The effect if the Fed's unconventional monetary policy on the Eurozone

(1)

Master Thesis Master Economics

Faculty of Economics and Business

Specialization: International Economics and Globalization

The effect of the Fed’s unconventional monetary policy on the

Eurozone

Name: Femke Blom

Student number: 6049672

E-‐mail: femke.blom@student.uva.nl Supervisor: Konstantinos Mavromatis Date: 20/04/2015

(2)

Abstract

The Fed could not solve the issues related to the recent financial crisis with conventional measures and therefore turned to unconventional monetary policy. This thesis analyzes the effect of the unconventional monetary policy announcements of the Fed on the Eurozone economy and the channels through which these announcements affected it. The results are computed by using descriptive analysis, dummy regression and the ARMA model. The results show that the unconventional monetary policy announcements significantly affected the Eurozone economy, but that there is insufficient proof to conclude that the signaling channel played a significant role.

(3)

1. Introduction

In August 2007 the housing bubble in the United States (the U.S.) burst. Asset-backed securities and in particular mortgage-backed securities (MBS) became illiquid. Asset-backed securities are securities created from a pool of non-mortgage assets, which are subsequently sold to investors. Mortgage-backed securities are created from a pool of mortgages before being sold to investors. Because mortgage markets were highly interrelated with other financial institutions, financial firms started having trouble to gain access to private liquidity. The crisis even worsened in September 2008 when the U.S. government took Fannie Mae and Freddie Mac (both government sponsored enterprises) into conservatorship, bought most of the equity of a major insurance company called AIG for which they got an emergency loan of the Federal Reserve (the Fed) in return and let Lehman Brothers go bankrupt. From the beginning of the financial crisis in 2007, the Fed reacted by using unconventional monetary policy measures (Labonte and Webel, 2010).

During the recent financial crisis conducting conventional monetary policy was much more difficult because of the severe impairment of the transmission mechanisms and because of the malfunctioning of the financial market. Mishkin (2010) defines the transmission mechanisms of monetary policy as the channels through which the money supply affects economic stability. In December 2008, after having lowered the federal funds rate in a series of ten moves towards a range between 0% and 0.25%, the ‘zero lower bound’ was reached. The Fed could no longer solve the financial problems with conventional measures and had to turn to unconventional monetary policy (Cecioni et al., 2011). From 2009 up until 2013 the Fed mainly conducted unconventional monetary policy in two ways. The first was through large-scale asset purchases (LSAPs), also known as ‘quantitative easing’. The second was through forward guidance (Bauer and Neely, 2014). The focus of this thesis will be on these two forms of unconventional monetary policy measures. Unconventional monetary policy mainly operates through two channels: the signaling channel and the portfolio-balance channel. The signaling channel, as described by Cecioni et al. (2010), works through communication of central banks in order to regain market confidence and to influence market expectations with respect to future decisions on economic or monetary policies. The portfolio-balance channel is the way through which purchases of securities (public and private) and credit provision to (non)-financial institutions can affect the economy.

(5)

literature has expanded and now also includes studies on the international effects of the Fed’s monetary policy. For example, Neely (2013) found that LSAPs in the U.S. had significant effects on bond and exchange rate markets in economies other than the U.S. economy. Bauer and Neely (2014) also found significant effects of the LSAPs on bond yields of foreign countries. They even established the channel through which the LSAPs affected international markets.

While most papers that focus on the international effects of the Fed's monetary policy include a few important economies dispersed over the world, this thesis will contribute to the current literature by specifically looking at the Eurozone. It will analyze the effects on fifteen of the seventeen Eurozone countries taking into account the Fed’s unconventional measures since the start of the recent financial crisis in September 2007 that have been proven to be the most important by earlier literature. The following research question will be answered: To what extent does the unconventional monetary policy of the Fed affect the economy of the Eurozone?

To answer this question, three empirical approaches are needed. All of them are event studies. The first approach is a model-free approach and consists of two parts. The first part is a descriptive analysis of the graphs of the 10-year and 2-year government bond yield. The second part is a dummy regression that shows which announcement dates led to a significant change in the 10-year or 2-year government bond yield. The second approach will look at the change of the 10-year government bond yields of the Eurozone countries after the announcement of the unconventional monetary policy measures. Using a time series regression model, also known as the autoregressive-moving average (ARMA) model, and a t-statistic, the significance of the effect of Fed announcements on Eurozone bond yields can be established. The third approach focuses on the channel through which the Eurozone economy is affected by the unconventional monetary policy of the Fed. The signaling and the portfolio balance channel both affect the long-term interest rates, but determining which channel dominates in affecting the long-term interest rate of a Eurozone country has been a challenge for researchers. Therefore, this research only focuses on the importance of the signaling channel by looking at the 2-year government bond yields. For this approach, the descriptive analysis, dummy regression (both already mentioned in the first and second approach) and ARMA model will be used as well.

The structure of this thesis will be as follows. The second section will provide a literature review. The third section will describe the monetary policy of the Fed. It will explain how conventional and unconventional monetary policy is conducted and more specifically how the

(6)

Fed conducts it. Based on the literature several hypotheses will be formulated. The fourth section will explain the empirical method that will be used to test my hypotheses. Based on this, the fifth and sixth section will consist of a discussion and conclusion, respectively.

2. Literature review

There are two strands of literature about the effects of unconventional monetary policy by the Fed. The first strand takes a broad approach including (almost) all unconventional monetary policies by the Fed. This strand will be covered in the first part of the literature review. It is mostly descriptive rather than stating clear conclusions about the effect of the policies. The second strand of literature focuses extensively on a particular unconventional monetary policy, clearly analyzing its effect on financial and macroeconomic factors nationally and internationally. This strand will be discussed in the second part of the literature review. It also includes papers that determine through which channel, the signaling or portfolio balance channel, the unconventional monetary policy of the Fed affects the national and international economy. This will be the subject of the last part of the review.

2.1 Unconventional monetary policies by the Fed since September 2007

At the beginning of the crisis the Fed responded to the financial and economic problems in an ad hoc manner. After the panic in September 2008 concerning Lehman Brothers, Fannie Mae, Freddie Mac and AIG a more systematic approach was needed (Labonte and Webel, 2010). The unconventional monetary policy measures are structured and covered in previous literature in different ways. In this section the most important unconventional policy measures of the Fed according to five papers will be presented. Not only the Fed but also the U.S. department of the treasury (the Treasury) and the Federal Deposit Insurance Corporation (FDIC) played a large role in introducing unconventional programs of their own. These programs will not be discussed because the focus of this paper is on the unconventional monetary policy of the Fed specifically.

