The effect of Quantitative Easing on the relation between the long-term interest rates and inflation in the euro area

The Effect of Quantitative Easing on the

relation between the Long-term

interest rates and inflation in the Euro

area

Abstract

This thesis analyses the effect of the Quantitative Easing program implemented by the ECB on the euro area. An event study is used to determine the significance of multiple

channels through which QE works. Furthermore, multiple models are used to determine the relation between the long-term interest rates and multiple macro-economic variables capturing information about economic growth and inflation. With these models long-term interest rates are predicted before and after the implementation of the program, to examine if the found relation is affected by QE. Findings suggest QE has significant effects through multiple channels. Furthermore, using multiple models contradicting results are found on a change in the relation between the long-term interest rates economic growth and inflation.

University of Amsterdam Bachelor Thesis

Student: Kjelle Boot 10791868

2

Statement of Originality

This document is written by Kjelle Boot 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 contents

Chapter Page

1. Introduction 4

2. Literature review 7

3. Data & Methodology 12

4. Results 17

5. Conclusion 20

6. Discussion 21

7. References 22

4

Introduction

To this day the European Union is still recovering from the downside effects of the financial crisis started in 2007. One of these effects is the drop in the inflation rate across the euro area. A primary goal of the ECB is to maintain the inflation level around 2% (ECB, 2017). Since the beginning of the crisis unconventional monetary policy tools like

Quantitative Easing (QE) were becoming popular. Central banks use these policy tools to lower interest rates and stimulate the economy to reach a 2% inflation level. central banks of countries like Japan (BoJ), England (BoE) and the United States (Fed) started

implementing QE and options for banks to loan at discount at the beginning of the crisis 2007-2009 (Martin & Zhang 2017). The ECB used other policy tools to achieve stimulation of the economy and the 2% inflation level.

These tools proved ineffective as the inflation levels kept going down to the point where there was even fear for deflation. As the current tools where not effective, other options needed to be evaluated. In 2011 the newly assigned president Draghi said to do ‘whatever it takes’ to reach the goal of 2% inflation level in the couple years to come (Dunkly, 2012). Even though positive effects of QE implementation were already visible in the U.S, England and Japan, did the ECB not start its first QE program until 2015 (Claeys, 2015).

The euro area faced very low inflation levels, almost even deflation levels. By the possibility of upcoming deflation, the spending by investors and consumers went down rapidly as spending in the future could be less expensive (Lautenschläger, S. 2014, July 7). In addition, investors held on to their money as credit ratings of companies went down and investing became more risky. Because of this downward spending spiral the trust in the economy began to fade, which led to economic stagnation and even decline (Lautenschläger, S. 2014, July 7). The first implementation of QE in the euro area was set to reverse the negative spiral effects, and sustain a more stable and growing economy.

At the moment the inflation rate across the euro area is still not near the ECB goal, at average 1,1% (Eurostat, 2017). There are indications that inflation is rising, and the fear for deflation is dwindling, but a big increase in inflation level that will reach for 2% will not be realized any time soon (FD, 2017). Furthermore, the ECB is calling for raising loans to stir up

the inflation levels. This indicates that the QE program did not have the effects the ECB 55hoped for, as other measures are being used. It is still unclear when inflation levels will rise again, making investors more prudent towards the market. As a result investors tend to hold on to their money, which holds back economic recovery (FD, 2017).

There are reasons to believe the implementation of the QE programs have been ineffective to comprehend to economic stimulation and higher inflation levels. This thesis therefore asks whether this ineffectiveness could be the result of a different relation

between the interest rates and inflation then the ECB expected. Hence, a different effect on the economy in the euro area. The analysis on the effects of QE will be done with the use of an event study and multiple simple regression models. The methodology of an event study will help to examine the effects of QE on the long-term interest rates. The multiple

regression models will be used to determine the relation between the long-term interest rates and multiple macro-economic variables which capture economic growth and inflation, through which the effects of QE should be visible. The research question of this thesis is:

"What are the effects of the implementation of Quantitative Easing Programs on the relation between long-term interest rates and inflation within the Euro area?’’

With an answer on the research question it could be stated if QE did have a significant effect on changes in long-term interest rates, and if QE contributed with increasing the inflation level towards the 2% goal of the ECB.

There has been much research on the effects of QE. However, because of the late implementation of QE by the ECB there is rather less known about its effects in the Euro area. The economical relevance of this question is an answer on the relation between long- term interest rates and inflation before and after the implementation of the QE program. This could give an insight on the effects of QE, and possibly how future implementations could be more effective.

