The impact of quantitative easing by the ECB on the 10-year term premium of Dutch government bonds

(1)

Faculty of Economics & Business

The impact of quantitative easing

by the ECB on the 10-year

term premium of Dutch government bonds

Michelle Nguyen

11045698

Abstract: The European Central Bank (ECB) resorted to unconventional

monetary policy in the wake of the recent global economic recession. The

most notable form of unconventional monetary policy is quantitative easing

(QE). The ECB’s QE program attempts to stimulate inflation by targeting

long-term interest rates. This paper tries to establish whether QE,

implemented by the ECB, has successfully lowered the term premium of

10-year Dutch government bonds. To research the effect, several ordinary

least squares regressions are performed for the period January 2005 till

March 2018. The results from this study indicate a statistically significant

negative relationship between quantitative easing and the term premium.

University of Amsterdam

Thesis Supervisor: Cenkhan Sahin

(2)

5. RESULTS ___________________________________________________________________14 Discussion of Results 6. CONCLUSION ______________________________________________________________18 7. REFERENCES ______________________________________________________________19 8. APPENDIX _________________________________________________________________22 Statement of Originality

This document is written by Michelle Nguyen 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.

(3)

1. Introduction

The year 2008 marked the eruption of the global financial recession. During this recession the conventional monetary policy instruments seemed to be ineffective at achieving the European Central Bank’s (ECB) mandate of price stability. Conventional monetary policy operates through short-term interest rates which are affected by open market operations. The short-term interest rates reached the zero lower bound and became disconnected from market interest rates which made it challenging for the ECB to fight the threat of deflation. As financial markets in the Eurozone experienced more turmoil the ECB decided to follow the footsteps of the Bank of Japan and introduced unconventional monetary policy measures. The most notable unconventional policy instrument is known as quantitative easing (QE), which consists of large quantity asset purchases, and was introduced in March 2015. Through these asset purchases the ECB strives to increase asset prices and lower the long-term yield of those assets and assets with comparable characteristics to the ones purchased (Joyce et al, 2012).

The research surrounding QE and its impact on the term premium of long-term government bonds is extensive. However, most of the existing literature focuses on QE implemented by the Bank of Japan, Federal Reserve or the Bank of England. The ECB is the latest central bank to have implemented QE. The literature is limited since the ECB’s QE program is still active. Considering that the state of the economy is dependent on the success of QE, as QE is expected to stimulate economic growth and prevent deflation, it is of interest to enhance the research in this area (Joyce et al, 2011).

This paper aims to identify whether the ECB’s QE has successfully lowered the term premium of 10-year Dutch government bonds. According to Kim and Wright (2005), long-term yields are composed of short-long-term yields and a long-term premium. The long-term premium reflects the interest rate risk associated with holding a security with a longer maturity. The focus in this paper lies with the term premium because Gagnon et al. (2011) argue that future short-term yields are expected to be kept consistently low. Therefore, any change in the long-term yield should result from a change in the long-term premium.

The effects of QE on the term premium will be estimated with an ordinary least squares (OLS) regression analysis. To establish whether the effects differ across euro area countries, the same regression analysis will be applied to the term premium of 10-year Italian government bonds. In order to compare the effects a panel data regression is applied. Italy is

(4)

chosen as its one of the PIIGS countries. The OLS regression results for The Netherlands as well as Italy show a negative influence of QE on the term premium. The panel data

regression findings also support the expected negative relationship. The QE coefficients from all the regression analyses are statistically significant at the 1 percent level.

This paper commences with an overview of previous literature in Section 2. Section 3 consists of a summary about the implementation of QE by the ECB and some theoretical background. Section 4 outlines the methodological approach as well as a description of the data. Section 5 offers an interpretation of the results obtained. Finally, Section 6 concludes.

