGS respectively. I use the term core Euro banks and sovereigns instead of non-PIIGS in the reminder of this paper. A detailed discussion follows in Section 5.

(9)

3

The remainder of the thesis is organized as follows. The subsequent section contains the theoret-ical background about unconventional monetary policy, how it differs from regular monetary policy and through which channels the real economy is affected. It further discusses existing literature related to QE and findings about the impact on various asset prices. I also introduce CDS and what they reveal about default risk. Based on this knowledge, I formulate two hypotheses in this section. Section 3 de-scribes the event study methodology and assumptions which are applied to answer the hypotheses. Sec-tion 4 outlines the data and examined events. In SecSec-tion 5, I present the results of the analysis, discuss the economical meaning and provide robustness checks to further strengthen the results and underlying assumptions. Section 6 presents my conclusion.

(10)

4

2. Literature Review

Extensive research is done about monetary policy and its impact on different assets such as government bonds, corporate bonds and stock prices. However, there exist not many articles and findings towards the effect of recent post crisis unconventional monetary policy communication on default risk measured with CDS spreads. This section introduces the fundamentals of monetary policy, the channels to the economy, credit default swaps and how they serve as a proxy for default risk.

2.1 Conventional Monetary Policy

Central banks play an important role in financial markets under governmental authority. Central banks’ actions, namely ‘monetary policy’ affect interest rates, money supply and amount of credit. The opera-tions impact not only financial markets but also aggregated output and inflation. During times of sound economic conditions central banks manage their activities mainly through the short term interest rate (Mishkin 2007).

As mentioned previously, conventional tools were no longer considered efficient enough after the breakout of the financial crisis. Therefore, other measures became essential. Which measures and how to distinguish them from the conventional ones is part of the following discussion. This section provides concepts, goals and mechanisms behind conventional and unconventional monetary policy rather than details of each program conducted by central banks. I also exclude the response from governments and their fiscal and regulatory actions in my analysis. Finally, it shows how monetary policy actions affect and stabilize the real economy.

Mishkin (2007) summarized the main goals of national banks as economic growth, low unem-ployment and stability of price, interest rate, financial market and foreign exchange. Economic growth and stability of financial markets play an important role to assess the scope of this research. Joyce et al. (2012) states that the recent financial crisis deteriorated financial market conditions and its ability to provide funds to people and business with good investment opportunities. This led to an even more severe contraction in the whole economic activity. Steady economic growth can be stimulated through investment and consumption by either providing tax incentives from fiscal authorities or a more favor-able funding rate offered by central banks.

Central banks cannot directly influence their own goals but rather use certain tools such as changes in discount rates, reserve requirements and open market operations. The quantities of such open market transactions were seen as almost insignificant before the crisis. Central banks solely changed reserve requirements and interest rates for banks in order to provide and reduce money supply. Central banks facilitate or constrain lending to banks by adjusting the interest rate at which they are willing to lend. Credit facility is another source to provide liquidity for financial institutions. Bond purchases and sales are determined as open market operations to adjust the monetary base. As a result of open market

(11)

5

purchases, reserves expand and the short term funding rate falls; while for sales, the short term funding rate increases.

The Fed for instance, executes these transactions in US Treasury and government agency securi-ties because this market is the most liquid and deep, therefore prices do not get excessively distorted. Mishkin (2007) and Joyce et al. (2012) described conventional monetary policy as the effect on short term interest rates by central banks, which purchase and sell securities from the banking system and therefore affect bank reserves. Under regular market conditions, the ECB (2017a) uses reserve require-ments and deposit facilities as conventional monetary policy tools. Reserve requirerequire-ments manage the liquidity and influence interest rates in the money market. By changing conditions under which the central bank is willing to enter transactions with financial institutions, it further steers interest rates by handling liquidity. Bernanke and Gretler (1995) explained that monetary policy affects market condi-tions such as interest rates and the financial position of borrowers rather than economic variables.

Macroeconomic variables such as inflation and output tend to respond only after a certain time period of roughly one to two years. Therefore, central banks work with forward looking operating tar-gets. Operating targets quickly respond and can be adjusted if not right on track. Central banks define target operations or instruments4_{to influence the amount of money in circulation or interest rates and}

guide them towards the desired level. The Feds pre-crisis monetary policy is described as successful even without using nominal anchors such as exchange rates, monetary aggregate or inflation. The Fed rather followed a ‘just do it’ policy which clearly deviates from policies and targets in the aftermath of the crisis. On one side, this strategy had the disadvantage of being less transparent and leading to uncer-tainty in financial markets. On the other hand, there was no reliance on a stable money-inflation rela-tionship that allowed for domestic consideration to determine the best possible setup (Mishkin 2007).

ECB (2017a) described the pre-crisis role and activities of monetary policy in a very similar way. The central bank influenced money market conditions at which banks trade with each other. A change in interest rates launched several mechanisms and actions by money market participants in the short run. Eventually, it influenced economic variables such as price level in the long run, after adjustments take place.

Smaghi (2009) labeled conventional monetary policy as a steering of overnight interest rates in the interbank money market by adjusting money supply through open market operations. In normal times, this takes place in form of repurchase transactions with collateral. Central banks therefore are neither involved in direct lending nor in outright purchases of debt instruments.

4_{Required bank reserves; rates at which central banks are willing to lend or deposit bank reserves (Fed: Federal}

(12)

6

2.2 Unconventional Monetary Policy

When conventional monetary measures can no longer achieve the central banks’ goals, they need to expand their range of monetary tools. A deeper analysis about unconventional monetary policy is pro-vide in this section. Opening the discussion, I quote Ben Bernanke, former chairman of the Fed. He was asked at a conference if quantitative easing really works. According to the Financial Times (2017), he responded:

“The problem with QE is that it works in practice, but it doesn’t work in theory.”

Despite his statement, I describe the scope of unconventional monetary policy and channel the theory to answer the first research question in this section.

The term ‘quantitative easing’ has its origins in Japan. Since the financial crisis, QE is frequently used in the context of unconventional monetary policy by various central banks. Around the year 2000, nominal interest rates reached a lower bound and the Bank of Japan purchased government bonds to float the market with liquidity. The aim was to stimulate lending, boost asset prices and the economy and put downward pressure on deflation. The term ‘quantitative’ links towards focusing on quantity variables (Joyce et al. 2012).