The paper is structured as follows. Firstly, this paper will explain the theory behind QE and discuss previous research and contradicting results on the subject. Secondly, an event study will be conducted to determine the effects on long-term interest rates by the first QE

program implemented by the ECB. Thirdly will simple regression models be used to examine the relation between long-term interest rates, economic growth and inflation. Fourthly,

there will be a discussion about the interpretation of the results, and finally an answer on the research question with accompanying conclusion where is stated what the answer means, and what this could say about the QE program.

Literature review

The literature review is divided in three subsections. Firstly, the theory will be discussed including explanation about multiple channels through which the effects of QE are realised. Secondly, research on the subject and contradicting results will be addressed. Thirdly will be stated how this thesis will examine the effects of QE and how its results are of economic relevance on the subject.

Quantitative easing is an unconventional monetary policy tool. It works through large scale purchases of securities to increase the monetary base and thus the economy (Claeys, 2015). The theory behind QE is shown in the figure below, and explained as follows. By purchasing securities, there is an increase in price and this creates money in the banking system. With the obtained money banks will buy other assets to replace the ones sold to the central bank. This leads to price increases of the assets and make the interest rates drop. Hence, a wide range of interest rates fall and loans become cheaper. The households and businesses are able to borrow more, as they have to pay less in interest. The result is a higher level of consumption and investment. These higher levels of investment and

consumption support job creation and economic growth.

The final stage of this theory is that as the prices rises an increase in inflation will follow. In case of the euro area the ECB will use this policy to reach an inflation level around the 2% (ECB, 2017).

The main reason to implement QE programs is to bring the inflation and also inflation expectations to a level that is in line with the central banks inflation goal, the stimulation of economic growth and lower the unemployment rate. The way this works is through lowering interest rates and devaluating the currency. The effects of QE are realized through multiple transmission channels. The most prominent ones are the portfolio rebalancing and the signalling channel (Gambacorta et al, 2014). These channels are mainly targeting the goal of lowering long-term interest rates. The signalling channel works through affecting

expectations of future long-term bond yields, while the portfolio rebalancing channel affects term premiums, which influences the interest rates. Furthermore, there are two minor channels: the exchange rate and the wealth channel.

The primary channel through which QE works is the portfolio rebalancing channel (Gambacorta et al, 2014). It affects the risk premium on the long-term assets bought by the central banks. The risk premiums on long-term assets are the extra amount of money investors are demanding for the risk of holding on to a long-term asset instead of a short- term asset. The way it affects the yields on both long-term as short-term assets is through imperfection of substitutes between short-term and long-term interest rates (Gagnon et al, 2010). The yield curve is affected by the relative supply of short and long-term bonds. Large scale purchases of long term bonds by the central bank lowers the term premium, because as demand increases the compensation for the assets decreases. Through arbitrage

opportunities the premiums of similar assets are affected the same way (Gagnon et al, 2010). A drop in the term premiums stir investors to start investing in higher yield

investment opportunities, dropping those expected returns as well. In short, the purchases drive up the prices of the assets and lower its yield. The alternative method is directly

buying private sector assets. These are substitutes of money and government bonds, as they tend to be more risky investments. The increase in demand of these assets directly drops the market risk premiums of these assets as well (ECB, 2009).

Before implementation central banks use forward guidance communication

yields (Campbell et al, 2012). This works quite the same for the signalling channel within QE programs. By informing the market there will be a large asset purchase program in the near future, it could already have a beneficial effect before implementation. It could lead to a decrease in long-term bond yields if the market believes in the central bank’s commitment to keep interest rates low for an extended period of time, even after the economy recovers (Krishnamurthy, Vissing-Jorgensen, 2011). The effects of implementing a QE program itself give the market this credibility. The central bank purchases assets, and if central banks raise interest rates in the near future, losses will be made on those bought assets. As assets are bought on large scale, so will be the losses on these assets, this gives the market reason to believe the Central Bank will not raise interest rates in the near future. Purchasing assets on large scale thus serves as a credible commitment to keep interest rates low. Another

possibility is that the large-scale asset purchases by the central bank signals to the market how bad the economy really is, so an extra disadvantage of an early exit would be that the negative signal effect outweighs the positive one who needs more time to complement. So, these two effects signal that the QE program will be in place for some time.

By the huge increase of the money supply and a successful decrease of interest rates QE programs could lead to depreciation of the exchange rate. This causes a stimulation of the economy through a third channel, the exchange rate channel. As the domestic currency becomes cheaper compared to foreign currency, export becomes more attractive. The increase in export has a positive effect on the economic growth. As economic growth causes domestic price increases, further increases in the inflation level will follow.

