Quantitative Easing by the European Central Bank : has the Asset Purchasing Programme of the European Central Bank had any effects on the interest rates of long-term bonds?

Quantitative Easing by the European Central Bank

Has the Asset Purchasing Programme of the European Central Bank had any

effects on the interest rates of long-term bonds?

Bachelor thesis Economics and Business Economics Student: Alex Wenker Student Number: 10810544

Statement of Originality

This document is written by student Alex Wenker who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are 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.

Table of Content Page number:

1 Introduction 4

2 Literature Review 5

3 Methodology and Empirical Analyses 12

4 Conclusion and Discussion 18

5 References 20

Introduction

Recently the European Central Bank (ECB) has announced that it would continue with its asset-purchasing programme (APP) that was started in March 2015 (European Central Bank, 2017). In a press release the ECB stated that the programme was necessary since the desired path of inflation that is consistent with an inflation-rate aim of less, but close to 2% has not yet been achieved. There is however a slight change in its policy: the monthly value of assets that are purchased will drop from 60bln to 30bln euro until at least September 2018. This policy of buying large amounts of (long-term) guilt paper and thereby increasing the money supply is commonly known as Quantitative Easing (QE) (Mankiw, 2012). The reason to implement this was because never in the ECB’s history, the expected inflation rate was as low as then, which caused a great risk of deflation (European Central Bank, 2015). To counter this, billons of asset backed securities, covered bonds and sovereign bonds (all with a longer maturity than two years) were purchased the last couple of years to stimulate the economy within the euro area and bring back the inflation rate close to 2%. When observing path of inflation over the last few years, it seems that after a rate of approximately zero, everything is on its way back to its obtained level (close but below 2%). Nevertheless this does not mean that this is due to the implementation of the APP, various other factors could have been of great influence. The aim of this paper is to find out if the in 2015 started QE programme of the ECB has had any effects on the interest rates of long-term securities. This will be answered by on the one hand analysing economic theory and on the other hand empirically, by doing an event study on the impact of important announcements related to the APP and its influence on market interest rates. These should be influenced since especially market interest rates of longer securities (more than one year maturity) depend on market expectations about the future. Announcements regarding quantitative easing policy should then lower the yields on these securities because the interest expectations about the future will decrease. The motivation for doing an event study is because it is a very useful instrument to quantify the effect of the APP. This is because financial variables are used to measure these effects because they fluctuate on a daily basis. They are based on market beliefs about monetary policy. Changes in monetary policy can thus be of influence on financial

variables when they differ from market expectations. By analysing if such changes has occurred on specific event dates the effectiveness of a monetary programme can be discussed. Moreover in my previous literature regarding QE, event studies are used for analysing the effectiveness of QE policies, such as the paper of Gagnon et al. (2010). In the literatures review QE will be explained and its potential effects on interest rates. This section will also explain the different components of long-term bonds and how they are defined. For some additional background and the results from previous QE policies will be summarized and lastly the theoretical elements of an event study will be discussed. In the methodology and empirical analyses section the most important ECB announcements regarding the APP will be discussed. Furthermore the implementation of the various models are discussed and an explanation is given on why which variables are used. Regarding the analyses the obtained results are discussed and some limitations are acknowledged. Finally the results will be evaluated in the conclusion and discussion part of this paper. The main finding of this study is that the ECB’s quantitative easing policy has indeed had a decreasing effect on the interest rate of long-term EU bonds. Moreover it is likely that this decrease is for the most part because of a decrease in the risk-premium of these long-term bonds.

Literature Review

The main instrument that central banks use to accomplish their goal(s) is through open-market operations (OMO) (Mankiw, 2012). By buying and purchasing government bonds from the public they can influence the interest rate and control or push the economy in a certain favourable direction. For example, when a central bank buys government securities from the private market it creates scarcity of those securities, what increases their price but lower their yield, because the risk carried by the holding agent becomes less. According to Mankiw this means that interest rates will drop and the money supply will increase. Such policy discourages savings and stimulates consumption and investments in the economy, for example investments in corporate bonds. QE is a form of open-market operations but instead of buying small/regular amounts of short term treasury bills, the central bank buys large amounts of, usually,

longer term securities. These large amounts are due to the fact that QE is a policy that occurs when short-term interest rates are very low or even zero, where regular amounts of OMO have little effect (Bernanke & Reinhart, 2004). This is called the zero lower bound. At this point the regular OMOs are useless since the short-term interest rate can’t shift any lower. So in achieving its goal(s), a central bank loses its main instrument for that because of the zero lower bound (Mishkin, 2007). This doesn’t mean that monetary policy cannot be effective according to Bernanke & Reinhart. First of all a central bank (CB) can create low interest rate expectations among investors and assure them that short rates will be kept low for a longer period than expected. Second, by changing the variety or composition of its assets a CB is able to influence the risk or liquidity characteristics of securities that are available, with the result that there is a shift in demand which causes a change in the securities’ prices. Furthermore Bernanke & Reinhart argue that a CB can expand the size of it balance sheet (QE). Since the ECB was in a situation of the zero lower bound and then implemented a balance sheet expansion programme, the APP, this is the most relevant option for this paper to analyse. Bernanke and Reinhart state that QE can affect the economy in different ways. First of all through portfolio substitution effects, or as Tobin (1969) called it “the portfolio balance effect”, where money is seen as an imperfect substitute for other financial assets. The large increase in the money supply due to QE makes investors changes their portfolio balance what raises prices and lowers yields on alternative financial assets. The effect is thus the same as with the earlier discussed regular OMO’s, only now the result is for the most part generated by the huge amounts of money that enters the market, and not by the decrease in the short-term interest rate, since this is relatively fixed at the zero lower bound. Furthermore Bernanke and Reinhart argue that QE can contribute as a monetary policy instrument by downgrading expectations about the future path of policy rates. This differs from the previous mentioned policy rate commitment the central bank can assure investors in the following way: By committing itself to keep its reserves rate high and thereby signal that the short-term interest will stay low for the future, the CB gives a more credible and visible signal to investors than only a verbal promise. At last QE can be effective because of its expansionary fiscal effects according to Auerbach and Obstfeld (2005). In their paper they state that the probability that a fully