The Fed was the first central bank adopting QE in the aftermath of the crisis, followed by the Bank of England, the Swiss National Bank and the ECB. QE is commonly seen as a balance sheet expansion while interest rates are at a lower bound of zero. Nevertheless, expansions have been going on as long as we have had central banks. The monetary injections were associated with financing of wars or the bail-out of the banking sector. Since the financial crisis, the balance sheet expansion is seen explicitly as a monetary policy tool. Interest rate adjustments lost its effectiveness after the crisis and central banks approached outright debt purchases funded by the creation of central bank reserves. QE led to a substantial increase of central banks’ balance sheets relative to GDP and debt outstanding by governments (Haldane et al. 2016).

Smaghi (2009) concluded that additional liquidity provided by central banks was sufficient enough to respond to the first signs of the crisis in 2007. In September 2008 and the collapse of Lehman Brothers, everything changed. Conventional tools no longer complied with economic conditions, market liquidity dried up, market participants lost confidence and spreads between short and long term rates increased. In such abnormal times, national banks were driven to start with unconventional monetary policy. The improvement of financing condition beyond short term interest rates could be achieved through influencing long term interest rate expectations and expanding or changing the composition of the balance sheets of central banks. Smaghi (2009) described that “unconventional measures can be

defined as those policies that directly target the cost and availability of external finance to banks, house-holds and non-financial companies. (…) The cost of external finance is generally at a premium over the

(13)

7

short term interbank rate.” QE aims to reduce this spread, thus affecting asset prices and the flow of

funding into the economy. One way of narrowing the spread and reducing cost of credit is by influencing long term interest rates by increasing expected inflation and thereby, real interest rates fall. Another tool to improve market conditions is through sales and purchases of different assets such as government bonds and commercial papers with different maturities.

To continue, Smaghi (2009) stated that central banks have three measures to modify the size of their balance sheet and attempt to fulfill their goals. First, buying long term government assets from banks which reduces yields of privately issued securities and long term interest rates, hence stimulating investment, aggregated demand and price stability. This is called direct quantitative easing. Second, purchases of commercial paper, corporate bonds and asset-backed securities (ABS) directly address li-quidity shortage in the corresponding market segment and institutions. During times of distress in the banking sector direct credit easing, as this measure is called, is even more striking. A third measure is to lend to banks at longer maturities against collaterals including assets of impaired markets. This di-rectly affects the yield curve over the period that the operation is conducted. One can see it as indirect quantitative or credit easing.

Blinder (2010) also described that unconventional instruments helped to reduce interest rates, flattened the yield curve and stimulated aggregated demand. Therefore, these measures were powerful even at the lower bound of nominal interest rates. He emphasized two unconventional tools: first, an “open mouth policy” to credibly commit for lower long term interest rates and thereby boost demand. Second, expand the balance sheet through purchasing risky or less liquid assets, thus reducing liquidity premiums and strengthening the monetary base.

Joyce et al. (2012) agreed that unconventional policy becomes necessary when interest rates hit the lower zero bound and central bank rates are disconnected from money market rates. Therefore, cen-tral banks had to expand their balance sheets by purchasing government and private bonds after the crisis. Central banks matched their greater asset side by increasing reserves held by the banking sector. The conclusion was that QE worked: asset purchases lowered long term interest rates, which had a positive impact on the economy. Another mentioned form is the use of negative interest rates to address inflation targets. Mishkin (2007) emphasized that open market purchases have several advantages in conducting monetary policy in general. First, central banks can control the volume which is not the case with discount rate operations. Second, the operations are flexible and precise. Third, they are easily reversible. And forth, the operations can be implemented very quickly.

(14)

8

2.2.1 Financial Crisis Response by the US Federal Reserve

According to the Fed (2017a) and Blinder (2010), the Federal Reserves’ QE program was to respond to the worst recession since the Great Depression. The program was designed to support liquidity of finan-cial institutions and improve finanfinan-cial market conditions. The main operation involved purchases of long term securities to put downward pressure on long term interest rates and ease credit conditions.

The tools used by the Fed can be divided into provision of short term liquidity, provision of li-quidity directly to borrowers and investors and open market operations. Short term lili-quidity provision was related to a central banks traditional role of providing short term liquidity to financial institutions. Since financial markets are international, the Fed also approved currency swaps with foreign central banks. Direct liquidity provision supported nearly defaulted borrowers and investors in credit markets. Open market asset purchases stimulated credit markets, flattened the interest rate yield curve and stabi-lized financial conditions (Fed 2017a).

Responding to the crisis, the Fed started cutting the Federal Fund Target Rate from 5.25% to 2% between July 2007 and June 2008 and further lowered it 0-0.25% until December 2008. In early 2008, the Fed began selling short term Treasuries and buying less liquid long term assets to reduce the liquidity premium and increase liquidity. This operation twist was netted to zero and did not affect the balance sheet but lowered long term interest rates. In the second stage, the Fed started increasing their assets and liabilities without increasing bank reserves. Additionally, they began to lend to primary dealers after Bear Stearns rescue5. In this early period, QE action was mainly crisis-driven, institution based and ad hoc. In October 2008, after Lehman Brothers failed, the Fed cut interest rates to zero and started ex-panding its balance sheet. A new era had begun. After that, the Feds actions and policy became more systemic, thoughtful and orderly. The Fed expanded its asset side by purchasing billions of dollars in securities such as agency debt, ABS and US Treasuries and ballooned bank reserves on the liability side (Blinder 2010).

Plot 1 on the following page displays the dimension of the Fed total asset development from 2008 till 2016. The balance sheet expanded from around 900 billion USD in 2008 to more than 4 trillion USD in 2014. It gives an indictor about the three different programs and how they subsequently influenced the size of assets. The drivers of the expansion were liquidity facilities and support for specific institu-tions in the beginning and outright purchases of MBS, agency debt and bonds after 2009.