The first reaction by the Fed to the financial problems was the creation of the Term Auction Facility (TAF). The TAF was introduced in December 2007 to auction reserves to banks for which the Fed wanted collateral in return. The Fed was convinced that the discount window was not used enough (Labonte and Webel, 2010). The next important unconventional policy measure was the creation of the Term Securities Lending Facility (TSLF) in March

(7)

Dealer Credit Facility (PDCF) was created. This was a discount window for primary dealers. Three other unconventional policy measures that are mentioned in the papers of Cecioni et al. (2011) and Labonte and Webel (2010) followed in 2008. The first two consisted of an Asset-Backed Commercial Paper Money Market Mutual Fund Liquidity Facility (AMLF) in order to stabilize the commercial paper markets and a Commercial Paper Funding Facility (CPFF). The third measure was a reaction to the problems in the asset-backed securities market. It marked the start of a more systematic approach of the Fed. In November 2008, the Fed announced the establishment of the Term Asset-Backed Securities Loan Facility (TALF). The TALF was financed by the Troubled Asset Relief Program (TARP), which was created to increase liquidity in the secondary mortgage markets by purchasing illiquid mortgage-backed securities. Neely (2013) discusses the fact that the TALF announcement and the first asset purchase announcement coincide. Research on previous event studies has proven that non-policy news did not have large effects on bond markets and the same announcement date should not complicate research (Faust et al., 2007).

From 2009 up until 2013 the Fed mainly conducted unconventional monetary policy in two ways. The first was through LSAPs, also known as ‘quantitative easing’. The second was through forward guidance (Bauer and Neely, 2014). These two ways of unconventional monetary policy are analyzed the most in literature about the effects of unconventional monetary policy of the Fed. The exact dates of the announcements of the LSAPs and forward guidance can be found in table 4 of appendix 1. This table also contains the announcement dates of all other unconventional monetary policy measures that have been discussed in this section. In section two, papers on LSAPs and forward guidance and their effects on financial and macroeconomic factors will be reviewed.

2.2 LSAPs and forward guidance

The literature about the effect of unconventional monetary policy of the Fed on foreign economies has mainly focused on LSAPs and forward guidance and their effects on economies abroad. This section will discuss the previous literature on the effects of LSAPs and forward guidance on the domestic economy but it will mainly focus on its effects on foreign economies.

(8)

2.2.1 LSAPs

There have been three Large Scale Asset Purchase programs in response to the recent financial crisis. Their effects on financial and macroeconomic factors have been researched extensively in the literature.

Gagnon et al. (2011) discovered that the LSAPs caused longer-term interest rates on numerous securities to fall. Even longer-term interest rates on securities that were excluded from the program decreased significantly. Baumeister and Benati (2010) used these results by Gagnon et al. (2011). They concluded that the Fed had averted the great risk of getting deflation and of having output collapses with the LSAPs. Williams (2011) assessed the effect of LSAPs by looking directly at its effects on the Treasury yield. Looking at different studies, he concluded that the average effect was 0.15- to 0.20-percentage point on the Treasury yield and argued that this was significant.

Recent research has focused more on the international effects of the Fed’s unconventional monetary policy. Both Neely (2013) and Bauer and Neely (2014) found that LSAPs in the U.S. had significant effects on bond and exchange rate markets in economies other than that of the U.S. Before that, Craine and Martin (2008) discovered that asset prices in Australia were affected by monetary policy measures in the U.S. Also, Hausman and Wongswan (2011) discovered that monetary policy measures by the U.S. affected bond yields in a lot of foreign countries. Bauer and Neely (2014) even established the specific channels through which the LSAPs affected international markets. Research on these channels will be reviewed in part three of the literature review.

2.2.2 Forward guidance

A credible central bank can potentially prevent the effects of the zero lower bound by using forward guidance (Williams, 2011). Williams (2011) shows the effect of forward guidance by referring to a Federal Open Market Committee (FOMC) statement in August 2011. This statement expressed a commitment to ‘exceptionally low levels for the federal funds rate at least through mid-2013’. Figure 1 shows the effect of this statement and thus of the effect of forward guidance.

(9)

Figure 1: Intraday U.S. Treasury yields

Research on the international effects of the Fed its unconventional monetary policy, exemplified by the papers of Neely (2013) and Bauer and Neely (2014), has mainly focused on the LSAPs and forward guidance. This thesis will also focus on these two forms of unconventional monetary policy in order to be able to compare the results with those of other countries.

2.3 Signaling channel and portfolio balance channel

The growing literature on the unconventional monetary policy of the Fed has not only focused on its effects but also on the channels through which it affected economies. The two channels through which unconventional monetary policy can affect the economy of a country are the signaling channel and the portfolio balance channel.

There is no agreement in the literature on the dominant channel of the effects of the unconventional monetary policy of the Fed. Gagnon et al. (2011) discovered that the portfolio balance channel dominates in the U.S., whereas Bauer and Neely (2014) found that the signaling channel was of greater importance They also established that the signaling channel plays a larger role for countries that already had large yield responses in time of conventional monetary policies of the U.S. Because the channel through which the unconventional monetary policy of the Fed affects economies is still unclear and because Bauer and Neely (2014) only looked at Germany as a Eurozone country, there is still an opportunity for research with respect to the Eurozone countries.

3. Monetary policy of the Fed

In times of economic stability the Fed traditionally conducts monetary policy through targeting the short-term interest rate, the federal funds rate. Through different channels and transmission mechanisms this federal funds rate targeting can fulfill the Fed’s goals of

3

Several studies have examined the effects of central bank communication more generally (see Gurkaynak et al. 2005, Kohn and Sack 2004, and Bernanke et al. 2004). They found that the Federal Reserve’s policy statements have significant effects on financial market expectations of future policy actions and on Treasury yields. Only a few studies have looked at the effectiveness of forward guidance policies at the

zero lower bound. One example was the Bank of Canada’s April 2009 statement that it expected to hold

the policy rate constant until the second quarter of 2010, which had an immediate effect on financial market expectations regarding short-term interest rates. The conditionality of the guidance worked as well. When the Canadian economy appeared to be recovering from the recession more quickly than anticipated, market participants began to expect interest rates to rise ahead of the previously announced date (see Chehal and Trehan 2009).

Of course, we at the Fed have our own recent case study that speaks to the effectiveness of forward guidance. The Federal Open Market Committee’s statement issued following our August meeting said, “The Committee currently anticipates that economic conditions—including low rates of resource utilization and a subdued outlook for

inflation over the medium run—are likely to warrant exceptionally low levels for the federal funds rate at least through mid-2013.” As Figure 1 shows, two-year Treasury yields fell by about 0.1 percentage point and ten-year Treasury yields fell by about 0.2 percentage point following the announcement. This provides prima facie evidence of the powerful effects of forward guidance at the zero bound.