A forth channel is the so-called wealth channel. As interest rates drop through the effects of other channels, investors will increase spending. They buy more stock and physical assets to contain the net present value that is not solvable with the former long-term bonds yields. The extra amount of money spends economy a boost, as demand rises prices will rise to. This leads to an increase of the inflation level, wanted by the central bank

(Krishnamurthy, Vissing-Jorgensen, 2011). When these four channels are proved significantly strong enough to lower long-term interest rates, and boost the economic growth, QE may have a positive effect on the increase in prices levels and hence an increase in the inflation level. The time that goes by before this effect is realised depends on the anticipated effects of the QE program and in which extend it is implemented.

So far there has been extensive scientific research on the effects of quantitative easing.

The results of these are somewhat contradicting. According to Joyce, Miles, Scott and Vayanos (2012) the fact that the economy of the euro zone remains weak gives a reason to believe that QE does not work. On the other hand, the results in this research state that QE does have a positive effect on the economy, so the question arises if QE really is ineffective, or are their distortionary factors that undermine the potential effects of QE. Another explanation could be that the scale on which it was implemented was not high enough to reach the optimal level that was hoped for. Another research by De Vries and Van Marle (2015) states that QE can be effective if implemented with the right timing, the ECB however was late, implementing QE when the economy was already recovering. This gave distortionary effects on the financial markets and government fiscal policies. So, the QE programs could have been effective as true timing some negative effects could have been prevented.

According to Kapitanos et al (2012), without the asset purchases implemented by the Bank of England in their first phase of quantitative easing program during March 2009 and January 2010, real GDP would have been even lower, and annual CPI inflation would have reached levels close to zero or even become negative. In this paper, tree models are used to determine the effects of QE on long term interest rates. Even though effects vary across the models, all are shown positive. From which is concluded that QE is a positive monetary policy tool to use in a financial crisis.

According to Neely (2011) the announcement of the QE program that would be implemented by the FED reduced the long-term yields of both U.S. bonds and foreign bonds. Furthermore, it led to a decrease in the spot value of the dollar. The changes in yields after the announcement were much too large that to be considered by chance. An explanation for this effect is given by the Signalling channel. The market interpreted the announcement of the QE program to come as a sign for bad news of the world economy. Which led to financial safety runs toward international bonds. In summary, evidence shows the announcement of the QE program in the U.S. had a strong portfolio balance effect on bond yields. It could also have caused an asset run towards foreign assets which led to a magnified effect on U.S. bond yields but reduced the predicted exchange rate effects.

According to Claeys et al (2015), the QE programs in the euro area did had a significant impact on the European public finances since mid-2014. The first program was launched in 2015; this would suggest the impact before this date was through the signalling channel, so a significant drop in yields because of the market anticipation of the QE program that still had to come. Furthermore, they state that with a quicker implementation of QE programs there could have been larger effects, and it could also have reduced the debt burden of the Euro-area governments, as the interest rates were higher.

Thornton 2014 did a research on the account of the portfolio balance channel, and in what terms its effect would be significant. There was no significantly statistical effect found between the risk premium and ten different public debt supply measures in the US, which represent the quantity of long-term assets. Hence, no empirical evidence has been found that the large-scale purchases of long-term assets lower the yields and risk premiums as the portfolio balance channel suggests it should have. The portfolio rebalancing channel has been proven insignificant by this event study. This is contrarily to Gagnon et al (2010). In this research the Federal Reserve’s QE program appears to successfully reduce the risk premium on long-term assets. In addition to this reduction the effect of the QE program seems to have an even greater effect on long-term interest rates on agency debts, this by improving market liquidity and by removing assets with high risk from private portfolio’s. Based on this evidence, Gagnon et al concludes the FED’s QE programs were successful at lowering long- term interest rates and stimulating economic activity. As both implications are from the US the results on the effect of the portfolio rebalancing channel are contradicting. So, the effects of the portfolio rebalancing channel should be questioned, regarding the possible effect it could have in the euro area.

To conclude, according to these studies QE programs have had effects through different channels. The significance of these effects, and the channels they apply to

however, differ across the studies. Therefore, the question remains if the implementation of QE in the euro area has had a significant effect on the recovery of the economy. In other words, has the QE program been effective? And if it has, through which channel this

(positive) effect has been realized. In particular this thesis will discuss what the effect of QE has been on the statistically relation between long-term interest rates and inflation within the euro area. When significant effects are found a discussion will be provided on the possible channels through which this effect has been realized. The main reason for the ECB

to implement QE was to reach the goal of a 2% inflation level. Based on the findings in this thesis I will be able to state if QE has had a significant effect on a (possible) change in this relation, and how far it helped to achieve the ECB goal.

Methodology & Data

In this thesis two approaches will be used to examine the implementation effects of QE by the ECB on long-term interest rates of government bonds. The first approach is used to determine the effects of QE on the long-term interest rates and to distinguish the channels through which the effects are realised. The second approach is used to examine if QE affects the statistical relation between inflation, economic growth and the long-term interest rates. With the results of this approach will be concluded if QE contributed to an increase in the inflation level.