(15)

9

Plot 1. Federal Reserve balance sheet development

Source: Fed (2017b); own input

Krishnamurthy and Vissing-Jorgensen (2011) studied the first two out of three different uncon-ventional programs by the Fed responding to the crisis. QE1 operations contained long term assets pur-chases such as MBS, Treasuries and agency securities, which were conducted after Lehman collapse in the late 2008 until 2009. The purchases significantly reduced yields of the purchased assets and addi-tionally, corporate bonds. Across five events during QE1, the ten year Treasury yield lowered by 107 basis points (bps) and the 15 year MBS yield by 98 bps. Furthermore, QE1 reduced the default risk on high yield corporate bonds and MBS. This program was addressed to support economic activity and the functioning of credit market by liquidity provision. The flight to safety assets and its demand might have reinforced the large effect on interest rate during this period. QE2 started in August 2010 and contained reinvestments of agency debt and MBS principals in long term Treasury. The Fed expanded the program by purchasing long term Treasury starting in November 2010. During this program, Treasury, agency and investment grade corporate bond yields declined moderately and mortgage rates and high yield corporate bonds even remained stable. Across three events of QE2 announcements, the ten year Treasury yield declined by 40 bps. QE1 events showed for all yields a stronger decline, which made QE2 not as successful as QE1 from this perspective. With QE3, the Fed (2017c) wanted to put further pressure on long term interest rates and to raise expected inflation since the economic recovery happened at a mod-erate pace and investments even slowed. In September 2012, the Fed decided to purchase MBS and continue to reinvest principal payments of their holdings in additional MBS.

Additionally, Gagnon et al. (2011) found evidence for the effectiveness of LSAPs, which reduced long term interest rates of various securities and flattened the yield curve. Across eight QE1 and QE2 events between November 2008 and December 2009, the ten year US Treasury yield decreased by 91

(16)

10

bps whereas the long term yields for agency debt and MBS decreased even more with 156 and 113 bps. Based on their findings, one concludes that private borrowing rates consequently lowered, which theo-retically stimulates economic activity. The Feds QE effected mortgages, Treasury securities and corpo-rate bonds which supported the potency of QE at the lower bound of interest corpo-rates. The event study by Neely (2015) confirmed the findings of Gagnon et al. (2011) that QE announcements subsequently re-duced long term US bond yields, long term foreign bond yields and additionally the dollar value. The ten year US Treasury yield fell by 94 bps over eight LSAP buy and sell announcements during QE1 and QE2. Across the same events, the German ten year bond yield lowered by 39 bps and the EURUSD exchange rate by 7.76%. These results suggest that the Fed was not without measures when short term interest rates hit the lower bound. Furthermore, the impact on foreign bonds and the exchange rate likely stimulates the US economy through higher exports. The international spillover is clear evidence that central banks should work together when it comes to such unconventional monetary policy.

The first attempt to exit unconventional monetary policy by the Fed was indicated by Janet Yellen during a speech in October 2014. The US economy was sufficiently strengthened, with unemployment and inflation right on track. Nevertheless, the Fed kept interest rates at a low level until December 2015, when the Federal Fund Rate was increased to 0.25-0.5% (BBC 2017a, Fed 2017d).

2.2.2 Financial Crisis Response by the European Central Bank

The Lehman bankruptcy, panic in stock markets, flight to safe assets (e.g. sovereign bonds) and fear of a complete failure of the financial system made asset prices fall and caused the whole economy to suffer. In the following, the deepest recession since the 1930’s hit Europe by 2009. The ECB started cutting the fixed rates for refinancing and deposit facilities from 3.75% respectively 3.25% in October 2008 to 1.25% and 0.25% in April 2009. The ECB further lowered the main refinancing fixed rate to 0.05% in September 2014 and to 0% in March 2016. The apparently lower bound of 0% for the deposit facility fixed rate was reached in July 2012. From June 2014 until March 2016, the deposit rate was subsequently reduced to -0.40% (European Commission 2009, ECB 2017b). The negative interest rate on deposits can be interpreted as an unconventional measure according to Joyce et al. (2012). Banks should have no incentive to deposit excess liquidity at the ECB but rather lend it to the economy, which stimulates investments, consumption and aggregated output.

The exposure to the US subprime market and highly leveraged financial institutions were the spark to spread the crisis over Europe. Financial markets suffered6 first and interbank spreads widened, which implied risk, uncertainty and troubles to recapitalize. Besides cutting the interest rates to zero or even below, the ECB provided additional liquidity through repurchase agreements to banks by expand-ing its balance sheet. The interbank spreads narrowed again as a result and prevented the market from a

6_{List of (almost) failing European institutions: Northern Rock, Landesbank Sachsen, Fortis, Dexia,}

(17)

11

collapse, but lending to businesses was still malfunctioning. Continuing, the ECB focused on risk shift-ing by exchangshift-ing risky long term assets for safe liquid assets and supportshift-ing financial institutions to recapitalize. The central bank thereby acted as the lender-of-last-resort and aimed to keep financial mar-kets liquid and to deleverage its institutions (European Commission 2009).

Since the conventional operations were not able to restore the functionality of the market, the ECB was driven to adopt unprecedented unconventional measures in the middle of 2009. Compared to the Fed, the ECB launched a bunch of less aggressive, non-standard monetary policy tools addressing liquidity and sovereign debt problems. European sovereign bond spreads widened relative to the German long term yield, especially from stressed European countries. Increasing spreads reflected anxiety about creditworthiness of governments and indicated a higher required risk premium by investors (Falagiarda and Reitz 2015).

Plot 2 displays the dimension of the ECB total asset development from 1999 till 2016. The balance sheet expanded from around 1.5 trillion EUR before the crisis to over 3.5 trillion EUR in 2016. It also gives an indication about the different programs and how they subsequently influenced the size of assets. On one side, the drivers of the expansion were the liquidity facilities (lending to euro area credit insti-tutions) in 2008 and 2012 and on the other hand, the large securities purchases after 2014.