Large-scale asset purchases

Let me turn now to large-scale asset purchases, or LSAPs, the main alternative to forward guidance in the unconventional monetary policy arena

today. LSAPs are central bank purchases of securities funded by an increase in reserves. Their history is older than forward guidance. It goes back at least to Operation Twist, the 1961 joint initiative of the Kennedy Administration and the Federal Reserve to purchase longer-term Treasury securities. More recently, the Bank of Japan began its so-called quantitative easing policy in 2001. It ultimately resulted in Bank purchases of almost ¥35 trillion of Japanese government bonds. In March 2009, the Bank of England announced it would purchase £75 billion of U.K. gilts, which was subsequently expanded to £200 billion. And the Federal Reserve carried out three rounds of large-scale asset purchases during the Great Recession. Two rounds of “QE1” took place in November 2008 and March 2009, during the financial crisis. The third round followed the “QE2” announcement in November 2010.

A number of theories consider the channels by which LSAPs affect Treasury yields and financial conditions more broadly (see Krishnamurthy and Vissing-Jorgensen 2011 for a thorough discussion). I will highlight two: signaling and portfolio. The signaling channel works through the effects asset purchases have on public expectations of future short-term interest rates. The portfolio channel works

Figure 1

Intraday Treasury yields

Note: Trading data from 9:30 a.m. to 4 p.m. EDT, at five-minute intervals. Source: Bloomberg. 2.0 2.1 2.2 2.3 2.4 2.5 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 8/8/11 8/9/11 8/10/11 8/11/11 Percent Percent 2-year yield (left axis) 10-year yield (right axis) FOMC announcement

(10)

maximum employment, stable prices and moderate long-term interest rates (Labonte, 2014). In the first section of this chapter, the conventional monetary policy of the Fed will be discussed. The severe impairment of the transmission mechanisms and the malfunctioning of the financial market made the conduct of monetary policy during the recent financial crisis much more difficult. The Fed had to turn to unconventional monetary policy because it could no longer solve the financial problems with conventional measures (Cecioni et al., 2011). The unconventional monetary policy of the Fed will be the subject of the second section of this chapter.

3.1 Conventional monetary policy

Until the subprime financial crisis in 2007, the Fed successfully conducted monetary policy without the use of an explicit nominal anchor1 to contain the price level. Mishkin (2010) describes this policy of the Fed as a ‘just do it’-policy. Although this approach had advantages like no dependency on a stable money-inflation relationship and demonstrated success in the past, there was also an important disadvantage of the lack of transparency. In the following chapters it will be shown that during the recent financial crisis the use of forward guidance2

eliminated this lack of transparency.

To conduct its monetary policy a central bank can choose different tools and instruments. The federal funds rate is determined in the market of reserves and affected by three possible monetary policy tools (Mishkin, 2010). The first tool is open market operations, which are the primary determinants of changes in the monetary base and short-term interest rates. By buying or selling bonds in the open market, the Fed can expand or contract reserves and thus the monetary base. This changes the money supply and short-term interest rates. The second tool is the discount window at which the Fed allows depository institutions to temporarily borrow directly. For direct lending at the discount window the Fed charges the discount rate, an interest rate slightly above the federal funds rate. In times of economic and financial stability the importance of the discount window is negligible but during a financial crisis it can act as a lender of last resort and serves as a source of reserves (Labonte, 2014). The final tool consists of changing reserve requirements, which is rarely used. Depending on the size of a bank it is required to hold 0% to 10% in reserves. An increase of reserve requirements can immediately cause liquidity problems for banks (Mishkin, 2010).

(11)

Having chosen the monetary policy tool, the central bank then decides on the policy instrument. The policy instrument is the variable that can signal the monetary policy stance. Reserve aggregates, such as monetary base, non-borrowed base and (nonborrowed) reserves, are examples of policy instruments. The instrument the Fed focused on in recent decades is the short-term interest rate, also known as the federal funds rate (Mishkin, 2010). The federal funds rate has a stronger link with for example the inflation goal than reserve aggregates. In December 2008, after having lowered the federal funds rate in a series of ten moves towards a range between 0% and 0.25%, the ‘zero lower bound’ was reached. The federal funds rate thus could no longer be used as a monetary policy instrument (Cecioni et al., 2011).

Having explained the tools and instruments to conduct monetary policy it is now needed to identify the different channels, also known as transmission mechanisms, that explain how changes in monetary policy feed through to the real economy. Mishkin (2010) distinguishes three main categories of transmission mechanisms, which contain several transmission channels. The first category is that of traditional interest-rate channels, the second is that of asset price channels (excluding interest rates) and the last is the credit view. The credit view operates through asymmetric information effects on credit markets. Figure 2 in appendix 5 shows all transmission channels identified by Mishkin (2010). The most important channels will be explained here.

The first transmission mechanism operates through the interest-rate channels. The theory of the interest-rate channel states that expansionary monetary policy decreases the short-term nominal interest rate and the real interest rate, causing a rise in current investment and consumption spending. The fact that expansionary monetary policy that lowers the short-term nominal interest rate also lowers the real interest rate is due to the phenomenon of sticky prices, a situation in which the aggregate price level adjusts slowly over time. The Fisher equation3 explains why the interest rate channel can still stimulate the economy when the zero lower bound has been reached because it works through the real interest rate, which in turn affects spending (Mishkin, 2010).

The second transmission mechanism operates through the asset price channels. The first of these channels that Mishkin (2010) mentions, is the effect of exchange rates. The dollar will depreciate when the real interest rate on domestic assets falls. Because the dollar becomes cheaper, exports will increase and hence aggregate output will increase. The second channel to look at is that of wealth effects. In this case consumption spending is considered to

3_{Fisher equation: i}

r= i – πe , the real interest rate is equal to the nominal interest rate minus the

(12)

be the lifetime resources of consumers. Expansionary monetary policy can increase the stock prices, thereby increasing the wealth of consumers and hence their lifetime resources. This results in higher consumption and aggregate output. The last channel to look at is the channel that operates through Tobin’s q. Tobin’s q is the ratio of the market value of firms to the replacement cost of capital. Again, expansionary monetary policy can cause stock prices to rise, which causes an increase in q and thus an increase in investment and aggregate output.

The third transmission mechanism is the credit view, which mainly operates through the bank lending channel and the balance sheet channel. Both channels emphasize the effect of monetary policy on the availability of credit. For the bank lending channel the expansionary monetary policy increases the bank deposits and therefore the bank loans available, which increases investment and consumption spending. The balance sheet channel explains the outcome of increased investment and consumption spending through a reduction of adverse selection and moral hazard problems (Mishkin, 2010).

3.2 Unconventional monetary policy

In the introduction of this chapter it was already mentioned that conducting monetary policy became much more difficult during the recent financial crisis. Therefore, the Fed had to start using unconventional monetary policy. In general, there are two reasons for doing so. The first reason is that the short-term nominal interest rate has been cut to zero; the zero lower bound is reached. The second reason is that the transmission mechanisms are severely impaired (this can be a reason even if the zero lower bound has not been reached) (Bini Smaghi, 2009). Which tools to use in these extraordinary circumstances depends on institutional characteristics, the banking sector and the types of shocks that hit the economy. Bini Smaghi (2009) uses the following definition for unconventional monetary policies: ‘Those policies that directly target the cost and availability of external finance to banks, households and non-financial companies’. In this section the main unconventional monetary policies will be discussed and linked to the unconventional monetary policy of the Fed. Three main unconventional monetary policy tools will be addressed. The first is market expectations, the second quantitative easing and the last is credit easing. The last section of this chapter will elaborate on two transmission channels that are important when discussing unconventional monetary policy.