First approach:

The first approach is an event study. Event studies look at the movement of a data series over a time frame, the so-called estimation window, and try to statistically assess the impact of an event. There are typically two approaches (MacKinlay, A, 1997). In this thesis the approach of a constant mean return model will be used. The constant mean model is probably the most simplistic model for measuring normal performance. Yield results found with the use of these models are similar to more sophisticated models. The models lack of sensitivity leads from the variance of the abnormal returns which is often not reduced with a more sophisticated model (MacKinlay, A, 1997). The second approach that could have been used is the market model. This model is a potential improvement on the constant mean return model because the variance of the abnormal return in the market model is lower than of the constant mean return model, this leads to a better ability to detect event effects. The market model would have been preferred if this thesis would look at the movement of a data series of one country within the euro area, using the other countries as a portfolio. Instead this thesis uses the combined average data of the 19 euro countries, to examine the effect of QE on the long-term interest rates within the euro area as a whole. Because

averages are used the constant mean return model is preferred (MacKinlay, A, 1997). Furthermore, is the combined average long-term interest rate of the 19 euro countries also used in the second approach, so consistency on results and conclusions is obtained using the same dependent variable in both approaches.

In this event study there will be two event dates with comprehending data sets. The first data set will be referred to as the data set of announcement, used on the first event

date, which is 22 January 2015 when the first QE program was announced. Secondly, the data set of implementation used for the second event date which is 10 March 2015 when QE was first implemented. With the use of this study I will be able to show if the movement in the series after the stated date is statistical significantly different from the estimation period (MacKinlay, A, 1997).

The methodology of the event study is derived from (MacKinlay, A, 1997). It contains the following steps:

Firstly, an event date is chosen with corresponding event window. In this case there are two event dates, with two windows. These dates are the day of announcement 22 January 2015 and the day of implementation 10 March 2015. With both event windows containing 10 days before and 10 days after the event date. Secondly, an estimation window is used for both estimations which exist from 95 days till 10 days before both events. Thirdly, using the constant mean model, the estimated normal returns are calculated. The normal returns show the expected returns if no event would take place. It is calculated with R(i,t) = 𝜇(i) + 𝜁(𝑖, 𝑡). Where 𝜇(i) is the mean return for interest rate I and 𝜁(𝑖,𝑡) is the disturbance term for interest rate I with an expectation of zero. Given the normal returns, the sample abnormal returns are calculated. The abnormal returns give the actual ex-post return of the long-term interest rate over the event window minus the normal return. The abnormal is calculated with AR(I,t) = R(I,t) - 𝜇̂(i) which gives the disturbance term of the constant mean return model

based on the sample. Fourthly, using the summation of the abnormal returns gives the cumulative abnormal returns (CAR). Dividing the CAR by its standard error gives the comprehending t-value to test the result on significance.

The estimation results of both event windows give the opportunity to see if one or more channels of QE have been of significant effect on the changes in long-term interest rates. The channels which are measurable are the signalling and the portfolio rebalancing channel. The signalling channel will be measured with the time frame of announcement, and the portfolio rebalancing channel with the time frame of implementation. The null

hypothesis related to both studies:

H0: The event has no significant effect on the long-term interest rate.

If normality of the data set is proved using a Jarque-Barre test which is based on the cumulative abnormal returns, a regular T-test can be used (MacKinlay, A, 1997). So, values of the T-statistic outside the interval [-1,96:1,96] will prove significand on a α = 5% significance level.

Second approach:

The second approach is with the use of two simple regression models containing multiple macro-economic variables. The variables are GDP that capture real growth and CPI that captures inflation rates. With these models there can be determined if the long-term

interest rate is dependent on changes in the real growth and inflation rate. Furthermore, will the lagged long-term interest rate be included. The reasoning behind this is that the lagged interest rates include much information about the current economic situation with respect to growth and inflation (Kapetanios et al 2012).

The variables in the models are based on those used by Kapetanios et al (2012). Kapetanios et al (2012) used the equations shown below to conduct multiple vector auto regression models to analyse the effects of QE implemented by the BoE. The difference in this thesis will be that the equations will be used within two simple regression models to examine the effect of QE implementation by the ECB on the relation between the

economic growth, inflation and the long-term interest rates. Only one model will contain the lagged interest rate as a variable. To have one regression without the lagged interest rate is to find the explanatory power of long-term interest rates on growth (measured by GDP) and inflation (CPI). The models will look as follows:

Model I:

Rt = β0 + β1 ΔGDP(t) + β2 ΔCPI(t) + ε(t )

Model II:

Rt = β0 + β1 ΔGDP(t) + β2 ΔCPI(t) + β3 R(𝑡 − 1) + ε(t)

The dependent variable Rt is the long-term interest rate on government bonds with

a ten-year maturity at time t. This bond type is used as it is the benchmark for the long-term interest rates (Kapetanios et al, 2012). Furthermore, is β0 the constant in the model.