Plot 2. European Central Bank balance sheet development

(18)

12

In May 2009, the ECB launched its first Covered Bonds Purchase Program (CBPP1) to improve liquidity and stimulate lending. In May 2010, the ECB was concerned about the indebtedness of some stressed euro area countries. In response, the ECB started the Securities Markets Program (SMP) in-cluding private and public debt to stabilize the dysfunctional financial sector, to ensure the monetary policy transmission mechanism and to conduct price stability in the medium term. In 2011, CBBP2 followed and additional liquidity actions to enhance bank lending. In 2012, Outright Monetary Trans-actions (OMT) was launched by purchasing sovereign bonds for financial stability and to improve the transmission mechanism. Both programs, the SMP and OMT, were implemented to address the stressed debt situation of European countries. In the late 2014, the ECB reinforced its effort and started a third CBPP3 including covered bonds and ABS. The program was addressed to facilitate lending to the euro area, spillovers to other markets and to target the desired inflation rate of 2%. In January 2015, the program expanded, implementing the Public Sector Purchase Program (PSPP) by adding sovereign bonds to list of purchasing securities (ECB 2017c).

Gerlach et al. (2010) found that widening Euro area bond spreads relative to the German ones are driven by government debt, the future fiscal deficit position and a country’s banking sector. Aggregated risk is highly related to the size of the banking sector compared to the GDP and high leverage ratios. Ghysels et al. (2014) discovered lower sovereign bond yields over the SMP lasting from May 2010 to February 2012. They observed intraday data of stressed European countries and found significantly lower bond yields with different maturities. Falagiarda and Reitz (2015) confirmed the results of Ghysels et al. (2014) that spreads narrowed for stressed euro area countries relative to the German ten-year bond. Furthermore, they observed narrowed spreads across long term refinancing operations, OMT and CBPP communications.

In contrast to the US QE, the European one is still alive. The 60 billion monthly asset purchase program, which was temporarily raised to 80 billion in March 2016, is expected to roll out by the end of 2017. Furthermore, the fixed rates for refinancing and deposit facilities remain 0% and -0.40% (Zeit 2017).

2.3 Transmission Mechanism of Monetary Policy

After defining conventional and unconventional monetary policy, its tools and goals, I briefly explain the transmission mechanism of monetary policy to the real economy and the importance of money sup-ply to the economic activity as explained in Mishkin (2007). Previously demonstrated, the financial crisis impaired the functioning of financial markets. Furthermore, the transmission of conventional mon-etary policy to the economy became difficult. The classic Keynesian theory, the interest rate channel,

(19)

13

indicates that an expansionary monetary policy decreases real interest rates,7 which lowers cost of cap-ital, stimulates investment, increases consumption and finally rises economic output. GDP growth may be translated into higher tax income and relieves sovereign balance sheets. Central banks normally in-duce short-term rather than long-term interest rates. The expectation hypothesis of the term structure states that long term interest rates reflect expected future short term rates. Even if nominal interest rates are at the floor of zero, lower real interest rates can still stimulate business and private investment. Expansion in money supply can increase the expected price level and therefore increase expected infla-tion, lowering real interest rates and stimulate consumption. The interest rate channel says that monetary policy can be effective even if nominal interest rates are at the lower bound. Previous sections showed that both QE announcements by the Fed and ECB affected short and long term bond yields and lowered the term structure of interest rates. Another channel illustrated by Mishkin (2007) is the banks’ lending channel. Banks play a special role in the economy and financial markets due to their business model of maturity and liquidity transformation. Expansive monetary policy increases bank reserves and deposits at national banks. Therefore, the number of bank loans to the economy increases and investments rise as a result. Despite this channel is widely accepted in theory, concerns rose about the channel not being powerful enough in practice and fear of it malfunctioning after the crisis. Plot 3 in Appendix A summa-rizes transmission channels of monetary policy mentioned in Mishkin (2007).

Cecioni et al. (2011) highlighted two channels through which unconventional measures are trans-mitted to economy. First, central banks use communications about the future evolution of expected short term interest rates, the assets they purchase or other tools to target dysfunctions of market segments. This channel is called ‘signaling channel’ and relies on central banks credibility, the success of the gained market confidence and how markets change their expectations. Important information for the markets are size, speed and general terms of the operations. Second, through the ‘portfolio rebalance channel’, central banks utilize their roles as monopolistic money suppliers. Central banks can conduct endless operations such as securities purchases, assets swaps and liquidity injections. Thereby, affecting asset prices and yields in the corresponding segment by regulating supply and demand of private assets and liabilities.

Falagiarda and Reitz (2015) labeled the signaling channel as central banks route the future eco-nomic path and thereby affect asset prices. The portfolio rebalancing channel implies portfolio shifts under the condition of imperfect substitutability among assets. As a result, asset prices rise and interest rates fall. This creates more favorable condition for the businesses and people to borrow money and invest. Higher asset prices imply a larger collateral for credits and lower interest rates imply lower

(20)

14

rowing costs. Another important aspect is the ‘liquidity premium’ or ‘market functioning channel’ es-pecially in times of dried out liquidity and high liquidity risk premium. Therefore, central banks need to improve conditions in such markets to further influence the economic system.

In the US study conducted by Gagnon et al. (2011) and the UK study done by Joyce et al. (2011), the portfolio balance channel is described as a replacement of private investors by central banks. In order to raise the acceptance of these operations, prices had to rise and the expected returns of purchased assets had to fall. Krishnamurthy and Vissing-Jorgensen (2011) defined channels as how QE was expected to affect assets. Long term asset purchases reduced the duration risk and the long term yields relatively to the short term yields which is labeled as the ‘duration risk channel’. In the ‘safety premium channel’ they described an increasing spread between high yield and investment grade bonds. Another important channel to point out is the ‘default risk channel’. Riskier bonds carry higher default risk. The default risk as well as the risk premium and risk aversion of investors fall if QE succeeds to stimulate and recover the economy. Research conducted by Blinder (2010), Joyce et al. (2012) and Haldane et al. (2016) support these theories and underline the importance of the aforementioned channels. Additional channels are mentioned in these papers such as the ‘exchange rate channel’ or ‘confidence channel’. I expect that communications of QE operations may affect the default probability of the affected entities and work through the presented channels.