(13)

3.2.1 Unconventional monetary policy measures

In order to provide monetary stimulus when the short-term interest rate has reached the zero lower bound, the central bank can use market expectations to guide medium to long-term interest rate expectations. There are several ways to use market expectations to conduct monetary policy. The central bank can commit to a low policy rate for an extended period of time. If credible, it should flatten the yield curve, reduce real long-term interest rates and increase aggregate demand (Blinder, 2010). This is also known as forward guidance, which has been used several times by the Fed during the recent financial crisis. Another use of market expectations is influencing expected inflation through which the real interest rate can be affected (Bini Smaghi, 2009).

Besides using market expectations there are two other unconventional monetary policy measures to provide a monetary stimulus when the zero lower bound has been reached. These are quantitative easing and credit easing, used to either change the composition or the size of the central bank’s balance sheet. In the literature on the Fed’s unconventional monetary policy often no distinction is being made between quantitative or credit easing and they are both referred to as quantitative easing (Blinder, 2010). This section will make a distinction between the two.

Quantitative easing is an open market purchase of assets when the zero lower bound is reached. The focus of quantitative easing is on the term structure by buying longer-term government bonds from banks. Quantitative easing is only useful when the short-term interest rate has reached or is close to zero because then banks will not choose to use the received liquidity to increase their reserves at the central bank (Bini Smaghi, 2009).

Credit easing focuses on risk or liquidity spreads. The central bank can buy a risky or less-liquid asset by selling treasuries and thus change the composition of its balance sheet. Alternatively, the central bank can increase the monetary base and thus change (increase) the size of its balance sheet (Blinder, 2010). Credit easing can target specific markets and also differs from quantitative easing in that it can be used when the zero lower bound has not been reached.

3.2.2 Signaling and portfolio balance channel

When explaining unconventional monetary policy and its international effects, two additional transmission channels have to be introduced. These are the signaling channel and the portfolio balance channel. To explain these channels equation 1 will be used.

(14)

𝑌_!! _{= 𝑛}!! _𝐸 ! !!!

!!! 𝑟!!!+ 𝑌𝑇𝑃!! (1)

This equation is taken from Bauer and Neely (2014). The first term on the right-hand side is the average expected overnight rate taken over the next n periods, where rt+1 is the policy rate.

The second term, YTPtn, is the yield term premium. These two terms together form the yield,

Ynt , which is present at time t on a bond for n periods. The signaling channel affects

long-term interest rates through direct or indirect signals regarding the future path of policy rates (Bauer and Neely, 2014). The signaling channel thus affects the expectation of the interest rate in equation 1. During the recent financial crisis the Fed has used direct signaling in the form of forward guidance. Asset purchase announcements can function as indirect signaling. It is important to notice that asset purchase announcements partly offset the signal of a decrease in the expected policy rate.

Where the signaling channel affects the expected future interest rates, the portfolio balance channel affects the term premium for long-term interest rates. LSAPs affect the composition of outstanding government securities and their maturities, which eventually determine the risk premia for long-term interest rates. The portfolio balance channel overstates the change in term premia because the signaling channel also partly causes a decrease in term premia (Bauer and Neely, 2014). In this thesis the signaling channel is defined as the channel that affects the expected future interest rate and the portfolio balance is defined as the channel that affects the term premium.

In the previous chapter the literature on international effects of the unconventional monetary policy by the Fed in the recent financial crisis was discussed. In addition, empirically proved ways of how the U.S. domestic monetary policy has affected international economies were explained. This chapter focused on the monetary policy of the Fed. The next chapter will present the empirical model through which the research question of this thesis will be answered. .

4. Empirical analysis

To answer the research question, empirical analysis is needed. The first part of this chapter will discuss some hypotheses. This will be followed by an explanation of the empirical method that will be used to test if these hypotheses are correct. The data that will be used are presented in the final section of the chapter.

(15)

4.1 Hypotheses and empirical method 4.1.1 Hypotheses

Taking the results of previous literature into consideration I will state two hypotheses. The first hypothesis is about the effect of the unconventional monetary policy of the Fed on the Eurozone economies in general. Taking the previous literature into account the hypothesis is that the effect of unconventional monetary policy measures of the Fed on Eurozone economies will be significant.

The second hypothesis is about the channel through which the unconventional monetary policy of the Fed affects the economy of the Eurozone. Bauer and Neely (2014) found non-negligible signaling effects for Germany. This is why the second hypothesis is that the economies of Eurozone countries will be significantly affected by the signaling channel.

To test these two hypotheses an empirical method is needed, which will be explained in the next section.

4.1.2 Empirical method

To answer the research question, three empirical approaches will be used. All approaches are event studies. The event study approach has been used in previous literature on unconventional monetary policy. Unconventional monetary policy by the Fed and its success have been analyzed by Gagnon et al. (2010) using event studies. Using the same approach, Joyce et al. (2010) examined unconventional monetary policy measures taken in the UK. Bauer and Neely (2014) also use the event study approach. This thesis also uses the event study approach as it is a common method to analyze unconventional monetary policy. In the section describing the data I will discuss the use of event study analysis in more detail by arguing that unconventional monetary policy effects largely depend on the time-frame chosen to analyze them. All three event study-based empirical approaches that will be used are discussed in this section.

The first approach consists of two parts. The first part is a purely descriptive analysis. By graphing the 10-year government bond yield over time, it can be shown for each Eurozone country separately whether a response occurred around the announcement dates of the Fed. To see whether there is a difference in the 2-year government bond yield responses and hence whether the signaling channel has played a role in the effect of the announcement of the Fed, the 2-year government bond yield will be graphed as well. In chapter 3 the following equation was used to explain how to see whether the signaling channel played a role by looking at the 2-year government bond yield:

(16)

𝑌_!! _{= 𝑛}!! _𝐸 ! !!!

!!! 𝑟!!!+ 𝑌𝑇𝑃!! (1)

As explained, the first term on the right-hand side is the average expected overnight rate taken over the next n periods, where rt+1 is the policy rate. The second term, YTPtn, is the yield term

premium. These two terms together form the yield, Ynt , which is present at time t on a bond

for n periods.Short-term yields are used to look at the effect of the signaling channel because in the short run the term premium is mostly stable. As was already pointed out in chapter 3, the signaling channel affects the expected future interest rates and the portfolio balance channel affects the term premium. Large changes in the 2-year government bond yield indicate that a change in expected short-term rates played a significant role in the change of the long-term government bond yield (Bauer and Neely, 2014).