β1 is the influence by real GDP (ΔGDPt) at time t on the long-term interest rate with a value

between [0;1]. β2 is the influence by the consumer price index (ΔCPIt) on the long-term

last variable is the lagged interest rate (Rt-1) by time t. The influence on the long-term

interest rate by this variable is measured by β3 with a value between [0;1].

The method to estimate the coefficients and test its significance will be with the use of an OLS regression performed in Stata. The calculated values will be tested against a 5% significance level (α = 5%). This means that the estimated coefficients are significant when the T-value lies at or outside the interval [-1,96;1,96]. To test if the estimated coefficients are independently significant the following hypothesises are constructed for both models:

H0: β1 = 0 H1: β1 ≠ 0

H0: β2 = 0 H1: β2 ≠ 0

H0: β3 = 0 H1: β3 ≠ 0

When proved significant the estimated coefficients will be used to determine the expected long-term interest rate from 2008 till the third quarter of 2017, to examine if there are significant changes in the relation within the crisis and expected values will be calculated from the first quarter of 2015 till the third quarter of 2017, to extinguish the particular change in the relation after implementation of QE.

Estimation model I:

R

̂t = β̂0 + β̂1 ΔGDP(t) + β̂2 ΔCPI(t )

Estimation model II: R

̂t = β̂0 + β̂1 ΔGDP(t) + β̂2 ΔCPI(t) + β̂3 R(t − 1)

Estimated values will be compared with the actual long-term interest rates of those periods to determine if the implementation of QE caused a change in the relation between the variables. To determine if the expected values are significantly different than the actual values, a Pearson Chi- square test will be used.

𝛸

= ∑

(𝑅̂𝑡 − 𝑅𝑡)

2

𝑅̂𝑡

𝑡

17

The outcome will be tested against the critical value of the chi-squared distribution with t-1 degrees of freedom.

The data used for the OLS regression is retrieved from the database of the Organisation for Economic Cooperation and development (OECD). This data consists of the combined information of the 19 euro countries, where GDP was retrieved with quarterly changes, and the long-term interest rates and CPI where retrieved annually, but redesigned to quarterly changes. The data set starts with the second quarter of 1995 and ends with the last quarter of 2007. The starting point has been chosen because of the quarterly GDP data points, which availability starts from the third

quarter of 1995. The end point is the last quarter of 2007 as the start of 2008 is taken as the starting point of the crisis before the announcement of QE by the ECB. This gives the data set 51 observations.

The data used for the event study is retrieved from Eurostat. This data contains the combined daily data on the long-term interest rate of the 19 euro countries. The data retrieved starts at 1 October 2014 and ends at 1 June 2015. The data set contains 262 observations. For the two studies the data set is split, giving both studies 3 months before and after the event window. The data used for the examination of the signalling channel will be referred to as data set of announcement. The data used for the portfolio rebalancing channel will be referred to as the data set of implementation. The event window contains 10 data points before and after the event. Because of missing data points the event window range of announcement is 15 days before and 13 days after the event. In case of the data set of implementation this is 14 days before and 11 days after the event.

18 1. Prove for normality is shown in the appendix by graphical display.

Results:

First approach:

As described in the methodology, two event windows are used to test the

significance of the effect on the long-term interest rates through two QE channels. The first window is 10 data points before and 10 days after the 22 of January, which is the day of announcement. The second window is 10 data points before and 10 days after the 10 of March when QE was implemented. Following the methodology, Normality for both data sets is proven using the Jarque-Berra test1. Corresponding T-values of the event dates are calculated using the standard t-statistic. Calculated values are displayed in the table below.