To sum up, there is a mutual consensus in academia about the questions when and how uncon-ventional monetary policy is effective. The most common form described in literature is a massive ex-pansion of central banks’ balance sheet by open market purchases. QE works through the signaling and portfolio rebalancing channel, which are among the most frequently applied in order to explain the transmission mechanism to the real economy. Unconventional monetary policy helps to ease markets and to relieve the pressure from bank and sovereign balance sheet.

(21)

15

2.4 Credit Default Swaps

A Credit Default Swap (CDS) is an insurance contract, in which the buyer gets protected against a credit event8_{of a reference debt security (e.g. bonds or loans). CDS are among types of credit derivatives}

traded Over-The-Counter (OTC). Therefore, maturity and notional value are negotiable and upon agree-ment. The buyer obtains the right to sell the security for its face value, when a credit event occurs and gets compensated for the loss. In this case, the issuer of the CDS has to purchase the debt security against its face value. CDS spread increase when the reference bond becomes more likely to default and can be interpreted as an investors’ perception of the credit risk (Hull 2006). Despite CDS looking alike a regular insurance contract, Jarrow (2011) stated that there is one important discrepancy: one can protect itself against the underlying default without even holding it. This raised concerns about paradox incentives all around the financial crisis and led to a ban of naked CDS trades in 20129 (ESMA 2017).

The premium that the protection buyer has to pay is defined as the CDS spread. It is measured in basis points of the contracts’ face value and payments are mostly settled on a quarterly base. For exam-ple: a CDS spread of 100 bps to insure ING bank debt with face value EUR 10 million costs EUR 100’000 annually or EUR 25’000 quarterly over a five-year term. A CDS spread has to be solely inter-preted as an insurance premium. A bond spread is the difference between two different entities; for instance a German bond relative to an Italian bond (Deutsche Bank 2017).

Ang and Longstaff (2013) described the key advantage of using CDS spreads over bond yields, that CDS directly measure the credit risk of the underlying entity. Bonds are not only driven by the entities creditworthiness but rather by many other factors such as interest rate movements, the availabil-ity respectively supply and illiquidavailabil-ity. Chen et al. (2011) attempted to make CDS data more transparent, since these instruments are traded OTC and data is scarce. The find low trading activity for many single name CDS traded less frequent than once a day and clustered in time. They state that liquidity of sover-eign and corporate CDS was not a concern during the crisis, even though trading was scarce.

Plot 4 displays the semiannual evolution of CDS notional amount outstanding from 2005 until 2016 reported by the Bank for International Settlement. The CDS notional amount outstanding is meas-ured in billion USD with a semi-annual frequency. On the left axis, the amounts for Sovereign and Financial CDS are displayed, on the right axis the total amount by all sectors of reference. Before the crisis the notional amount sharply increased to hedge credit exposure, whereas the amount steadily de-creased in the aftermath. Augustin et al. (2016) explained the decrease in CDS notional amount after 2007 with occurred credit events, the subsequent regulatory wave for OTC markets such as the naked CDS trades ban and a netting portfolio compression of long and short positions. The regulations require

8_{Credit events can be either bankruptcy, failure to pay, restructuring or debt repudiation}

9_{European Securities and Markets Authority banned naked sovereign CDS on 1 November 2012; one must have}

(22)

16

financial institutions to hold more capital for OTC traded single-name CDS. Despite large quantities of issued bonds, financial institutions did not hedge their credit exposure with single-name CDS but rather centrally cleared indices, given low interest and default rates in the years prior 2016.

Plot 4. CDS notional amount outstanding

Source: BIS (2017); own presentation 2.4.1 Corporate and Sovereign CDS

Ang and Longstaff (2013) distinguished corporate and sovereign default based on three factors. First, corporates can get sued by bond holders to hand over assets to them. Sovereigns cannot hand over their domestic located assets in such a case. Second, sovereigns’ assets are protected by immunity, which prevents them from getting sued by individuals. Third, there is no international process for sovereigns’ defaults. Fontana (2010) found evidence for differences as well as commonalities between sovereign and corporate CDS spreads. Both experienced a significant repricing with the crisis; risk premium and the likelihood of tail events thereby play an important role. The CDS differ in the flight to safety mech-anism and the importance of technical default. For sovereigns, the flight to safety effect creates a wedge with positive basis between the price of the underlying debt security and the corresponding CDS. For corporates in contrast, the basis is negative. This suggests for corporates spreads a liquidity premium and sovereigns a liquidity discount. Furthermore, they find limited trading activity of sovereign CDS around crisis years, which affected price discovery.

2.4.2 What CDS Spreads reveal about Default Risk?

The relationship between the annualized CDS spread and the probability of default can be expressed as

𝐶𝐷𝑆 𝑠𝑝𝑟𝑒𝑎𝑑

(1−𝑅) = 𝑝, (2.1)

where R is the recovery rate of the claim and p the annual default probability. One can interpret the formula as follows: the higher the CDS spread, the higher the default probability of the underlying debt security (Deutsche Bank 2017). According to Fitch Ratings (2017), CDS spreads reflect the market view

(23)

17

on debt and are commonly used market-based indicators for risk analysis. CDS are used to derive the estimated default risk of the underlying debt instrument. The pricing could be driven by additional fac-tors such as market risk aversion, creditworthiness or liquidity conditions. In other words, the spread is a measure of risk and reflects the market perception of the credit risk of the underlying reference.

Duffie (1999) showed under no arbitrage conditions that a CDS spread can be priced as the spread between the risk free rate and a par floating-rate bonds. Longstaff et al. (2005) found that the default risk of a corporate bonds is mainly explained through its corresponding CDS spreads. For the best rated underlying investment grade bonds, CDS spreads explain over 50% of default risk and even more than 80% for low quality debt. However, bonds include many other risk drivers, which makes them more inadequate than CDS spreads to measure default risk. The nondefault component of bonds is driven by bid ask spread, the outstanding notional amount and the overall liquidity of fixed income markets.

Companies of the same industry or region tend to be similarly affected by external shocks and may experience simultaneously financial distress. Default rates vary over time and are related to eco-nomic conditions. Furthermore, default of one company may cause the default of other companies through a contagion channel. If the trigger company is a systemic bank, it could affect several sectors and even governments. This effect is described as default correlation in Hull (2006).