2. Literature Review

In 1999 and the early 2000’s the Bank of Japan adopted a monetary policy that included a Zero-Interest Rate Policy (ZIRP) and a Quantitative Easing Monetary Policy (QMEP). Oda and Ueda (2007) provide an empirical study about the effect of this monetary policy on the medium- to long-term interest rates of government bonds. The paper

implements a macro-finance model which combines a macro model with the no-arbitrage asset pricing framework in finance. The macro-finance model, in combination with Monte-Carlo simulations, allows for a decomposition of the interest rates into a risk premium and expectations component. Once the interest rates have been decomposed, the effect of the signaling channel on the expectations component and the effect of the portfolio rebalance channel on the risk premium is determined with a maximum likelihood function. From the results Oda and Ueda (2007) conclude that the Bank of Japan’s monetary policy effectively lowered the medium- to long-term interest rate through the signaling channel. The portfolio rebalance channel had no significant effect on the risk premium component.

Joyce et al. (2012) assess the impact of QE1 and QE2 by the Bank of England on gilt yields. The paper emphasizes that the main impact of QE1 and QE2 on gilt yields is more likely to occur at the time of purchase announcement rather than at the time of purchase implementation. Following that reasoning that paper commences with an event study that focuses on the immediate change of gilt yields in response to the QE announcements. The results indicate that the first round of QE decreased gilt yields by approximately 100 basis points. According to Joyce et al. (2012) this yield reduction could be traced back to a reduction of liquidity premia on gilts and increased scarcity of gilts. In comparison with the first round of QE, the results for QE2 show a smaller impact on gilt yields, even when the

(5)

different volumes of gilt purchases are accounted for. On average the gilt yields actually increased. Joyce et al. (2012) offer various explanations for this increase in gilt yields. First, due to earlier communications from the monetary policy committee, a continuation of asset purchases might have been anticipated. Another reason for the increase could be that during the QE2 period other economic news emerged which potentially also affected the gilt yields. Therefore, it is difficult to distinguish the movement in gilt yields that is attributed to the announcement of QE2.

Furthermore, Joyce et al. (2012) investigated the impact of the two QE

announcements on other financial assets. The results show that the decrease in corporate bond yields was larger during QE1 than during QE2. For the first round the decrease amounted to 150 basis points and the second round showed a decrease of 40 to 50 basis points. The difference in results could reflect an increase in concerns about the euro area which depressed the prices of riskier assets. It could also be possible that investors are less willing to rebalance their portfolio with riskier assets. This effect is stronger during a worse outlook for the state of the economy and heightened risk aversion. Thus, the portfolio balance effect on the corporate bond yields may experience a delay. Joyce et al. (2012) conclude by acknowledging that it is extremely difficult to isolate the QE announcement effects in gilt yields from other factors. Therefore, the results are uncertain which makes it hard to draw conclusions.

Christensen and Rudebusch (2012) analyzed the decline in government bond yields as a result of the Fed’s large-scale asset purchases (LSAP) and the Bank of England’s QE announcements. In particular, the paper aims at identifying which transmission channel dominated during the decline of the government bond yields. The paper notes that the effects of the LSAP and QE announcements on the overnight index swap (OIS) were contradicting. The OIS rates in the US have decreased proportionately with US Treasury yields, while the OIS rates in the UK only experienced a small decline in comparison to UK gilt yields. These observations suggest that different transmission channels impacted the government bond yields. To identify through which channels the purchase programs influenced the government bonds Christensen and Rudebusch (2012) applied an event study methodology with the use of a dynamic term structure model (DTSM). This model decomposes the UK and US yields into a term premium and a short-term interest rate. The results indicate that the Fed’s LSAP mostly focused on lowering expectations surrounding future short-term interest rates, thus causing a decline in US Treasury yields through the signaling channel. In contrast, the UK QE announcements induced a decline in gilt yields through a reduction in term premiums.

(6)

Therefore, the UK gilt yields were affected through the portfolio balance channel. According to Christensen and Rudebusch (2012) possible explanations for the difference in transmission channels are related to financial markets structure and policy communication.