Announcement Implementation

Table 1: Abnormal returns, cumulative abnormal returns and t-values of corresponding events. ***, **, * denote significance at 1, 5 and 10 percent levels respectively

Event window AR CAR T-value AR CAR T-value

-10 -0.31 -0.31 -1.82* -0.26 -0.26 -1.53 -9 -0.29 -0.61 -1.70* -0.22 -0.47 -1.29 -8 -0.26 -0.87 -1.53 -0.20 -0.67 -1.17 -7 -0.33 -1.20 -1.94* -0.25 -0.91 -1.47 -6 -0.37 -1.58 -2.17** -0.27 -1.18 -1.58 -5 -0.41 -1.99 -2.40** -0.30 -1.48 -1.76* -4 -0.46 -2.45 -2.69*** -0.41 -1.88 -2.42** -3 -0.47 -2.93 -2.75*** -0.43 -2.31 -2.54** -2 -0.46 -3.39 -2.69*** -0.43 -2.74 -2.54** -1 -0.42 -3.81 -2.46** -0.40 -3.13 -2.36** 0 -0.41 -4.23 -2.40** -0.41 -3.54 -2.42** 1 -0.56 -4.79 -3.27*** -0.36 -3.89 -2.12** 2 -0.57 -5.36 -3.33*** -0.35 -4.24 -2.06** 3 -0.55 -5.92 -3.21*** -0.38 -4.62 -2.24** 4 -0.54 -6.46 -3.15*** -0.35 -4.96 -2.06** 5 -0.56 -7.02 -3.27*** -0.33 -5.29 -1.94* 6 -0.56 -7.59 -3.27*** -0.32 -5.61 -1.88* 7 -0.58 -8.17 -3.38*** -0.31 -5.91 -1.82* 8 -0.56 -8.74 -3.27*** -0.31 -6.22 -1.82* 9 -0.56 -9.30 -3.27*** -0.33 -6.54 -1.94* 10 -0.55 -9.85 -3.21*** -0.36 -6.90 -2.12**

19 The table shows the abnormal returns are constantly lower than the normal returns. During

the event window the change in long-term interest rates is significantly lower than if there was no event at the time. Both event dates are proven significant at a 5% level. It is clear that the days towards both events, especially the event of announcement also show significant abnormal returns. An explanation for this result could be that the market anticipated on the upcoming events through for example unofficial statements or leaked information. The H0

hypothesises for both data sets are rejected. With the use of this event study the effect of QE through both the signalling as the portfolio rebalancing channel on the long-term interest rates are proven significant.

Second approach:

In the theory is explained how QE works. Lowering the long-term interest rates should support economic growth and increase the inflation level (ECB, 2017). With the second approach in this thesis I tried to evaluate if this way of events is indeed the case. This is done by examining if macro-economic variables containing information of economic growth and inflation indeed have a significant relation with the long-term interest rate.

As described in the methodology does this thesis contain of two models. These models try to capture the effect of the quarterly changes in multiple macro-economic variables on the Long-term interest rate. OLS regressions are used to find the corresponding coefficients of the variables and to test its significance. The coefficients are estimated by using Stata and are displayed in the table below:

20 2. Actual and expected long-term interest rates shown in the appendix

R Coefficients (1) Std Error (1) Coefficients (2) Std Error (2) Constant -0.054*** 0.012 -0.043*** 0.012 GDP 0.068*** 0.021 0.054*** 0.020 CPI 0.141*** 0.046 0.134*** 0.048 Rt-1 - - 0.204** 0.12 Number of Obs Prob > F Adjusted R-squared 51 0.0003*** 0.3099 51 0.0012*** 0.3445

Table 2: Number between brackets displays the model corresponding to the values. ***, **, * denote significance at 1, 5 and 10 percent levels respectively.

The calculated beta values give the following estimation models:

Δ

R

̂t

= −0.054 + 0.068 ΔGDP(t) + 0,141 ΔCPI(t )

(5)

Δ

R

̂t

= −0.043 + 0.054 ΔGDP(t) + 0,134 ΔCPI(t ) + 0,204 𝑅(𝑡 − 1)

(6)

For both models all the coefficients have T-values outside the 5% significance interval {-1,96;1,96], hence the H0 hypothesises are rejected and all coefficients are proved significant.

Both models have significant F values. So both models are statistically able to predict future long-term interest rates2. The R-squares of both regressions can be compared as both dependent variables are the same. The adjusted R-squared of the second model is higher, so the variance of the model is better explained by the variables it contains. This states the seconds models predictions would be more precise. Performing the chi-squared tests for both models gives the following outcome:

Period χ² (1) χ² (2) Number of Obs

2008Q1-2015Q1 5.06 3.43 27

2015Q1-2017Q4 1.66 17.58** 11

Table 3: Number between brackets displays the model corresponding to the values. The subscripts added to the periods denote the quarter as starting point from which the year is taken into account. ***, **, * denote significance at 1, 5 and 10 percent levels respectively.

The Chi-square test values provided by both models give contradicting results. In model 1 the expected values of the long-term interest rates are not in any of the tested periods significantly different from the actual values. Expected values in the second model do not prove significant in the period 2008-2015. This means the crisis did not affect the relation between the tested variables. However the second model does prove significant on changes in the long-term interest rates after the implementation of QE. This states that QE did have a statistically significant effect on a change in the relation between the tested variables. So both models have contradicting outcomes. The adjusted R-squared of the second model is higher than that of the first one, this could state the variance of the second model is better explained by its variables. So the second model could be ‘better’ at predicting values, and the outcome of model two could be taken higher into account. But it is still hard to conclude which model would be right and if the statistical relation between the variables after the implementation did indeed change.