Two effects of QE reflection on default probabilities can be derived from the previous sections. On one side is the supporting activity through QE for the financial sector and the whole economy, which reduces the probability of default. On the other hand, credit easing may reinforce risk taking by cheaply and excessively lending and borrowing, which implies an increasing default probability. Therefore, debt holders are eventually affected by a change in default probabilities through QE, which is expected to be reflected in CDS spreads.

2.4.3 What CDS Spreads reveal about the Economic Situation?

CDS spreads and what they reveal in relation to unconventional monetary policy is not widely re-searched yet. Mizen et al. (2014) found increasing CDS spreads and default risk before the recession emerged after the crisis. Their results suggest that simple single-name CDS averages and CDS indices indicate distress in financial markets and predicts real economic activity up to four quarters ahead. Both CDS variables have significantly negative coefficient on real GDP, investments and employment. Fur-thermore, they identified larger negative effect on the real economy by non-financial CDS relative to financials. They emphasized, that CDS indices are good predictors of macroeconomic variables and reacted ahead compared with bond yields. Bleaney et al. (2012) discovered similar results by observing European sovereign bonds, real GDP growth and financial distress. Bond spreads and economic varia-bles have a negative relationship up to two years ahead.

(24)

18

Gerlach (2015) observed volumes of ECB repurchase agreements, CBPP1 and SMP on sovereign and bank CDS. The purchases of covered bonds during CBPP1 and SMP helped to reduce the CDS spreads whereas repurchase agreements did not affect them. She finds further evidence, that some ECB unconventional measures had a paradox effect. CDS spreads rose, which could have been interpreted as a sign of ambiguity and disrupted markets. Krishnamurthy and Vissing-Jorgensen (2011) found lower in CDS spreads through a reduction in default risk amongst other factors during Fed’s QE1 and QE2. 2.4.4 Components and Drivers of CDS Spreads

Ang and Longstaff (2013) divided the credit risk into a systemic and a sovereign-specific component and find higher systemic risk for European countries compared to US states. Systemic default risk is highly correlated between the US states and European countries and strongly related to stock market and corporate CDS indices. Fender et al. (2011) focused on the Fed’s monetary policy communication spillover to emerging countries sovereign CDS spreads before the crisis. They found that emerging country CDS are strongly related to global and regional risk rather than country-specific factors. Their results confirm Ang and Longstaff (2013) that CDS spreads are related to broad stock market and CDS indices. Furthermore, they found important economic influence of US monetary policy actions on CDS markets. Decisions of the Fed are important drivers of financial markets and calm down the markets outside the USA. Fontana (2010) observed that CDS pricing is due to common factors related to credit market developments reflected in the iTraxx CDS index and global risk aversion measured by implied volatility. Furthermore, their results suggest that CDS spreads are not linked to bonds outstanding over GDP and CDS market liquidity.

Takam (2014) found that global variables are the most significant estimators of CDS spreads. Local stock markets, oil price, exchange rates and local short term yield can give evidence for most emerging market CDS. An additional point is the importance of the banking sector for the economic stability and vice versa. A stressed and large banking sector can raise concerns about debt sustainability of a country and therefore affects CDS spreads. Ang and Longstaff (2013) and Takam (2014) charac-terized the caveat of the observed after-crisis period by excess liquidity provided by central banks and flight to safety in various markets. Gerlach (2015) in contrary identified the advantage of CDS that they do not respond to a flight to safety, since there are no substitute assets as there are for bonds. CDS respond to new information and investors change their opinion about the default probabilities, which is reflected in a higher or lower spread. Arce et al. (2013) observed the relative basis of the deviation between a bond and a CDS relative to the CDS spread. The main drivers are the counterparty risk of the CDS issuer, the relative liquidity of CDS to bonds, stock market volatility and ECB sovereign bonds purchases. Bonds purchases widened the basis because bonds were stronger affected than CDS spreads. Annaert et al. (2013) stressed that CDS of Euro area banks have changing dynamics with credit risk, liquidity, business cycle and market wide conditions. During the recent crisis, credit risk was the main

(25)

19

CDS driver but also individual and market wide liquidity played an important role. Especially the im-portance of liquidity is contrary to Chen et al. (2011) and Fontana (2010). Nevertheless, in the analysis of Annaert et al. (2013), the bid-ask spread as the liquidity measure is not statistically significant at any level. Concluding, CDS spreads serve as a good proxy for default risk and are likely to be affected through unconventional monetary policy communications.

2.5 Hypothesis

Previously, I described unconventional monetary policy and through which channels sovereign and bank default risk is affected. In the reminder, these findings serve as the basic concept to conduct the empirical analysis of central banks’ communication effects on credit default swaps.

Gerlach et al. (2010) emphasized the importance of the banking sector for European economies. Governments cannot let major banks to go down, which basically implies a governmental guarantee for the financial sector. This guarantee or implicit liability weights on sovereign default risk. They found that the size and capitalization of the banking sector are important drivers of sovereign bond spreads and risk. Ang and Longstaff (2013) found that European and US systemic shocks are highly correlated and that sovereign risk is strongly related to financial markets. The relationship of the banking sector, the real economy and sovereigns could be summarized as follows: the banking sector affects the real economy via investment and consumption through various channels such as the interest rates and lend-ing, this impacts sovereign balance sheets and creditworthiness via unemployment and taxes. Panetta et al. (2011) described channels how sovereign risk affects banks: losses on sovereign debt securities on banks’ balance sheets, lower value of collateral available for funding, credit rating downgrades for sov-ereigns go hand in hand with downgrades for domestic banks and a lower value of governmental guar-antee. Given the aforementioned literature, I channel the information and formulate two hypotheses regarding the impact on CDS spreads, which I analyze in the following parts.