In addition, Christensen and Rudebusch (2012) analyzed cross-country effects on government bond yields. A market-wide signaling channel exist when an announcement of bond purchases by US leads to lower expectations for future policy interest rates in the UK. Alternatively, a cross-country portfolio balance channel is present when an announcement of bond purchases in the US leads to a higher probability of bond purchases in the UK. This causes a reduction in the supply of UK bonds which is associated with a decrease in term premiums. Christensen and Rudebusch (2012) focus on the response of UK gilt yields to the Fed’s LSAP announcements. The results suggest a spillover effect on UK gilt yields caused by the Fed’s monetary policy through the signaling channel. However, no evidence was found for a cross-country portfolio balance channel.

Gagnon et al. (2011) use two different approaches to measure the impact of the Fed’s LSAP on the market interest rates. The paper applies an event study analysis followed by an OLS regression analysis. The event study analyses the influence of eight announcements about LSAPs on the interest rate of various financial assets. The evidence indicates a decline of 150 basis points on long term interest rates during crucial LSAP announcements. Gagnon et al. (2011) performed a regression analysis that looks at the influence of variables such as core CPI inflation and publicly held Treasury bonds on the term premium of 10-year US Treasury bonds. The monthly data for this regression runs from January 1985 to June 2008. The results of the OLS-regression show a positive relationship for all the independent variables with the dependent variable. A more extensive description for the OLS-regression analysis is offered in the next section of this paper. The paper concludes by finding a reduction of 30 to 100 basis points in the term premium associated with LSAPs.

The literature outlined in this section discussed the effects of quantitative easing implemented by the Bank of Japan, the Federal Reserve and the Bank of England. The Bank of Japan, the Federal Reserve and the Bank of England implemented a QE program in 2001, 2008 and 2009 respectively. The existing literature is substantial because these central banks have concluded several rounds of QE over the years. The ECB on the other hand introduced QE relatively late. The QE program is still active as of March 2015. The availability of literature surrounding QE by the ECB is limited as the program kicked of only three years ago. This paper attempts to make a contribution to the limited literature about the ECB’s QE program by establishing its impact on the 10-year term premium of Dutch government bonds.

(7)

3. Literature

The ECB and Quantitative Easing

On 22 January 2015 Mario Draghi announced the ECB’s expanded asset purchase programme (APP), commonly known as quantitative easing. In addition to the asset-backed securities purchase programme (ABSPP) and the covered bond purchase programme

(CBPP3) the ECB would start purchasing securities issued by European governments under the public-sector purchase programme (PSPP). The securities under the PSPP are bought from financial institutions, such as banks, insurance companies and pension funds by printing money. The ECB announced that it would purchase a total amount of €60 billion worth of securities each month as from 9 March 2015 to September 2016. During this first

announcement of QE Draghi emphasized that the purchases would stop once the euro area inflation approached 2 percent and that the program would be extended if it failed to improve inflation as expected. From these statements it can be concluded that the APP was launched to meet the ECB’s primary objective of maintaining price stability (The European Central Bank, 2015).

A graph from the ECB with a timeline of the APP is shown below. The y-axis expresses the amount of asset purchases in billions of euro’s and the x-axis shows the timeframe for the APP. As shown by the legend, the monthly amount of asset purchases is composed of the PSPP, ABSPP, CBPP3 and CSPP.

(8)

The graph shows a slight increase in the asset purchases during the months May and June of 2015 and a decrease in July and August 2015. This is due to expected lower liquidity during the summer holiday period (Cœuré, 2015). On 3 December 2015 the ECB announced an extension of the APP until at least March 2017 (Draghi & Constâncio, 2015).

On 10 March 2016, the ECB decided to further expand the APP by increasing the monthly asset purchases from €60 to €80 billion (The European Central Bank, 2016). Furthermore, the central bank also announced it would initiate a program to purchase investment-grade corporate bonds, known as the corporate sector purchase programme (CSPP) (Draghi, Constâncio & Nowotny, 2016).