22

Conclusion

This thesis gave an insight in the theoretical framework around Quantitative Easing. Explaining which macro-economic variables are affected by QE and how it is used by central banks to achieve certain goals. Furthermore is described how multiple channels exist

through which QE has its possible effects. This thesis made use of two approaches to

examine the realised effects of the QE program implemented by the ECB. The first approach was an event study to determine the effects of QE on the long-term interest rates through the two major channels, the signalling channel and the portfolio rebalancing channel. The second approach was used to examine if QE affects the statistical relation between inflation, economic growth and the long-term interest rates. With those results an answer could be given on the research question of this thesis.

"What are the effects of the implementation of Quantitative Easing Programs on the relation between long-term interest rates and inflation within the Euro area?’’

With the event study is proved that QE did have a significant effect on the changes in the long-term interest rates during the periods of Announcement and Implementation of the QE program. The second approach gave contradicting results, which makes it hard to

conclude on the research question. One model did prove a statistical change in the relation between inflation economic growth and the long-term interest rates. The other model showed no significant change in this relation. Model two showed a higher adjusted R-squared which makes its prediction more statistically relevant over the other model. Based on this model is concluded that the relation between the long-term interest rates and inflation did change after the implementation of QE. From this result is concluded that QE did contribute at increasing the inflation rate since the moment of implementation. Furthermore does the uncertainty of QE implementation effects based on the contracting results found in the two models be an explanation that QE did not give the substantial increase in inflation towards the 2% goal the ECB had hoped for.

To conclude, the findings in this thesis show that QE did have a significant effect on the long-term interest rates but it remains uncertain if this measure is capable to help achieve the 2% inflation level goal anytime soon.

Discussion

This thesis consists of multiple caveats. The event study performed made use of an event window containing 21 days in total. The channel effects of QE do have effects on the short, but also on the long-term. With the event study the long-term effects are not taken into account in this thesis. Furthermore did this thesis not look at other possible events during the stated event windows that could have been of influence on changes in the long-term interest rates. This thesis contains two models up to three variables. More variables should be added to give more precise estimations on the effects of QE. Furthermore do regressions of the models not consist of many observations. More data should be gathered to give the variables a better ability to make more precise and reliable predictions.

24

References

Campbell, J. R., Evans, C. L., Fisher, J. D., & Justiniano, A. (2012). Macroeconomic effects of Federal Reserve forward guidance. Brookings Papers on Economic Activity, 2012(1), 1-80.

Claeys, G., Leandro, Á., & Mandra, A. (2015). European Central Bank quantitative easing: the

detailed manual (No. 2015/02). Bruegel Policy Contribution.

Claeys, Grégory; Leandro, Álvaro; Mandra, Allison (2015) : European Central Bank quantitative easing: The detailed manual, Bruegel Policy Contribution, No.2015/02.

Claeys, G., & Leandro, A. (2016). The European Central Bank’s quantitative easing programme: limits and risks. Bruegel Policy Contribution, 4.

Dunkly, J (2012, July 26). Debt crisis: Mario Draghi pledges to do 'whatever it takes' to save euro retrieved from telegraph.co.uk:

http://www.telegraph.co.uk/finance/financialcrisis/9428894/Debt-crisis-Mario-Draghi- pledges-to-do-whatever-it-takes-to-save-euro.html

ECB (2009).Conventional and unconventional monetary policy retrieved from: https://www.ecb.europa.eu/press/key/date/2009/html/sp090428.en.html

ECB (2017). Measuring inflation – the Harmonized Index of Consumer Prices (HICP) retrieved from ecn.europa.eu:

https://www.ecb.europa.eu/stats/macroeconomic_and_sectoral/hicp/html/index.en.html

Eurostat (2017). HICP- inflation rate, annual avrage rate of change (%) table.

Financieel dagblad 2017 Joost van den Kuppeveld ‘markt kijkt naar Dragi voor beleid opkoop programma’s.

Gambacorta, L., Hofmann, B., & Peersman, G. (2014). The effectiveness of unconventional monetary policy at the zero lower bound: A cross‐country analysis. Journal of Money, Credit

and Banking, 46(4), 615-642.

Federal Reserve: did they work?

Joyce, M., Miles, D., Scott, A. & Vayanos, D. (2012). Quantitative Easing and Unconventional Monetary Policy – an Introduction.