Many of the cited papers found positive effects of unconventional monetary policy for the do-mestic economy, the banking sector and even spillovers to other economies (Krishnamurthy and Vissing-Jorgensen 2011, Neely 2015, Gagnon et al. 2011, Fender et al. 2010, Fratzscher et al. 2016 for the US, Gerlach et al. 2010, Ghysels et al. 2014 and Falagiarda and Reitz 2015 for Europe). They all discovered significant effects on various asset prices and confirm the supporting intention by the central banks. QE programs brought stability to the economy, reduced financial turmoil and facilitated a recov-ery path, which was beneficial for the stressed banking sector and sovereigns. Bond prices rose and central banks bought toxic assets from financial institutions, which relieved the pressure of banks’ bal-ance sheet. This leads to the first hypothesis, that CDS spreads of banks and sovereigns decrease respec-tively the daily changes are negative around QE announcement dates:

“The effect of unconventional monetary policy communications by Federal Reserve and European Central Bank reduce bank and sovereign CDS spreads.”

(26)

20

Regarding the timeliness when observed QEs were announced, it is likely to have varying effects. On one side, the US QE started earlier and is proven to have beneficial international spillover effects, which may have also reduced or at least mitigated European default risk. In event studies of Gagnon et al. (2011), Krishnamurthy and Vissing-Jorgensen (2011) and Falagiarda and Reitz (2015), earlier an-nouncements had larger effects on the observed asset prices. Effects of later anan-nouncements of QE3, CBPP3 and PSPP are not studied yet to my best knowledge. The existing results and the rising con-sciousness of market participants over time suggest, that US QE would have more impact on default risk. On the other hand, the financial crisis hit Europe delayed and erupted the Euro debt crisis and default risk in 2010. Due to timeliness if QE events and market circumstances, the ECB may stronger affect CDS spreads. Trading this two possibilities against each other I state my second hypothesis that Fed announcements have larger effects on CDS spreads:

“Unconventional monetary policy communications of Federal Reserve have had a larger impact on CDS spreads than European Central Bank communications.”

(27)

21

3. Methodology

CDS contracts are frequently traded, the spread fluctuates based on market perception that the underly-ing entity will default. When the debt issuer is more resilient to economic shocks, debt holders are more likely to be fully repaid. Safer entities therefore face lower funding costs and sellers of CDS on these debt instruments will demand lower premiums. Measuring the immediate effect of QE announcements on CDS spreads can provide insights into the stabilization effect of sovereigns and banking institutions. Under the efficient market theory of Fama (1965), every relevant information or trend, which is available for market participants, immediately implements in the intrinsic value of assets. The new information may be an actual or anticipated change, which affects an entities’ condition or perspective. Therefore, the event study is a suitable approach to measure the effect of QE on default risk, using CDS as a proxy. McWilliams and Siegel (1997) determined that an event study is valid under the assumptions that mar-kets are efficient, the events unanticipated and no confounding effects happen during the event window. My methodology is based MacKinlay (1997), De Jong and De Goeij (2011) and Greatrex (2015). The methodology is hereby modified for CDS spread changes rather than stock returns. Beginning with raw daily CDS spreads, I take the natural logarithm and compute the daily changes as follows:

∆ log 𝐶𝐷𝑆_𝑡= 𝑙𝑜𝑔𝐶𝐷𝑆_𝑡− 𝑙𝑜𝑔𝐶𝐷𝑆_𝑡−1 (4.1) The market model is used to estimate the expected normal CDS changes. The assumption bases on a constant and linear relationship between individual CDS changes and the market change.

∆𝐶𝐷𝑆_𝑖,𝑡 = 𝛼_𝑖+ 𝛽_𝑖𝐶𝐷𝑆_𝑚,𝑡+ 𝜖_𝑖,𝑡 𝑤𝑖𝑡ℎ 𝐸[𝜖_𝑖,𝑡] = 0 𝑎𝑛𝑑 𝑉𝐴𝑅[𝜖_𝑖,𝑡] = 𝜎_𝜖,𝑖2 (4.2) 𝐶𝐷𝑆_𝑖,𝑡 and 𝐶𝐷𝑆_𝑚,𝑡 are daily logarithmic spread changes of CDS 𝑖 and CDS index 𝑚 at time 𝑡. The size-weighted market index is self-constructed and based on four CDS indices10 according the quan-tity of included entities. The abnormal spread change is indicated as 𝜖_𝑖,𝑡. The model parameters 𝛼_𝑖 and 𝛽_𝑖 are estimated by ordinary least squares regression. The abnormal spread change (𝐴𝑆𝐶) across the event window is defined as the difference between the realized change and the estimated model-based spread change.

𝜖_𝑖,𝑡 = 𝐴𝑆𝐶_𝑖,𝑡 = 𝐶𝐷𝑆_𝒊,𝒕− 𝛼̂ − 𝛽_𝒊 ̂ 𝐶𝐷𝑆_𝒊 _𝒎,𝒕 (4.3) The parameters are estimated by daily spread changes in the estimation window, which contains 150 trading days. It begins 160 days and ends 11 days prior the event date 𝑇₀.

10_{CDS Indices with number of components in brackets: CDX IG (125), CDX HY (100), iTraxx Europe (125),}

iTraxx Europe Crossover (75); the weights to construct the overall market index: CDX IG (30%), CDX HY (22.5%), iTraxx Europe (30%), iTraxx Europe Crossover (17.5%)

(28)

22

The cumulative abnormal spread change (𝐶𝐴𝑆𝐶) for each individual event and CDS 𝑖 across the event window11 𝑇₁≤ 𝑇₀ ≤ 𝑇₂ is defined as follows:

𝐶𝐴𝑆𝐶_𝑖(𝑇₁_,𝑇₂₎= ∑𝑇2 𝐴𝑆𝐶_𝑖,𝑡

𝑡=𝑇1 (4.4)

The cumulative average abnormal spread change 𝐶𝐴𝐴𝑆𝐶 across CDS sub-samples is defined as: 𝐶𝐴𝐴𝑆𝐶_(𝑇₁_,𝑇₂₎ = 1

𝑁∑ 𝐶𝐴𝑆𝐶𝑖(𝑇1,𝑇2) 𝑁

𝑖=1 (4.5)