Moreover, on 8 December 2016, Mario Draghi announced that the ECB would continue the APP until the end of December 2017 and if deemed necessary longer. The monthly amount purchased would be maintained at € 80 billion until March 2017, after March the purchases would be reduced to € 60 billion. During this announcement Mario Draghi also stated that the chance of deflation “has largely disappeared” (Draghi & Constâncio, 2016).

In the preceding months to 9 March 2017 the inflation rate rose to 2 percent and thus reached the desired level. The ECB acknowledged that the APP had been successful but concluded that it would be premature to start reducing the amounts purchased because the inflation was driven by food and energy prices (Draghi & Constâncio, 2017).

On 26 October 2017, the ECB decided that the monthly purchases would be halved from January 2018 to € 30 billion (The European Central Bank, 2017). A further halving, starting September 2018, to €15 billion was announced during a press conference on 14 June 2018 (Draghi & de Guindos, 2018).

Monetary Transmission Channels

Quantitative easing is expected to operate through several monetary transmission channels. Joyce et al. (2011) make a distinction between the channels operating through asset prices and other channels. Asset prices are affected by the signaling, liquidity and portfolio rebalance channel. Other channels include the confidence and bank lending channel which operate through economic growth and inflation. Since this paper focuses on the term premium, only the channels that operate through asset prices will be considered.

(9)

Signaling Channel

Through this channel the central bank can influence interest rates by informing market participants about their intention for the future path of monetary policy. Market participants will adjust their expectations for future policy rates if the central bank signals a desire to keep the rates at the zero-lower bound. As the market has an immediate response to a change in expectations, this will be reflected in the yield of assets. However, Eggertsson and Woodford (2003) stress that unconventional monetary policy is only successful in reducing long-term bond yields when a central bank can credibly commit to preserving low interest rates even after a recovery of the economy. Clouse et al. (2000) believe that such a signal implies that the central bank implements QE and thus purchases large amounts of long-term assets. This gesture is perceived a credible since the central bank would incur a loss on the assets

purchased if the bank decides to raise the interest rate again. Furthermore, according to Krishnamurthy and Vissing-Jorgensen (2011) the signaling channel influences bond market interest rates through the expectations theory. The expectations theory states that long-term interest rates are composed of current interest rates and a forecast of future short-term interest rates, which are expected to be kept low.

Liquidity Channel

During times of deteriorated economic outlook, the functioning of financial markets may be impaired due to credit strains and poor liquidity of assets. If the central bank implements QE during those times, it will provide a consistent demand for long-term assets. The presence of QE ensures active trading because it allows investors to take on long-term securities knowing that the securities could be sold to the central bank when needed (Gagnon et al, 2011). Hence QE improves the market’s trading conditions which leads to enhanced liquidity and

consequently a reduction in the liquidity risk premium. However, as these effects stem from a consistent presence of QE, there is a possibility that the effects of the liquidity channel only persist while the central bank actively conducts asset purchases (Joyce et al, 2011).

Portfolio Rebalance Channel

When the central bank purchases a large amount of assets held by the private sector, the relative free float of the assets purchased changes. These asset purchases are financed by creating money and the proceeds of the sale will appear on bank deposits. If it were the case that bank deposits and the assets purchased were regarded as perfect substitutes, the private sector would not have to rebalance their portfolio and it is uncertain whether yields would