In order to get graphs that are easier to interpret, only the yields of the following five Eurozone countries will be graphed: Germany, Finland, France, the Netherlands and Spain. I have chosen these countries because they have reacted differently to the unconventional monetary policy announcements of the Fed. In addition, I will only take into account six announcement dates in the descriptive analysis, more specifically the announcements on November 25th 2008, December 16th 2008, January 28th 2009, August 10th 2010, September 13th_{2012 and December 18}th_2013.

The second part of the first approach is an event study approach for which dummies are being used. The fourteen dummies capture the response of the 10-year and 2-year government bond yield on the fourteen dates that the Fed announced unconventional monetary policy measures. I will use the following formula:

Yt= β0 + β1UCADummy1 + β2UCADummy2 + β3UCADummy3 + β4UCADummy4 + β5UCADummy5 + β6UCADummy6+ β7UCADummy7 + β8UCADummy8

+ β9UCADummy 9+ β10UCADummy10 + β11UCADummy11+ β12UCADummy12+

β13UCADummy13+ β14UCADummy14+ εt (2)

UCADummy1 (Unconventional Announcement Dummy 1) will be the variable for the

announcement date 25/11/08. This means UCADummy1 will be 1 at the 25th of November

(17)

date 1/12/2008, so 2/12/20084. Each dummy that follows will be the dummy for the day after the next selected announcement date in chronological order. The coefficients will be estimated for each separate Eurozone country by Ordinary Least Squares (OLS). The dummy coefficients show if the coefficient for the change in the 10-year or 2-year government bond yield on the announcement date itself (for the 25/11/08-announcement) or the day after the announcement, was significant. When the coefficients for the dummies are 95% or 99% significant, it can be concluded that the change in the 10-year and 2-year government bond yield is significant.

The second approach is a stationary time series analysis. I have introduced this stationary time series analysis in addition to the dummy regression because it can better capture the LSAP and forward guidance effects. This will be discussed more extensively in the last part of this section. By not only computing a dummy regression but also introducing a stationary time series analysis I will try to capture the LSAP and forward guidance effects more carefully.

Yt is a stationary time series function if there is no change over time in its probability

distribution (Stock and Watson, 2007). By looking at the change of the 10-year government bond yields of the Eurozone countries around the announcement of the unconventional monetary policy measure, I will try to establish whether there was a significant change. To find the significance of these changes in the bond yields I will use a time series regression known as autoregressive-moving average (ARMA) model to forecast the change in the 10-year government bond yield for the day after the announcement. Where a pth order autoregressive model only represents Yt as a linear function of p and its lagged values, the

ARMA model also models ut as serially correlated. Ut is modelled as a distributed lag, also

known as moving average of another unobserved error term (Stock and Watson, 2007). Formula 3 is an example of what a pth order autoregressive model would look like. Formula 4 is an example of what the ARMA model will look like.

𝑌_! = 𝛽_!+ 𝛽_!𝑌_!!!+ 𝛽_!𝑌_!!!+ ⋯ + 𝛽_!𝑌_!!! + 𝑢_! (3)5

𝑌_! = 𝛽_!+ 𝛽_!𝑌_!!!+ 𝛽_!𝑌_!!!+ ⋯ + 𝛽_!𝑌_!!! + 𝑢_!+ 𝑢_!!!+ 𝑢_!!!+ ⋯ + 𝑢_!!! (4)

4_{In the data section later in this chapter, it is explained why only for the date 25/11/08 the reaction of}

the government bond yield on an announcement of the Fed will be measured on the same day as the announcement. For all the other announcements the reaction of the government bond yield will be measured on the day after the announcement date.

(18)

First the coefficients for the ARMA model will be calculated for each Eurozone country per announcement of the Fed. The number of lags of the government bond yield and the number of lags of the error term for a specific Eurozone country will be determined by looking at the correlogram and partial correlogram. After determining the number of lags for the yield and error term for a specific Eurozone country, there will be an additional check by predicting the residuals and by looking at the correlogram and partial correlogram of the residuals. When the number of lags has been determined correctly, the coefficients for the yield are calculated using the ARMA model. This is followed by a two-sided t-test for the significance of the deviation of the actual 10-year government bond yield change from the expected yield change with respect to the timeframe that includes the day of the announcement, the day after the announcement and the day after that.

The third approach focuses on the channel through which the Eurozone economy is affected by the unconventional monetary policy of the Fed. As I mentioned in chapter 3 of this paper the signaling and the portfolio balance channel affect the long-term interest rates. Determining which channel dominates in affecting the long-term interest rate of a Eurozone country has been a challenge for researchers. This thesis therefore only focuses on whether the signaling channel was of great influence. To determine the effect of the signaling channel on the long-term interest rate, I will analyze the 2-year government bond yield of the Eurozone countries. Short-term yields are used because in the short run the term premium is mostly stable. Large changes in the 2-year government bond yield indicate that a change in expected short-term rates played a significant role in the change of the long-term government bond yield (Bauer and Neely, 2014). Earlier in this section, I explained that the descriptive analysis and the dummy regression will also be conducted for the 2-year government bond yield. This is also the case for the stationary time series analysis and thus for the ARMA model. The ARMA model calculations for the 2-year government bond yield will be done in the same way as the calculations for the 10-year government bond yield. So by using the ARMA model and a two-sided t-statistic I will establish whether the signaling channel was of significant influence in channeling the effect of the unconventional monetary policy of the Fed to the Eurozone economies.

This paper analyzes whether the effect of the unconventional monetary policy announcements on the 10-year and 2-year government bond yields of the Eurozone countries is significant, but as explained by Bauer and Neely (2014) it is very difficult to estimate the

(19)

announcements were expected and this build up makes the total effect of the next LSAP announcements difficult to estimate. This does not mean that it is not possible to analyze the effect of a LSAP announcement and its significance, but it should be taken into account that the effect of a LSAP may not be fully captured in the event window. In this paper it is implicitly assumed that all effects on the 10-year and 2-year government bond yield occur during the time that market participants adjust their expectations and hence that markets are efficient. I also assume that an event that is being used in this paper incorporates all effects of the announcement on the future LSAP and that the expectations with respect to this LSAP have only been affected by these announcements (Gagnon et al., 2011).

Before applying the empirical method to obtain the results, the data that will be used are illustrated in the next section.

4.2 Data

There are data on fourteen announcements by the Fed that are presented in table 4 (see appendix 1). Table 4 was created by analyzing the press releases of the Fed, the minutes of the FOMC and by using literature already available. All the announcements in table 4 that are in italics will be included in the empirical analysis to analyze the effect of the announcements of the Fed on the economies in the Eurozone. These announcements have been chosen because they had a great influence domestically, were not affected by other announcements both domestically and internationally or were covered in previous literature meaning that the results can be compared.