permanent liquidity trap (zero-lower bound) is quite implausible. Market participants thus expect an increase of the short-term interest rate somewhere in the future, because of the existence of government debt in the market, what causes them to believe that current or future tax will rise. A QE policy will transfer a lot of interest costs of the government from the market to the CB if the CB commits to its new higher level of reserves for a longer period. By doing so the expected tax burden for the public will decline and what will cause public spending to increase. This thus results in a fiscal transfer where a direct tax e.g. on labour is replaced by the inflation tax. As a conclusion Auerbach and Obstfeld even advise that when a QE policy is implemented, it would be more effective when it is accompanied by a tax cut, to initiate the fiscal transfer. The latter will not be further discussed in this paper though, since it has not been executed by the ECB with its APP (and probably never will since the ECB may not interfere with policy making). What can be noticed is that the short-term interest rate that is set by the CB has an influence on long-term assets but first of all there should be a clear definition between the already discussed types of bonds. In this paper the same maturity breakdown is being used as is by the ECB: all securities with an original maturity of one year or less are considered short-term debt securities (also considered to be risk free). Securities with a longer original maturity of one year are long-term deb securities, where these can be divided into three subcategories: Fixed interest rate debt securities, variable interest rate debt securities and zero coupon bonds issued at discount (European Central Bank, 2018). Here, the latter doesn’t have interest payments but is issued at a discount. Nevertheless the amount of discount can (for the most part) be represented by the interest rate that is collected over the lifetime of the bond. From financial theory we know that the prices and yields of long-term assets consist of two components. Firstly it is based on expectations. Financial market participants have a certain believe about the future path of short-term interest rate (Bernanke, Reinhart, & Sack, 2004). This expected path of short-term yield over time will set the minimum amount of interest the longer term asset hold by a market participant agent should pay. It determines the rate for which agents would be indifferent between short- and long-term assets if risk for all bonds were zero. Since this is never the case in practice, the yield on long-term bonds is always higher since the risk of default on those bonds is bigger, which is called the risk premium. Phrased

differently: Bernanke, Reinhart and Sack state that instead of a ‘safe’ investment with buying (risk free) short-term securities over and over, the market participant is only willing to buy a long-term security if there will be a compensation for the higher risk of default. Regarding the fixed interest rate debt securities this risk premium is substituted by the term premium. It is more focused on the future path of short-term interest rates since the long-term yield is fixed (Gagnon et al., 2010). It requires the additional return for investors over and above the expected short-term yields, to compensate for the risk of holding a long-term asset. So a part of the yield on a long-term asset is set by the current short-term interest rate and it’s expected development over the duration of the long-term asset, plus an additional premium for carrying a higher default risk. Hence, this premium is thus also based on financial market participants’ expectations about future developments (which can be monetary, fiscal, etc.). A great impact on the interest rates of assets in the market can thus be the result of a shock or policy announcement. This will be shown by examples of QE policies by the Bank of Japan (BOJ) and the Federal Reserve Bank (Fed) that are quite similar to the APP of the ECB. The results of QE by the Bank of Japan In the early 1990s Japan was faced with the burst of the asset price bubble. Due to this bubble Japan faced a threat of deflation in the late 1990s with its short-term interest rate already at the zero lower bound (Ugai, 2006). To counter this recession the BOJ implemented a QE programme in 2001, one that would last until 2006. In his research Ugai examines previous empirical analyses and papers on the QE programme of the BOJ. He first assessed what the effect of QE commitment was on the future path of short-term interest rates. Furthermore he tried to identify through which channel the new composition of the extended balance sheet would have its effect. In his results he found that the QE programme, that was committed to create a stable or increasing inflation rate, had more effect on lowering the yields of overall assets than the zero interest rate policy it had implemented to counter deflation. A second result was that little effect was noticed on narrowing the bonds’ risk premiums due to the increase of the BOJ balance sheet of Japanese government bonds. QE’s effect on macroeconomic level was a stabilizing one. Due to the BOJ’s committing policy