(10)

change (Joyce et al, 2012). However, Tobin (1961) and Brunner and Meltzer (1973) argue that due to imperfect substitutability the central bank can influence the yield of different assets. Bank deposits and assets can be seen as imperfect substitutes due to the preferred habitats theory. If investors sell government bonds to the central bank, a long-duration asset, the government bond, is swapped for a short-duration asset, the bank deposits. This may be undesirable as the duration of their portfolio changes. Following the preferred habitat theory (Modigliani and Sutch, 1966), investors may want to match the duration of the assets with the liabilities and therefore attempt to reinvest the proceeds from the government bond sales into other riskier assets. This process will increase the price of these other assets, such as

corporate bonds, and reduce the associated yields and term premiums. Thus, the portfolio rebalance channel lowers the yield of the assets purchased as well as other riskier assets. The reduction in yields leads to a reduction in borrowings cost for firms and therefore eases credit conditions. Investors will experience an increase in wealth when the firms generate capital gains from the raised funds. If a part of the newly obtained wealth is consumed or invested in capital markets, demand will be stimulated which in turn could cause inflation to rise (Joyce et al, 2012).

4. Data & Methodology

Dataset and Construction

The dataset consists of monthly data on various macroeconomic variables for The Netherlands and Italy. Most of the data was available on a monthly basis. The data that was only available in daily or quarterly values have been converted to monthly values. For daily values this process involved averaging the daily values of each month whereas the quarterly values have been divided by three to obtain monthly values. The timespan for the data is January 2005 to March 2018. This period is chosen because it includes a period without QE followed by the implementation of QE in March 2015. The majority of the data is retrieved from the OECD. Other sources include Datastream and the ECB database.

(11)

The Econometric Model

To measure the impact of QE on the 10-year term premium of Dutch government bonds various macroeconomic variables are used. Following the research method of Gagnon, Raskin, Remache and Sack (2011) this paper will explain the historical time variation in the term premium with macroeconomic variables related to the business cycle, uncertainty about economic fundamentals and the net public sector supply of long-term debt securities. The set of variables consists of the unemployment gap, core consumer price index inflation, a long-term inflation disagreement, historical volatility, net issues of long-long-term debt security and asset purchases under the Public-Sector Purchase Programme. The last variable represents the impact of QE on the term premium.

In accordance with the research method of Gagnon et al. (2011), the historical

variation in the term premium for Dutch and Italian government bonds will be estimated with an ordinary least squares (OLS) regression. In additional, a panel data regression with fixed effects will be performed to compare the effects. Gagnon et al. (2011) also estimated the model with a dynamic ordinary least squares (DOLS) regression in order to estimate the long-term relationship between the term premium and the explanatory variables. This paper will only apply an OLS regression considering that the results from the DOLS regression were practically identical to the results from the OLS regression.

The baseline regression takes on the following form: 𝑡𝑝_𝑖,𝑡10_{= 𝛽}

0 + 𝛽1𝑈𝐺𝐴𝑃,𝑖+ 𝛽2𝐶𝑃𝐼𝑖+ 𝛽3𝐼𝑛𝑓𝐷𝑖𝑠𝑖+ 𝛽4𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦𝑖 + 𝛽5𝑃𝑢𝑏𝑙𝑖𝑐𝑖+ 𝛽6𝑄𝐸𝑖

The dependent variable, 𝑡𝑝_𝑖,𝑡10_{, is the term premium of the nominal 10-year yield. Kim}

and Wright (2005) defined the components of the nominal Treasury bond yield as “the sum of the compounded expected future short-term interest rate over the maturity of the bond and a risk or term premium” (p. 1). The risk or term premium is a compensation for investors as returns are uncertain when the bond is held over a period less than the maturity. Therefore, this paper measures the term premium as the difference between the yield of 10-year government bonds and the short-term yield; in this case the 3-month European Interbank Offered Rate is used. The data is retrieved from the OECD.

The following explanatory variables are included in the model to explain the term premium variation associated with the business cycle and fundamental uncertainty. The unemployment gap, 𝑈_𝐺𝐴𝑃, is defined as 𝑈_{𝐺𝐴𝑃,𝑖} = 𝑈_𝑡,𝑖− 𝑈_{𝑁𝐴𝐼𝑅𝑈,𝑖} with 𝑈_𝑡,𝑖 as the monthly

(12)

unemployment rate. The Non-Accelerating Inflation Rate of Unemployment is selected as an estimate for the natural rate of unemployment as this unemployment rate sustains a stable level of inflation (Walsh, 1998). The data on the NAIRU is only available per annum.