On November 25th 2008 there was a press release by the Fed announcing the purchase of $100 billion in agency debt and $500 billion in MBS. This was the start of the first large-scale asset purchase during the recent financial crisis. This LSAP was also known as Quantitative Easing I (QE I) (Bauer and Neely, 2014). The next important date to include when looking at the effect on international yields is December 1st 2008 when the chairman of the Fed announced that the Fed was able to purchase long-term Treasuries. Not long after that on December 16th 2008 the Fed stated that it expected the federal funds rate to be of ‘exceptionally low levels for some time’. This statement and that of January 28th 2009 also included the announcement that the Fed was considering to purchase more Treasuries and MBS. In March 2009 the Fed specified this announcement by releasing the amounts of long-term Treasuries ($300 billion), agency debt ($200 billion) and MBS ($1.25 trillion) it expected to purchase (Gagnon et al., 2011). A period of increasing purchases of securities by the Fed, was followed by a period during which the purchase of securities was slowed down.

(20)

On August 12th 2009 the FOMC announced a slowing down of Treasury purchases followed by an announcement of a slowing down of agency debt and MBS purchases on September 23rd 2009. On the 4th of November 2009 the FOMC eventually announced that the purchase of agency debt would end at $175 billion instead of $200 billion (Neely, 2013).

On August 10th 2010 the Fed started what is now known as Quantitative Easing II (QE II). An FOMC statement that was released on the 21st of September 2010 revealed that the financial crisis was not over yet because the committee believed that ‘inflation is likely to remain subdued for some time before rising to levels the Committee considers consistent with its mandate’ (FOMC, 2010). As part of QE II an additional purchase of $600 billion in Treasury securities was announced on November 3rd 2010. The predictions by some members in the minutes of the FOMC of the 22nd of August 2012, introduced Quantitative Easing III (QE III). Some members mentioned that additional monetary accommodation would be provided in the near future. This was specified in a press release on September 13th 2012, which stated that if the labor market would not improve, the Fed would start purchasing $40 billion MBS per month (Bauer and Neely, 2014). The last date that will be included to estimate the effect of unconventional monetary policy measures of the Fed on the economy of the Eurozone is the 18th_{of December 2013 (FOMC, 2013).}

For the empirical analysis of the announcements just described, daily 10-year government bond yields and daily 2-year government bond yields are needed from the following countries: Belgium, Germany, Finland, France, Greece, Ireland, Italy, Luxembourg, the Netherlands, Austria, Portugal, Spain, Slovenia, Malta and Slovakia. Currently Latvia is also part of the Eurozone but it only entered the Eurozone in 2014. Because the major announcements of the Fed used in this thesis include a period up to and including 2013, Latvia will not be included in the empirical analysis. Cyprus and Estonia will not be included either, because of lack of relevant data.

The timing of the data that are being used is crucial to examine the effect of the unconventional monetary policy of the Fed on the economies of the Eurozone. Most of the announcements made by the Fed were announced after the yield data of the European bond markets were already collected. For these announcements I have to look at yield changes from t to t+1. Only the LSAP announcement on the 25th of November 2008 took place at 8:15 am eastern standard time. This was before the collection of the European bond markets data, so for this announcement I can look at the yield changes of that day (Neely, 2013). In order to choose the correct window length, two things should be taken into consideration. Firstly, the

(21)

effects. Secondly, the window should not be too long to include new information that can affect yields (Gagnon et al., 2011).

The last thing to take into consideration when looking at the effect of the announcements of the Fed is the effect of announcements of the central banks of the Eurozone countries and particularly the European Central Bank (ECB). From the press releases of the ECB it became clear that four of the announcement dates of the Fed can have been affected by an announcement that the ECB did the day before, the same day or the day after. These announcements were on the 24th of November 2008, the 15th of December 2008, the 19th of March 2009 and lastly the 24th of September 2009.

5. Results and discussion

In this part the results of the empirical analysis will be presented. After introducing the descriptive graphs and the dummy regression results, the ARMA model results will be discussed.

5.1 Descriptive analysis results

The section on the empirical method already established that the first step in obtaining results about the effect of the unconventional monetary policy announcements by the Fed consists of graphing the 10-year government bond yield and the 2-year government bond yield. I will first consider the 10-year government bond yield. The next six graphs present the 10-year government bond yield for the ten days around the six announcement dates that were chosen in the empirical method.

(22)

Graph 2: 10yr government bond yield around 16/12/08

Note: In these graphs the 10-year government bond yield (in %) of five Eurozone countries is graphed over time.

When looking at the graphs on this page, a few key findings can be stated. Firstly, there is a difference in response between the individual Eurozone economies. The reaction of the 10-year government bond yield of Finland after the announcement of the Fed on the 10th of August 2010 is especially pronounced. This might have something to do with a domestic announcement about the domestic economy around that date. Secondly, although the response

(23)

announcement on the 18th of December 2013). Also, regarding the size of the reaction of the countries to the announcements, the reactions to the 25/11/08-, 16/12/08- and 10/08/10-announcement are clearly visible whereas the reactions to the 13/09/12- and 18/12/13- announcement are less pronounced, even with different scaling. Finally, the changes in the 10-year government bond yield appear to start the day before the announcement or the announcement date itself and the change in the yield continues until the day after the announcement date. After that, the 10-year government bond yield stabilizes again.

I will now proceed with the descriptive analysis of the 2-year government bond yield.

(24)

Graph 11: 2yr government bond yield around 13/09/12 Graph 12: 2yr government bond yield around 18/12/13

Note: In these graphs the 2-year government bond yield (in %) of five Eurozone countries is graphed over time.

The graphs of the 2-year government bond yield do not explicitly show different reactions to the six announcement dates selected for the descriptive analysis compared to the reactions observed for the 10-year government bond yield. But the reaction of the 2-year government bond yield to the 25/11/08-announcement is less pronounced in comparison to that of the 10-year government bond yield. The ranges of the graphs of the first four announcements are almost the same for both the 10-year and 2-year government bond yield. The only difference is that the yield of the 2-year government bond is lower. Hence, these four graphs can be easily compared. Only the ranges of the graphs of the 10-year and 2-year government bond yield for the announcements on 13 September 2012 and 18 December 2013 differ. But even though they differ, both graphs show that the responses of the Eurozone countries are not that big. Three more things stand out based on the graphs. Firstly, the reactions of Finland to the announcement of the Fed seem to coincide more with the reactions of the other Eurozone countries when comparing the 2-year government bond yield to the 10-year government bond yield. Secondly, with regards to both the 10-year government bond yield and the 2-year government bond yield, Spain has deviated from the other Eurozone countries from the year 2010 onwards. Lastly, the changes in the 2-year government bond yield also appear to start the day before the announcement or the announcement date itself and the change in the yield continues until the day after the announcement date, as was the case for the 10-year government bond yield. After this the 2-year government bond yield stabilizes again.