financial markets remained stable and further deterioration of Japan’s economy was prevented. Ugai concludes his research with the fact that the largest effect of the QE programme was on the expected future short-term interest rates due to communication. Credible commitments of the BOJ that were communicated with the private sector are critical in achieving policy effects. QE effectiveness of the Federal Reserve For a review on the recently implemented QE by the United States Federal Reserve Bank (Fed) the paper of Gagnon et al. (2010) delivers some empirical evidence. The Fed started its policy (the large-scale asset purchases (LSAPs) in November 2008 due to the financial crisis by buying up to 600 billion dollar of mortgage-backed securities and housing agency debt. A few months later, in March 2009, the Fed announced that it would expand its QE up to 1,75 trillion dollars whereas the federal funds target rate was already lowered to a range of 0-0,25 basis points. It should be clear that again this is an example unconventional monetary policy at the zero lower bound. Gagnon et al. analysed this QE policy effect on market interest rates with two different approaches: (1) they did a time series analyse on long-term Treasury supply and (2) by doing an event study on the effect of QE announcements/communications of the Fed. The latter will also be done in this paper only regarding the ECB’s APP. Concerning the Fed’s policy there are clear results. By the reduced supply of long- term assets the Fed has successfully reduced the term premium of the market. For 10-year securities the drop in the term premium was measured between 30 and 100 basis points. Due to the increased market liquidity the effect on interest rates of securities with an even longer duration was even more powerful. The yield on those assets decreased had dropped even further because of less ‘risky’ securities in the financial market. Empirical models of an event study Were Gagnon et al. performed two types of research in their study, this paper will only contain one: An event study on the ECB announcements regarding the APP. This is due to the fact that there is limited time for this paper. Since an event study like Gagnon et al. use is effective in measuring the impact of a change in environment, there is chosen for

this type of empirical research (MacKinlay, 1997). Moreover the idea of analysing the ECB’s APP policy came up because of its announcements, so investigating if these announcements itself already have any influence comes quite natural. In order to do so the implementation of the ECB’s QE policy and corresponding dates with important information regarding the APP has to be clarified. But first the theory of an study will be discussed The Constant Mean Return Model The models used in this event study are the Constant Mean Return Model (CMRM) and the Market Model (MM) according to MacKinlay (1997). With the CMRM: 𝑅_!" = 𝜇_! + 𝜁_!" 2 𝑤𝑖𝑡ℎ 𝐸 𝜁!" = 0 & 𝑣𝑎𝑟 𝜁!" = 𝜎!! ! where the return on an asset 𝑅_! in the period-t is estimated as the mean of the returns over that time period plus an additional error term (𝜁!") with an expected value of zero and a variance 𝜎_!!_{!. The sensitivity of this model may be a little less compared to more} complex models. This is due to the fact that the simplicity of the model increases the variance of the abnormal returns. Nevertheless MacKinlay states that it is still very usable since it usually generates the same results as the more complex models, just a little less precise. To see if an event itself has had an effect and generates abnormal returns it is therefore good in use. The abnormal return is the actual return of an asset at time t minus the expected return of that asset at time t. The CMRM and also the market model are instruments that are used to give an expectation of the returns on assets (for when the event takes place), using the actual returns in the estimation window. This is the reason why the days of the event (event window) should not be taken into the estimation window, because the event should provide abnormal returns. By not taken this into account the estimators are not influenced by the event and the ‘normal’ returns are projected so that the event impact can be captured by the abnormal returns. For the CMRM the abnormal return is 𝐴𝑅_!" = 𝑅_!" − 𝜇_! (3) 𝑤𝑖𝑡ℎ 𝜎! _𝐴𝑅 !" = 𝜎!! ! ₊ 1 𝐿_! 1 + 𝜇_!! 𝜎_!!_!

Since the influence of an event or even the event itself can last for multiple days, the aggregation of the AR (CAR) is just as important. This holds that the CAR measures across securities and along time. The CAR is thus the sum of the AR during the event window with a variance: 𝜎_!! 𝜏!, 𝜏! = (𝜏!− 𝜏!+ 1)𝜎! 𝐴𝑅!" . As shown, the variance of the AR (and thus of the CAR) contains of two parts: the estimated variance of the CMRM and an additional term that allows for cross correlation of the abnormal returns. For most dates it is not necessary to add this because the estimation window (L1) is very large (250 data points) what will make the additional term approach zero. Nevertheless, it should be added to the variances of the used CAR of 02 and 15 January 2015 since those dates have clustering in their (10-day) event window. This means that there is an overlap of event windows and the results are biased. By using the variance of equation 3 and substitute it in the variance of the CAR for the two mentioned dates and use the length of the event window (10) for the value of L1, the problem of serial correlation is solved according to MacKinlay, but more on that in the methodology. The Market Model In this study also the Market Model will be used. To evaluate if the QE policy of the ECB has had any effect it is also interesting to see if any fluctuations can be measured in the yields of the government bonds of a euro country for the separated events. Therefore the Market Model can be applied, stated as: 𝑅_!" = 𝛼_! + 𝛽_!𝑅_!" + 𝜀_!" 4 𝑤𝑖𝑡ℎ 𝐸 𝜀_!" = 0 & 𝑣𝑎𝑟 𝜀_!" = 𝜎_!!_! here, the return on a security is depended on an estimator alpha, an estimator Beta times the return on the market portfolio and on a distortion estimator with an expected value of zero. For the Market Model the abnormal return has the same definition as for the CMRM and also the interpretation of the variance is the same, except that the AR now looks like this: 𝐴𝑅!" = 𝑅!"− 𝛼! + 𝛽!𝑅!" (5) 𝑤𝑖𝑡ℎ 𝑤𝑖𝑡ℎ 𝜎! _𝐴𝑅 !" = 𝜎!!!+ 1 𝐿_! 1 + (𝑅_!"− 𝜇_!)! 𝜎_!!