Kapetanios, G., Mumtaz, H., Stevens, I., & Theodoridis, K. (2012). Assessing the economy‐ wide effects of quantitative easing. The Economic Journal, 122(564).

Krishnamurthy, A., & Vissing-Jorgensen, A. (2011). The effects of quantitative easing on

interest rates: channels and implications for policy (No. w17555). National Bureau of

Economic Research.

Lautenschläger, S. (2014, July 7). Low inflation as a challenge for monetary policy and financial stability?

https://www.ecb.europa.eu/press/key/date/2014/html/sp140707.en.html

MacKinlay, A. C. (1997). Event studies in economics and finance. Journal of economic

literature, 35(1), 13-39.

Martin, J, & Zhang, J. (2017). Impact of QE on European sovereign bond market.

Neely, C. J. (2010). The large scale asset purchases had large international effects (pp. 1-45). Federal Reserve Bank of St. Louis, Research Division.

Thornton, D. L. (2014). QE: is there a portfolio balance effect? Federal Reserve Bank of St.

Louis Review, 96(1), 55-72.

26

Appendix

Graph 1: outcome of the Jarque-Berra test for normality on the data set of announcement

Graph 1: outcome of the Jarque-Berra test for normality on the data set of implementation

2,5 1,5 0,5 -0,60 -0,50 -0,40 -0,30 -0,20 -0,10 -0,50,00 -1 Reeks1 -0,70 -1,5 -2 -2,5 2,5 1,5 0,5 -6,00 -4,00 -2,00 _-0,50,00 -1 Reeks1 -8,00 -1,5 -2 -2,5

Table 4: Actual and expected values long-term interest Table 5: actual and expected long-term interest rates based on model 2 rates based on model 1

R E[r] R-E[r] 4.1472 4.403785 -0.25659 4.5051 4.405981 0.099119 4.604933 4.343978 0.260955 4.166767 4.295318 -0.12855 4.1459 4.110058 0.035842 4.1825 3.812893 0.369607 3.949467 3.647425 0.302042 3.837567 3.254471 0.583096 4.065134 2.937148 1.127986 3.8405 3.123769 0.716731 3.512733 3.202857 0.309876 3.705367 3.183485 0.521882 4.299933 3.179411 1.120522 4.465933 3.222137 1.243796 4.278533 3.226428 1.052105 4.186267 3.189006 0.997261 3.643833 3.130649 0.513184 3.437467 3.063994 0.373473 2.891667 2.964507 -0.07284 2.216967 2.906694 -0.68973 2.758933 2.796042 -0.03711 2.870233 2.661673 0.20856 3.2014 2.66116 0.54024 3.215667 2.639882 0.575785 3.062267 2.58138 0.480887 2.480733 2.536091 -0.05536 2.000133 2.471999 -0.47187 1.5873 2.363715 -0.77642 1.146667 2.236238 -1.08957 1.2854 1.79655 -0.51115 1.468433 1.500471 -0.03204 1.184133 1.435856 -0.25172 1.028633 1.56058 -0.53195 0.937933 1.406209 -0.46828 0.656667 0.956438 -0.29977 1.099267 0.364768 0.734499 1.406333 0.52824 0.878093 1.1708 0.841734 0.329066 1.1224 0.874261 0.248139 R E [r] R-E[r] 4.1472 4.40164 -0.25444 4.5051 4.407421 0.097679 4.604933 4.338876 0.266057 4.166767 4.253967 -0.0872 4.1459 4.026243 0.119657 4.1825 3.692418 0.490082 3.949467 3.50755 0.441917 3.837567 3.086867 0.7507 4.065134 2.762295 1.302839 3.8405 2.96076 0.87974 3.512733 3.03306 0.479673 3.705367 3.020845 0.684522 4.299933 3.032801 1.267132 4.465933 3.066488 1.399445 4.278533 3.027669 1.250864 4.186267 2.971308 1.214959 3.643833 2.90633 0.737503 3.437467 2.8305 0.606967 2.891667 2.740992 0.150675 2.216967 2.681927 -0.46496 2.758933 2.586143 0.17279 2.870233 2.482453 0.38778 3.2014 2.425968 0.775432 3.215667 2.390134 0.825533 3.062267 2.297999 0.764268 2.480733 2.24535 0.235383 2.000133 2.18086 -0.18073 1.5873 2.103806 -0.51651 1.146667 2.00803 -0.86136 1.2854 1.589255 -0.30386 1.468433 1.332123 0.13631 1.184133 1.234371 -0.05024 1.028633 1.331429 -0.3028 0.937933 1.207385 -0.26945 0.656667 0.757372 -0.1007 1.099267 0.149695 0.949572 1.406333 0.384589 1.021744 1.1708 0.567967 0.602833 1.1224 0.541155 0.581245