5.2 Results econometric approaches

(25)

5.2.1 Dummy regression results

In appendix 2 the results of the dummy regressions can be found. In table 5 and 6 the results of the dummy regression for the 10-year government bond yield are stated. These first two tables of the appendix show that seven of the fifteen Eurozone countries in the tables have had a change in their 10-year government bond yield on the day after an announcement date which was 95% or 99% significant, in three or more cases. Three of those seven countries had a significant change in their yield after four or more announcement dates. After the unconventional monetary policy announcements on the 16th of December 2008, the 10th of August 2010, September 21st 2010 and the 13th of September 2012 five or more of the Eurozone countries had a significant change in their 10-year government bond yield.

For the dummy regression of the 2-year government bond yield the results are less significant than for that of the 10-year government bond yield. The results can be found in table 7 and 8 in appendix 2. These tables show that two of the fifteen Eurozone countries have had a significant change in their 2-year government bond yield in three or four cases on the day after an announcement date. After the unconventional monetary policy announcements on August the 22nd 2012 three Eurozone countries had a significant change in their 2-year government bond yield and after the 13/09/12-announcement five Eurozone countries had a significant change in their 2-year government bond yield.

5.2.2 ARMA model results

In table 1, table 2, appendix 3 and appendix 4 the results of the ARMA model are presented. Table 1 on the next page and table 9 in appendix 3 show the t-values of the fifteen Eurozone countries for the different announcement dates, having used the 10-year government bond yield. Table 2 on page 27 and table 10 in appendix 3 show the t-values of the fifteen Eurozone countries for the different announcement dates, having used the 2-year government bond yield.As explained before, I first looked at the correlogram and partial correlogram in order to determine the number of lags of the government bond yield and the number of lags of the error term for a specific Eurozone country. After determining the number of lags for the yield and error term for a specific Eurozone country, I checked them by predicting the residuals and looking at the correlogram and partial correlogram of the residuals. When the number of lags was determined correctly, the coefficients for the yield were calculated using the ARMA model. This was then followed by a two-sided t-test to test the significance of the deviation of the actual 10-year government bond yield change from the expected yield change with respect to the timeframe that includes the day of the announcement, the day after

(26)

the announcement and the day after that. This t-value from the two-sided t-test is shown in table 1 and 9. In appendix 4 the coefficients for the five countries used in the descriptive analysis (Germany, Finland, France, the Netherlands and Spain) for the 16/12/08-, 10/08/10- and 13/09/12- announcement are stated. For both the 10-year and 2-year government bond yield these ARMA coefficients are presented in appendix 4.

Because I used a two-sided t-test the significance of the |t|-value has to be 1.64 for a 90% significance, 1.96 for a 95% significance and 2.58 for a 99% significance. All t-values that are at least 90% significant have been marked in tables 1,2,9 and 10. The t-value for the 10-year government bond yield is at least 95% significant for 49 of the 192 results in table 1 and 9. Belgium, Finland, Greece and Luxembourg have a 95% or 99% significant t-value for the day of and days after an announcement for at least five of the fourteen announcements. The announcements on the 2nd of December 2008, 24th of September 2009, 22nd of September 2010 and 23rd of August 2012 resulted in a 95% or 99% significant t-value for the 10-year government bond yield for six or more countries.

Table 1: T-test 10 year (1)

This table shows the 10-year government bond yield ARMA model t-values for nine Eurozone countries and 14 announcement dates. The ***, ** and * stand for respectively 1%, 5% or 10% significance.

Belgium Germany Finland Greece France Ireland Italy Netherlands Austria

25/11/08 -0.9674 -1.0098 -1.4457 0.2733 -0.7287 -0.6619 -1.6648* -0.4964 -0.8830 01/12/08 -3.3849*** -2.2655** -20.4100*** -11.5816*** -1.5449 -0.6684 -1.7130* -1.4004 -7.3936*** 16/12/08 -15.5541*** -0.9071 -2.1797** -0.0262 -1.2320 -1.3202 -1.2241 -1.2466 -3.2477*** 28/01/09 -1.9489* 0.7389 -0.2141 -0.9594 -0.0468 -0.7300 -1.2336 0.2099 -0.4360 18/03/09 -1.0813 -0.6835 -0.5892 -6.3636*** -1.0209 -1.2609 -1.3082 -0.1660 -0.8140 12/08/09 -2.4537** -0.6654 -0.9083 0.3573 -1.6640* -0.4803 0.3411 -0.2367 -2.7009*** 23/09/09 -1.6779* -4.1115*** -1.0133 -2.9030*** -5.1584*** -5.4542*** -2.1326** 5.5443*** -0.5314 04/11/09 0.4232 2.5134** 4.1866*** 2.3922** 3.6049*** 0.8208 1.2855 0.1420 1.6781* 10/08/10 -0.9049 -0.8691 -2.4883** -0.4600 -0.4697 1.8159* 0.4951 -0.8086 -1.4106 21/09/10 -2.1728** -1.4758 1.1232 -6.6193*** -6.7394*** 0.2957 -0.7708 -2.4514** -0.7984 03/11/10 -0.2840 -0.9128 2.3023** 0.4390 -0.4170 3.0154*** 0.7324 -0.4666 -5.8592*** 22/08/12 -4.7385*** -2.9759*** -1.5525 -1.0384 -1.9568* -0.3788 1.3422 -4.0651*** -1.2082 13/09/12 0.5546 0.3871 0.3873 -1.0699 0.3387 -0.8993 -5.6812*** 0.4499 0.7966 18/12/13 -0.6423 0.2406 1.4277 -1.1209 3.4288*** 0.1611 1.0053 -0.3804 1.0267

Taking into account the values of the t-test with a 90%, 95% and 99% significance, the significant t-values for the 2-year government bond yield per announcement date and country have been marked in table 2 and 10. For the 2-year government bond yield the t-value is at least 95% significant in 45 of the 198 results in table 2 and 10. France, Ireland and Slovenia

(27)

least five of the fourteen announcements. The announcements on the 29th_{of January 2009,}

13th of August 2009 and 4th of November 2010 resulted in a 95% or 99% significant t-value for the 2-year government bond yield for six or more countries.

In appendix 4 the coefficients for both the 10-year and 2-year government bond yield for the five countries used in the descriptive analysis (Germany, Finland, France, the Netherlands and Spain) with respect to the 16/12/08-, 10/08/10- and 13/09/12- announcement are stated. For the calculations of the ARMA model, 25% or more of the coefficients for the 10-year government bond yield are significant for three countries, 33% or more of the coefficients for the 10-year government bond yield are significant for four countries and 50% or more of the coefficients for the 10-year government bond yield are significant for five countries. For the 2-year government bond yield 25% of the coefficients are significant for three countries, 33% or more of the coefficients are significant for three countries and 50% or more of the coefficients for the 2-year government bond yield are significant for seven countries.