Again here the second term is only used for serial correlation and thus will approach zero if there is no clustering of event windows, since L1 is large then. The CAR is stated in the same way as is done for the CMRM, as is its variance, but than applied with the AR of equation 5.

Methodology

MacKinlay (1997) states that, given rationality in the market, the effects of an event will immediately be reflected in security prices, so that and event study becomes very useful. In this statement there is already one assumption that will also be assumed in this paper: markets are rational and efficient. This implies that market participants immediately update their expectations on interest rates when new information regarding policy changes is available. The market will thus respond to the day of an announcement regarding policy and not specifically to when the actual policy is implemented. In obtaining any valid results the next assumption will be important as well: the set of events used in this paper will be complete and will exist of all the necessary announcements regarding the ECB’s APP. This will be done by adequately analysing the ECB’s official announcements during a period of 2014 until 2017 and by using the paper of Altavilla, Carboni & Motto (2015). Moreover the methodological approach assumes that the events used are the only events that have affected the APP expectations. Lastly the used windows will be assumed wide enough to contain all the effects regarding the APP but not that wide that yields will be affected by other influences. Events regarding APP announcements or speculations of the ECB 1) 18 September 2014; The ECB announced a new programme that consisted of eight parts of longer-term refinancing operations (TLTROs) (European Central Bank, 2014a). The programme was initiated to support lending to the real economy, which was a sign to the market that a QE policy announcement of the ECB could be expected, somewhere in the near future. This is support by the Financial Times and several other newspapers that started to speculate about this the next day (Financial Times, 2014). 2) 2 October 2014; A press release in which the ECB announced that it would implement a more than two year lasting policy of buying bonds, starting in the in

the fourth quarter of 2014 (European Central Bank, 2014b) (European Central Bank, 2014). 3) 02 January 2015; The president of the ECB raised expectations about a coming up QE programme to fight deflation. In an interview with the German financial daily Handelsblatt, Draghi said that realisation of the obtained 2% target level of inflation was drifting away more and more, with could be a signal for the markets that a QE policy could be announced on the banks next meeting of 22 January (The Guardian, ECB Boss Hints at Stimulus to Fight Deflation, 2015). 4) 15 January 2015; In an interview with ‘Die Zeit’ Mario Draghi stated that an expansionary monetary policy was the necessary thing to do to be able to counter deflation and recession at that time (Draghi, 2015) 5) 22 January 2015; The ECB announced at a press conference its expanded asset purchase programme in more detail. It would start in March 2015 and would contain the buying of a variety of securities, up to a monthly value of 60 billion euro. Moreover in the same press conference the ECB announced that there would be a modification to the interest rate. This in support of the future targeted longer-term refinancing operations (European Central Bank, 2015).

6) 10 March 2016; The ECB announced that it would push up their monthly purchasing rate up to 80 billion (European Central Bank, 2016).

7) 26 October 2017; In their monetary policy decisions the ECB explained that it would continue with its QE policy until January 2018. After that, the APP will still be in progress until at least September 2018, but with half the amount of monthly purchases as the intended amount of 60 billion (European Central Bank, 2017). Implementation The impact of the above listed dates is analysed with AAA rated euro area central government bonds on their zero-coupon spot rate. The bonds that were used varied in maturity: a 1-year maturity short-term bond and 2-, 5-, 10-, 20- and 30-year long-term bonds. For the Market Model also the Italian 10 year government bond was used, but that will be explained later.

All data was obtained from Eurostat. The estimation window was set on 250 days following MacKinlay (1997). The Constant Mean Return Model was used on the returns of the yields of those assets*_. The euro area central government bonds were used because the impact of policy measures by the ECB should be of a direct influence on those bonds. Regarding the use of them in the Market Model, they represent the AAA rated EU government bonds and are used as the market portfolio. Furthermore the Italian government bonds are used because they have had a quite stable BBB rating for the period of 2014-2017 according to the credit rating agencies (Trading Economics, 2017). This means that they should be more volatile to policy changes than the market portfolio. Where the estimation window was set on 250 days for every single event, the event window has been set on one day (the day of the announcement itself) for the CMRM and the Market Model. Moreover for the CMRM there a ten-day event window is used and a five days window for the Market model, including the day of the event of course. This is done because with this method the immediate market response can be seen, but with the extended window also the long run effects of the event can be measured. In doing so the abnormal return is used and tested on the one-day event window, whereas the cumulated abnormal return over ten days is used and tested in the long run analyses. As stated in the introduction and explained in the literature review, the yield on especially the longer-term assets should decrease because of a lower premium risk caused by the QE policy of the ECB. In the CMRM this should reflect a negative return on the day of events, so a negative abnormal return. Considering the long run effects this should reflect a negative CAR for the long-term bonds. Since of course the variance, and thus the standard deviation, can’t be negative, the results of the student-t-tests for both cases should be negative as well. Still, the exact reaction of the market may not follow theory. Market expectations could be very different and thus also an increase in yield could be possible. Because of this, the performed t-tests are two-sided with and 𝐻_! stated as: the change in the interest rates of bonds due to policy announcements is zero; with 𝐻_!: there is a change in the interest rates of bonds due to policy announcements. * _{This caused a few data errors for the bonds with maturity of 5 years or less, since the yields on those} assets went negative at some point in the estimation window and sometimes remained at a zero interest level. So for some of those bonds a couple of days had to be removed what resulted in a sometimes slightly smaller estimation window.