The standard cross-sectional test statistic of 𝐶𝐴𝐴𝑆𝐶 is based on abnormal spread changes and considers the number of days in the event window. The testing framework is parametric. The test statistic is based on the normal distribution and homoscedasticity assumptions of ASC, that the expected CASC for each entity equal zero across the event window. Therefore, the 𝐶𝐴𝐴𝑆𝐶 are also expected to be zero, that the event has no impact on mean and variance. The null hypothesis 𝐻₀ can be constructed:

𝐻₀: 𝐸[𝐶𝐴𝐴𝑆𝐶_(𝑇₁_,𝑇₂₎] = 0 (4.6)

The standard deviation with 𝑁 − 1 degrees of freedom12

is calculated per sub-sample across the event window as:

𝑠 = √ 1

𝑁−1∑ (𝐶𝐴𝑆𝐶𝑖(𝑇1,𝑇2)− 𝐶𝐴𝐴𝑆𝐶(𝑇1,𝑇2)) 2 𝑁

𝑖=1 (4.7)

The t-test for the null hypothesis then is: 𝑇𝑆 = √𝑁𝐶𝐴𝐴𝑆𝐶(𝑇1,𝑇2)

𝑠 ≈ 𝑁(0,1) . (4.8)

This procedure is performed for each event individually. Additionally, I aggregate the 𝐶𝐴𝐴𝑆𝐶 for every Fed and ECB announcements and check it with the same standard cross-sectional testing proce-dure as in (4.7) and (4.8).

For the further analysis I implement additional variables in a regression analysis to explain ab-normal spread changes around central bank announcements. The regression to explain abab-normal spread changes across the three-day event window is set up as follows:

∆𝐴𝑆𝐶

𝑡

= 𝛼

0

+ ∑

𝑗𝑖=1

𝛽

𝑖

∗ ∆𝑀

𝑡

+ 𝜀

𝑡

(4.9)

For the independent market variables 𝑀, I also take the logarithm of the daily observations and calculate the daily changes according the daily CDS spread changes in equation (4.1).

11_{I use a three-day event window in the main analysis, which contains the day before (T}

1), the day of (T0) and the day after (T2) the announcement of a central bank.

(29)

23

4. Data

The dependent variable in this analysis is the CDS spread of banks and sovereigns. Being more precisely, the daily changes of the natural logarithm of CDS spreads. This variable is used to approximate the adjustment of default risk of underlying entities when certain unconventional policy measures by the Fed and ECB were announced.

The sovereign and bank sample setup depends on a few basic criteria. First, European sovereigns have to be part of the Euro area over the whole period. Banks are selected among components of the two indices S&P 500 and EuroStoxx 600. Second, the availability of CDS spreads is crucial. Therefore, I exclude entities with no available CDS, with more than 30% of daily zero spread changes and which have no spread available for five and more events. Third, I exclude banks which were delisted, merged or failed during the period of interest. For the remaining entities, if there was a missing value, I carried the last observation forward. Otherwise the data would show a -100% change from a day with observa-tion to one without, which would strongly influence either the estimaobserva-tion or event window. The final sample consists of 13 Sovereign and 44 Bank CDS spreads and is divided into three sovereigns and four bank sub-samples. The sovereigns are divided into the USA, PIIGS and core Euro area countries. Banks are grouped by their country of origin and according the allocation of the sovereigns. Furthermore, I added a sample of European non-Euro area banks to show the interconnectedness of the banking sector. In Table 1 in Appendix B, each entity is listed according to their economic and political affiliation. Based on default correlation in Hull (2006), I assume them to react similar when new information about future economic path comes public. Banks are grouped according to the sovereigns since their default risk is strongly related to the corresponding sovereign and its creditworthiness. Regarding the event study and to construct the market return, I collected four equally-weighted CDS market indices13 with underlying investment grade and high yield senior debt securities respectively, two each for US and European markets.

The single-name CDS spreads with five year tenor are collected from Bloomberg, CMA and Datastream. Most of the literature such as Greatrex (2015), Gerlach (2015) or Fender et al. (2011) con-firmed that 5 year CDS are the most liquid and frequently traded. Nevertheless, they noticed that CDS data tend to show more zero changes than other assets. CDS are traded OTC and therefore not com-pletely accurate. I asserted also that daily CDS spreads do not vary as much as stock prices for instance. The reasons for zero changes can either be no available data, no traded volume or traded volume but risk perception did not change. Since the accessible database do not report volumes, I assume zero changes to imply no risk shifts given that CDS spreads were available. Comparing the databases I further experienced deviations in availability and CDS pricing. CMA only reported until 2009 and did not match

13_{US: CDX IG and CDX HY; Europe: iTraxx Europe (IG) and iTraxx Europe Crossover (HY), where IG stands}

(30)

24

Bloomberg and Datastream data due to different reporting approaches. Furthermore, Datastream data showed significantly more zero spread changes than Bloomberg. Therefore, I solely rely on data from Bloomberg to conduct my analysis. Mayordomo et al. (2014) compared six databases which provide market indices and find similar discrepancy among them due to timing, availability of information and pricing.

Plot 5 and show the evolution of daily five-year CDS spreads in basis points from 2008 to 2016. Plot 5 reveals that the financial crisis around Lehman Brother collapse in September 2008 created a divergence between different credit default swaps. In early 2009, a period of lowering spreads and vol-atility began. CDS started rising again in 2010 when concerns about the creditworthiness of stressed European countries emerged. After 2012, the recovery phase started and CDS spreads converged again. The only exception was in July 2015, when Greece rejected the bailout referendum and both PIIGS CDS jumped (BBC 2017b). CDS spreads vary over time and Plot 5 suggests that CDS follow a macroeco-nomic trend. During periods of overall deterioration around Lehman Brother failure and the Euro debt crisis, all spreads hiked and their variation changed during such periods. Another observation is that CDS differ by group: for instance both PIIGS14 samples, which I assume to be less creditworthy, show higher spreads than other groups. Plot 6 in Appendix B shows, that the average logarithmic five-year CDS spread changes were surprisingly less volatile from 2010 to 2013. In 2014 and 2015, CDS became again slightly more volatile.

Plot 5. Sovereign and bank daily absolute CDS spreads

Source: Bloomberg; own calculation and presentation.

14_{PIIGS stands for the initial letters of European countries, which are mostly determined as stressed countries:}