Table 2: T-test 2 year (1)

This table shows the 2-year government bond yield ARMA model t-values for nine Eurozone countries and 14 announcement dates. The ***, ** and * stand for respectively 1%, 5% or 10% significance.

Belgium Germany Finland Greece France Ireland Italy Netherlands Austria

25/11/08 0.6340 0.7118 4.2374 -1.2954 0.8572 -4.5241 -0.7098 1.3768 -0.6976 01/12/08 0.6665 0.7118 4.2374*** -0.5972 1.0129 1.2936 -0.7866 1.1085 -0.3763 16/12/08 -0.5793 0.3060 -1.2949 1.1064 -0.5606 0.1743 0.5114 1.7547* 2.0035** 28/01/09 -2.5344** -3.1472*** -1.5555 -5.2091*** -2.9352*** -1.4289 -3.7803*** -1.7400* -0.5123 18/03/09 -0.4844 -0.0923 -0.5046 -1.0731 -0.8128 -0.9085 -0.9085 -0.3499 -1.0239 12/08/09 -2.2992** -1.8340* -3.0112*** -2.0052** -13.6038*** -3.1275*** -2.1245** -1.5938 2.6746*** 23/09/09 -0.4041 -1.8295* -0.7045 -1.7843* -0.6725 -2.3294** -1.1254 -0.6611 1.0753 04/11/09 0.2147 1.7285* 0.9539 1.0682 0.0716 1.0671 0.7939 -0.3788 0.7638 10/08/10 -1.1721 -2.9017** -2.2299** -0.5444 -3.3869*** 2.6664** 0.8603 -2.7422*** 0.5203 21/09/10 -4.9899*** -1.3479 -1.2406 -1.5020 -9.1242*** 1.0468 -0.9632 -2.0893** 0.1628 03/11/10 1.1659 -0.9438 -1.2382 3.3001*** -10.1672*** 2.1979** 1.1934 -1.3680 -1.3796 22/08/12 0.1427 1.4594 -0.5118 -0.5877 -0.3814 0.1640 1.7503* 0.8027 0.0013 13/09/12 -1.2197 0.8609 0.8992 1.1230 0.7913 -0.9880 1.4989 1.1368 2.0691** 18/12/13 -0.3262 1.4239 -0.2083 -0.6600 -1.1248 -3.7992 -0.2569 -0.0326 -0.0749 5.3 Discussion

In this section the results will be discussed. I will interpret the results and discuss what the results mean with respect to the hypotheses. I will also state what has to be taken into consideration when assessing these results.

(28)

5.3.1 Descriptive analysis and dummy regression results

The model-free approaches already present evidence that supports my first hypothesis that the effect of unconventional monetary policy measures of the Fed on Eurozone economies will be significant. Although it is not possible to determine whether the response is significant with this approach, the descriptive analysis does show that the 10-year government bond yield and, although less pronounced, the 2-year government bond yield of Eurozone countries did react to the unconventional monetary policy announcements of the Fed. This strengthens the statement made in my hypothesis that the effect of the unconventional monetary policy measures of the Fed on the Eurozone economies will be significant. The descriptive analysis showed less pronounced reactions in the 2-year government bond yield to the announcements of the Fed. Therefore these results do not necessarily strengthen the hypothesis that the economies of the Eurozone countries will be significantly affected through the signaling channel.

The second part of the results consists of the dummy regression results. The coefficients of the dummies indicate whether the effect of the announcements of unconventional monetary policy by the Fed is significant for a Eurozone country. In these results the first hypothesis stated in my empirical method is also strengthened. For the announcements of the Fed on the 16th of December 2008, the 10th of August 2010, the 21st of September 2010 and the 13th of September 2012, the day after the announcement the change in the 10-year government bond yield was 95% or 99% significant for five or more countries. Belgium, Germany, France, the Netherlands, Austria, Portugal and Luxembourg had the largest number of significant reactions in their 10-year government bond yield on the days after announcements of the Fed. Bauer and Neely (2014) already found a significant effect of the unconventional monetary policy announcements by the Fed on the German economy, but now I also found three or more significant effects on the economies of Belgium, France, the Netherlands, Austria, Portugal and Luxembourg with this relatively easy regression.

The dummy regression results for the 2-year government bond yield do not show enough significant changes after an announcement to be able to say that these results underline my second hypothesis stating that the signaling channel was of importance in affecting the Eurozone economies. Only Ireland and Luxembourg show three or more significant reactions the day after an unconventional monetary policy announcement by the Fed in their 2-year government bond yield. Only the announcement on the 13th of September 2012 had a significant effect on the 2-year government bond yield of five Eurozone countries

(29)

announcement. Although I cannot make a general statement with these results that the signaling channel played a significant role in channeling the effect of the Fed announcement to the Eurozone economy, there are some individual Eurozone countries for which I can make this statement for particular announcement dates. For example, it is interesting to see that for the announcement on September 13th 2012 five Eurozone countries reacted significantly with respect to changes in their 10-year government bond yield as well. Four countries both had a significant reaction in their 2-year and 10-year yield: Germany, France, the Netherlands and Luxembourg. It can be concluded from the dummy regression that the economies of these four countries were significantly affected by the 13/9/12-announcement and this effect was significantly channeled through the signaling channel. Again, Bauer and Neely (2014) already found that the unconventional monetary policy of the Fed has significantly affected the German economy. They also found that the effect was significantly channeled through the signaling channel. With this dummy regression it can be said that I also found this result for the 13/09/12 announcement specifically. In table 3 all other combinations of countries and dates for which it can be concluded from the dummy results that their economies were significantly affected by a particular announcement and that this effect was significantly channeled through the signaling channel are presented.

Table 3: Significant effect through signaling channel (dummy results)

This table shows the 10-year and 2-year government bond yield dummy results for the combinations of Eurozone countries and announcement dates for which the significant effect of an unconventional monetary policy announcement by the Fed was significantly channeled through the signaling channel. The ***, ** and * stand for respectively 1%, 5% or 10% significance.

Announcement date Germany France Netherlands Austria Portugal Luxembourg

16/12/08 10yr dummy -0.0337*** 16/12/08 2yr dummy -0.0529*** 10/08/10 10yr dummy -0.0317*** 10/08/10 2yr dummy -0.0473*** 03/11/10 10yr dummy _0.0605** 03/11/10 2yr dummy _0.0972*** 22/08/12 10yr dummy _-0.0273** _-0.0319** _-0.0228** 22/08/12 2yr dummy -0.110*** -0.145** -0.0536*** 13/9/12 10yr dummy 0.0685*** 0.0310*** 0.0486** 0.0350*** 13/9/12 2yr dummy 0.335** 0.143*** 0.213** 0.0595***

The dummy results have strengthened my first hypothesis that the effect of the unconventional monetary policy measures of the Fed on the Eurozone will be significant. The second hypothesis about the fact that the Eurozone countries are affected through the