This means that, with 𝛼 = 0,05, significant differences are represented when the t-test score is smaller than ± 1,96. The same applies for the Market Model and the changes of the 10-year Italian government bonds. Empirical analysis The results of the constant mean return model with a one-day event window show some clear results, represented in table 1 (appendix). The results show that for all the events in January 2014 there has been a significant result in the yields of the longer-term bonds, the ones of 5-, 10-, 20- and 30- year maturity (except for the 5 year maturity on 22-Jan-15 and the 30 year maturity on 15-Jan-15). This result shows that the market immediately picked up the signals about an upcoming QE policy of the ECB and for the 2nd _{and 15}th _{of January anticipated on it with a lower short-term interest rate} expectation for the future or with a reduction in the risk premium on long-term assets. The latter is assumed to have the biggest effect of those two, because the table shows that the short-term interest rate (1-year) was not or hardly shifted on those dates. Furthermore, earlier policy announcements of the ECB had already informed the market that the short-term interest rate would remain low and stable for the upcoming future. So the market expectation about the future path of interest rates shouldn’t have changed much on the event dates, what suggests that the decrease in yields was due to a decrease in the risk-premium on those bonds. * Another result that can be found in the same table (1) is what happens on the first event. Although the short-term interest rate remained quite stable, the yields on 5- and 10 year maturity bonds increase significantly. So here the opposite of the expectation has happened. In the ECB announcements nothing is mentioned regarding the future path of the short-term interest rate. This combined with the small, but not significant, decrease in the short-term yield, the significant increase can be assumed as * _{A bit extreme are the significance values on the fourth event considered the 5-year maturity bond. These} are due the fact that the estimation window approached to a zero interest rate and on the actual date (15-Jan-2015) the yield jumped from 0,05 to 0,00. A huge (negative) return is realised what continues in the following days when the yield of the next day could easily be doubled what, combined with a low standard deviation measured in the estimation window, results in unusual high returns. For the next event this effect of the turning point around zero has been countered by a much higher standard error. The latter also applies for the short-term and 2-year maturity bond.

an increase in the risk premium. The announced purchases of TLTROs may have been received by the market as a sign of weakness, what explain the significant increases in the risk premium of the bonds with 5- and 10-year duration. Hence, exactly the same happened a few days later, on the 2nd _{of October, with the} announcement of two other assets purchasing programmes. Apparently the first reaction of the market on these announcements was not as the theory suggests. The same argument holds, the increase in yields of the 5- and 10-year bonds is most likely due to an increase in the risk premium due to little trust in effectiveness of the ECB’s announced policies. The last thing regarding the one-day event window is that no significant results are obtained in the announcements made during the APP (except for the 5 year maturity bond on the 26th _{of November 2017). This suggests that the up scaling of the monthly} purchase rate from 60 bln euro per month up to 80 bln euro per month didn’t came as a surprise for the market. What can be observed is that the yield of the long-term bonds now follows the behaviour of the short-term interest-rate and decrease simultaneously, although not significantly. This can also be seen for those last two events in the results of the ten-day event window. The announcements made during the APP didn’t create an effect on the market, what means that the expectations of market participants were in line with the with the policy measures. Except for the 5-year maturity bond, this one shows a significant increase in its yield on the day of the announcement regarding the extension and future downsizing of the policy. For the bonds with the highest maturities (20- and 30- years) a significant decrease in yields is shown in the results for every announcement before the APP implementation (except for the 30 year maturity on the 15th _{of January 2015). Where} the short-term interest rates do not show any significant decrease, again it can be assumed that the risk premium on these securities decreases. Furthermore what’s interesting to see is that where in the one-day event window there was observed an increase in the yields of 5- and 10-year maturity bonds on the first two event dates, in the 10-day event window there is a significant decrease for the first event. So over the days in this event window the expectations of market participants have probably changed. Since this is regarding the first event date, and an actual QE policy wasn’t yet implemented, the expectations about the ECB’s policy could have varied a lot. So the

increase of yield at the first moment can be well explained by the earlier mentioned insecurity about ECB policy. The same applies for the fifth event. Although the yields for the longer-term bonds significantly increase on the announcement day itself, in the following days still an overall decrease is measured. Thus, in the end the yields of those bonds have a by theory expected decrease just like the other events in January 2015. Strange is the fact that on the other January events the markets respond wasn’t the same as on the APP announcement date. This shows that maybe market participants were still a bit shocked on the 22nd _{(maybe because of the duration and/or amount of purchases) but eventually} absorbed the policy quite well. Several financial media institutions like The Guardian (ECB 'Takes Out Bazooka' with bigger than expected QE stimulus package, 2015) discuss that the amounts of monthly purchases were beyond expectations. This could cause a shock in the yields of long-term bonds but with knowledge from the literature review this shock would be expected to be negative instead of positive. Since the ECB would purchase a bigger amount of securities the risk of default on the bonds left in the market should decrease. Theory thus suggests a lower risk-premium on those bonds and with still a stable short-term interest rate expectation the yields should drop. Nevertheless, the results show an increase but it the explanation for that remains ambiguous. The other events in January show the same results for the 10-day event window as for the one-day event window: significant decreases in yields for the 5-, 10-, 20- and 30 year bonds (except for the 10 year maturity bond on 15-Jan). Results of The Market Model For the market model the results are not as clear as those of the CMRM. For starters it can be seen that for all events the goodness of fit of the model is low (R-square less than 0,3). This means that the yield of the Italian 10-year maturity government bond is not very well explained by the market portfolio. Since 10- year government bonds of a Italy is the only one with available daily data, the effects of yields and the quality of the model cannot be analysed by other regressions of Italian bonds. Therefore it can be criticized that the obtained results, discussed in the following paragraphs, are very debatable and do not represent more than an indication. Tables one and two show that for the one-day event window only the 3th _{and the} 5th _{event have a significant result. Both represent a negative abnormal return on the}

yield of the Italian 10-year maturity government bond. Especially the 5th event (22-January-2015) is interesting since it is known from the CMRM that on that day the interest rate on the 10-year EU bonds showed an increase. This doesn’t mean that the Italian government bond didn’t increase on that day as well, but that at least it increased significantly less than was expected from the market model. Regarding the five-day event window the only significant results can be obtained from the period during the APP, so the sixth and seventh event (Table 4). Where the market portfolio didn’t changed at all on this date, as we know from the CMRM, the Italian 10-year to maturity bond shows a negative abnormal return. This is based on the assumption that without a significant result from the one- and 10-day event window, it can be assumed that there is not a change for the market portfolio in for 5 days as well. With still stable short-term interest rate policy by the ECB and an unchanged market portfolio, this means that there must be a decrease in the risk premium of the Italian government bond. However, as mentioned earlier, the goodness of fit of the model remains debatable as will be further discussed in the discussion, what means that the results from the data can maximally be used as an indication of a phenomenon and not as an exact instrument for measuring these decreases.

Conclusion and Discussion

With short-term interest rates at the zero lower bound and a threat of deflation it is known from previous examples that unconventional monetary policy can be used as an instrument to counter this deflation risk. As is known, the QE programme by the Bank of Japan showed that an important and very influential part of a QE policy is signalling to or communication with the market. By doing an event study the influences of these signals or communication tools on the yield of euro area bonds are analysed with the expectation that these would decrease because the APP would lower the expectations about the future path of short-term risk and furthermore decrease the risk premium of long-term bonds. From the results it can be concluded that the announcements about the implementation of the APP did lower yields of long-term bonds EU government bonds. As obtained from the event study on EU bonds with the constant mean return model, the 10-day event window has shown that the yield of the most mature bonds (20- and 30-years to maturity) have decreased on every event before implementation (except for

one event regarding the 30-year bond). That these decreases are due to lower risk-premiums is very plausible since the ECB has communicated from the beginning that it would hold short-term rates low and stable for a long time. Since in this paper the ECB definition of a long-term bond holds, it cannot be concluded that the yield of every long-term bond decreased before implementation. Another conclusion that can be drawn is that the significant announcements regarding the APP are the third and fifth. On these days the announcements made caused a change in the yield on long-term bonds (excluding the 2-year to maturity bond). Overall it can be said that these announcements have caused for a decrease in the yields and most of all due to a decrease in the risk-premium. Nevertheless the announcement with al the specific information regarding the APP still shocked the market and increased yields on long-term bonds on the day itself. Financial media reports of that day point out that the APP was bigger than expected but that would suggest a decrease in yields. The real reason for the increase in yields on that event could be analysed in future research. Furthermore it this study shows that the announcements made during the APP were of little influence on the interest rate of long-term bonds. This means that the upcoming policy changes that were officially announced on these dates were well expected by the market. Although the market model analysis counters this conclusion with significant changes on yields of the Italian government bond, the low value for the goodness of fit of this model doesn’t represent enough evidence. Due to its limitation in time this study has not been able to figure out what other explanatory variables for the Italian government bonds should be present in the market model regression. Further research on this topic would be helpful in two ways: firstly in detecting all the explanatory variables that influence Italian government bonds. Secondly the impact of the ECB’s QE policy on a euro area country can be analysed.

References

Altavilla, C., Carboni, G., & Motto, R. (2015). Asset purchase programmes and financial markets: lessons from the euro area. Frankfurt am Main: European Central Bank. Auerbach, A. J., & Obstfeld, M. (2005). The Case for Open-Market Purchases in a Liquidity Trap. The American Economic Review , 95 (1), 110-137. Bernanke, B. S., & Reinhart, V. R. (2004, May 03). Conducting Monetary Policy at Very Low Short-Term Interest Rates. The American Economic Review , 85-90. Bernanke, B., Reinhart, V., & Sack, B. (2004). Monetary Policy Alternatives at the Zero Lower Bound: An Empirical Assesment. Brookings Papers on Economic Activity , 2, 1-100. Draghi, M. (2015, January 15). Interview with Mario Draghi, President of the ECB. (G. d. Lorenzo, Interviewer) European Central Bank. (2017, Oct 26). Asset Purchase Programmes. Opgeroepen op Dec 19, 2017, van ecb.europa.eu: https://www.ecb.europa.eu/mopo/implement/omt/html/index.en.html European Central Bank. (2014, Sep 18). ECB allots €82.6 billion in first targeted longer-term refinancing operation. Opgeroepen op Jan 13, 2018, van ecb.europa.eu: https://www.ecb.europa.eu/press/pr/date/2014/html/pr140918_1.en.html European Central Bank. (2015, Jan 22). ECB announces expanded asset purchasing programme. Opgeroepen op Dec 19, 2017, van ecb.europa.eu: https://www.ecb.europa.eu/press/pr/date/2015/html/pr150122_1.en.html European Central Bank. (2016, Mar 10). ECB announces new series of targeted longer-term refinancing operations. Opgeroepen op Jan 13, 2018, van ecb.europa.eu: https://www.ecb.europa.eu/press/pr/date/2016/html/pr160310_1.en.html European Central Bank. (2014, Oct 02). ECB announces operational details of asset-backed securities and covered bond purchase programmes. Opgeroepen op Dec 19, 2017, van ecb.europa.eu: https://www.ecb.europa.eu/press/pr/date/2014/html/pr141002_1.en.html European Central Bank. (2018). Key ECB interest rates. Opgeroepen op Jan 02, 2018, van ecb.europa.eu: https://www.ecb.europa.eu/stats/policy_and_exchange_rates/key_ecb_interest_rates/h tml/index.en.html European Central Bank. (2018, Jan). Securities issues. Opgeroepen op Jan 07, 2018, van ecb.europa.eu: https://www.ecb.europa.eu/stats/financial_markets_and_interest_rates/securities_issu es/html/index.en.html

Financial Times. (2014, Sep 19). Weak ECB Loans Demand Paves Way for 'QE'. Opgeroepen op Jan 11, 2018, van ft.com: https://www.ft.com/content/89b404cc- 3fee-11e4-a381-00144feabdc0 Gagnon, J., Raskin, M., Remache, J., & Sack, B. (2010). Large-Scale Asset Purchases by the Federal Reserve: Did They Work? Federal Reserve Bank of New York Staff Reports (441). MacKinlay, C. A. (1997). Event Studies in Economics and Finance. Journal of Economic Literature (35), 13-39. Mankiw, G. N. (2012). Macroeconomics. New York: Worth Publishers. Mishkin, F. S. (2007). Monetary Policy Strategy. London: The MIT Press. The Guardian: a) (2015, Jan 02). ECB boss Hints at Stimulus to Fight Deflation. Opgeroepen op Jan 13, 2018, van theguardian.com: https://www.theguardian.com/business/2015/jan/02/european-central-bank stimulus-deflation-mario-draghi The Guardian. (2015, January 22). ECB 'Takes Out Bazooka' with bigger than expected QE stimulus package. Opgeroepen op January 25, 2018, van theguardian.com: https://www.theguardian.com/business/2015/jan/22/ecb-big-bazooka-bigger- than-expected-qe-stimulus Tobin, J. (1969, Feb). A General Equilibrium Approach to Monetary Theory. Journal of Money, Credit and Banking , 15-29. Trading Economics. (2017, Oct 27). Italy - Credit Rating. Opgeroepen op Jan 17, 2018, van tradingeconomics.com: https://tradingeconomics.com/italy/rating Ugai, H. (2006). Effects of the Quantitative Easing Policy: A Survey of Empirical Analyses. Monetary Affairs Department. Tokyo: Bank of Japan.

Tables

Table 1: changes in yield on the event days based on the CMRM. Data source: Eurostat daily data on euro yield curves Table 2: source: Descriptive statistics of the CMRM with a 1 day event window. Source: Eurostat daily data on euro yield curves

1-year 2-year 5-year 10-year 20-year 30-year

Date Event

before APP

18-09-14 First QE speculation -0,02 -0,02 ***0,04 *0,04 0,02 0,01

02-10-14 Press release -0,01 -0,01 **0,02 *0,03 0,02 0,01

02-01-15 Interview Handelsblatt 0 0 **-0,01 ***-0,04 ***-0,06 ***-0,06

15-01-15 Interview Die Zeit -0,02 -0,03 ***-0,05 ***-0,05 **-0,05 ***-0,07

22-01-15 Details of expanded APP -0,01 -0,04 -0,01 ***0,06 ***0,12 ***0,13

during APP

10-03-16 Higher purchasing rate -0,01 -0,01 -0,02 -0,02 -0,04 -0,05

26-10-17 Continuation of APP -0,02 *-0,03 **-0,04 -0,04 -0,03 -0,03

Maturity of bonds

Event window of one day, so only the day of the announcements itself ***;**;* denote significance at the 1, 5 and 10 percent levels

Statistics of the Constant Mean Return Model, 1 day event window

Date EU gov. Bond 1-year maturity 2-year maturity 5-year maturity 10-year maturity 20-year maturity 30-year maturity 18-09-14 Mean 0,0325 -0,0192 -0,0044 -0,0024 -0,0014 -0,0012 St. Dev 0,3069 0,3305 0,0432 0,0186 0,0116 0,0115 AR -0,2325 -0,2030 0,1423 0,0363 0,0111 0,0061 t-test score -0,76 -0,61 ***3,29 *1,95 0,96 0,53 02-10-14 Mean 0,0342 -0,0168 -0,0049 -0,0027 -0,0016 -0,0014 St. Dev 0,3058 0,3314 0,0458 0,0190 0,0117 0,0114 AR -0,1342 -0,0741 0,1001 0,0324 0,0124 0,0066 t-test score -0,44 -0,22 **2,19 *1,70 1,06 0,58 02-01-15 Mean 0,0272 -0,0164 -0,0089 -0,0047 -0,0029 -0,0026 St. Dev 0,2928 0,3336 0,0618 0,0230 0,0150 0,0144 AR -0,0272 0,0164 -0,1340 -0,0569 -0,0409 -0,0380 t-test score -0,09 0,05 **-2,16 ***-2,47 ***-2,72 ***-2,64 15-01-15 Mean 0,0296 -0,0157 -0,0094 -0,0050 -0,0035 -0,0030 St. Dev 0,2909 0,3343 0,0677 0,0254 0,0177 0,0160 AR 0,1704 0,2465 -0,9906 -0,0859 -0,0403 -0,0501 t-test score 0,59 0,74 ***-14,63 ***-3,38 **-2,28 ***-3,12 22-01-15 Mean 0,0308 -0,0155 -0,0052 -0,0049 -0,0036 -0,0028 St. Dev 0,2947 0,3344 0,2983 0,0270 0,0180 0,0169 AR 0,0248 0,3792 -0,1948 0,1140 0,1127 0,1013 t-test score 0,08 1,13 -0,65 ***4,23 ***6,27 ***6,01 10-03-16 Mean 0,0041 0,0051 -0,0124 0,0050 0,0031 0,0029 St. Dev 0,0441 0,0548 0,8606 0,0873 0,0511 0,0474 AR -0,0241 -0,0243 0,0791 -0,0638 -0,0431 -0,0433 t-test score -0,55 -0,44 0,09 -0,73 -0,84 -0,91 26-10-17 Mean 0,0006 0,0006 0,0010 0,0100 0,0030 0,0026 St. Dev 0,0223 0,0255 0,0843 0,1009 0,0393 0,0347 AR 0,0257 0,0423 0,1729 -0,0840 -0,0280 -0,0234 t-test score 1,15 *1,66 **2,05 -0,83 -0,71 -0,68 ***;**;* denote significance at the 1, 5 and 10 percent levels

Table 3: Cumulative Abnormal Returns based on the CMRM with an event window of 10 days. Data source: Eurostat daily data on euro yield curves

CAR t-test score CAR t-test score CAR t-test score

Date Event

before APP

18-09-14 First QE speculation -3,079 -0,32 0,74468 0.71 0,2439 -1,78 02-10-14 Press release -0,4489 -0,46 0,2763 0,26 -0,0568 -0,39 02-01-15 Interview Handelsblatt 0,012 0,01 0,698 0,66 -1,1397 ***-5,77 15-01-15 Interview Die Zeit 0,4071 0,44 0,5241 0,50 -2,8558 ***-13,22 22-01-15 Details of expanded APP -0,2213 -0,24 0,8823 -0,83 -2,5932 ***-2,75 during APP

10-03-16 Higher purchasing rate -0,0763 -0,55 -0,1248 -0,72 0,0869 0,031 26-10-17 Continuation of APP 0,0345 0,49 0,9223 1,15 0,4927 *1,84

CAR t-test score CAR t-test score CAR t-test score

before APP

18-09-14 First QE speculation 0,1281 **-2,18 -0,0874 ***-2,39 -0,072 **-1,98 02-10-14 Press release -0,0941 -1,57 -0,081 ***-2,40 -0,0866 **-2,28 02-01-15 Interview Handelsblatt -0,2155 ***-2,80 -0,1952 ***-3,80 -0,1428 ***-2,93 15-01-15 Interview Die Zeit -0,1168 -1,39 -0,1196 **-1,99 -0,0852 -1,57 22-01-15 Details of expanded APP -0,162 *-1,9 -0,1558 ***-2,74 -0,1768 ***-3,32 during APP

10-03-16 Higher purchasing rate -0,0384 -0,14 -0,1053 -0,65 -0,1066 -0,71 26-10-17 Continuation of APP -0,4372 -1,37 -0,171 -1,37 -0,142 -1,29 ***;**;* denote significance at the 1, 5 and 10 percent levels

Cumulative Abnormal Returns for an event Window of 10 days

1 year maturity 2 year maturity 5 year maturity

Table 4: Descriptive statistics of the market model. Data source: Eurostat daily data on euro yield curves and on euro area countries

Event Date 18-sep-14 02-okt-14 02-jan-15 15-jan-15 22-jan-15 10-mrt-16 26-10-17 Statistics

Intersect -0,0018 -0,0017 -0,0022 -0,0023 -0,0023 0,0012 0,0002

Slope 0,2623 0,2618 0,1219 0,1294 0,1454 0,1084 0,0881

R-Square 0,1251 0,1315 0,0277 0,0344 0,0475 0,0789 0,1527

Stand. Dev. 0,0129 0,0128 0,0167 0,0175 0,0176 0,0324 0,0210

AR on Italian bond (1day) -0,0071 0,0067 -0,0528 -0,0140 -0,0606 -0,0424 -0,0329 CAR on italian bond (5 days) 0,0182 0,0354 0,0010 -0,0392 -0,0209 -0,1731 -0,0858 T-test scores

1-day event window -0,55 0,53 ***-3,17 -0,80 ***-3,45 -1,31 -1,57

5-day event window 0,63 1,24 0,03 -1,00 -0,53 ***-2,39 *-1,83

Statistics of the Market Model performed on the return of Italian 10-year maturity government bonds (1-day event window)