However, it still qualifies as an estimate for the natural rate of employment as it is considered a slowly adjusting variable (Walsh, 1998). This relationship between the term premium and unemployment gap is expected to be positive. According to Gil-Alana and Moreno (2012) a higher unemployment level suggests that there is uncertainty about the future state of the economy. This uncertainty is undesirable for investors, therefore the longer-term bonds become relatively less attractive which increases the term premium. The data is retrieved from the OECD.

The second variable is the core consumer price index inflation. Core CPI inflation is measured as the year on year percentage change in CPI excluding energy, food, tobacco and alcohol prices since these prices are relatively more volatile. Mankiw, Reis and Wolfers (2004) found a positive correlation after examining the relationship between the level of inflation and uncertainty. Wright (2008) investigated the relationship between inflation uncertainty and the term premium and found a positive relationship. Therefore, the variable core CPI inflation is expected to have a positive relationship with the term premium. The data is retrieved from the OECD.

Furthermore, the long-run inflation disagreement is included in the model. In the regression analysis of Gagnon et al. (2011) this variable represents the public’s expectations of the path for inflation for the next 5 to 10 years. These expectations have an influence on the term premium through inflation, thus the relationship is expected to be positive. In this paper the consumer opinion survey by Statistic Netherlands is used as a proxy for the long-run inflation disagreement. The relevant indicator in this survey is ‘consumer prices, future tendency’. This indicator looks at how the public perceives the development of consumer prices for the next 12 months, in comparison with the past 12 months. This is a useful proxy since the survey reflects investors’ rational expectations about the future path of inflation. The unit of measure is percentage and the data is retrieved from the OECD.

Lastly, the six-month realized daily volatility of the on-the-run 10-year Treasury yield is added to the model to account for interest rate uncertainty. When the uncertainty

surrounding the interest rate is high, it is expected to increase the term premium. Investors are usually risk-averse and require additional compensation for taking on a more volatile bond (Osterrieder & Schotman, 2017). The daily volatility of the yield of Dutch government bonds

(13)

is retrieved from the Bloomberg Database. These daily rates were converted into monthly rates by taking the average of the daily rates for each month.

Other explanatory variables in the model represent the impact of changes in the net public sector supply of long-term debt. The variable that Gagnon et al. (2011) included is publicly held Treasury securities with at least one year to maturity. This variable represents the debt that the government owes the public. Gagnon et al. (2011) included this variable as it accounts for the change in the yield through the price of government bonds. An increase in the supply of long-term government bonds will most likely lead to a decrease of the price for those bonds, simultaneously causing the yield to increase. Thus, a positive relationship is expected for this variable and the term premium. In this paper net issues of long-term debt securities issued by general government is used. Net issues are measured as gross issues minus redemptions. The general government is defined as the central, state and local governments and social security funds. The debt securities have at least one year left until maturity and are taken as a percentage of nominal gross domestic product (GDP). Data for nominal GDP was only available quarterly, thus the quarterly data was divided by three to get the monthly GDP. The data is retrieved from the database of the ECB.

The explanatory variable QE is added to the regression to capture the impact of the ECB’s net purchases of debt securities. QE affects the term premium by driving up the price of long-term government bonds and thus decreasing the overall yield on those bonds. The ECB implemented QE on March 2015; the data follows from the Public-Sector Purchase Programme. Data on this variable runs from March 2015 to March 2018, the QE variable will take on the value of 0 before March 2015. This variable is normalized by nominal GDP. The data is retrieved from Datastream.

(14)

5. Results

Table 1: OLS regression for The Netherlands Coefficient

Constant 1.494 (0.209) *** Unemployment Gap -0.005 (0.064) Core CPI 0.321 (0.085) *** Inflation Disagreement -0.028 (0.005) *** Volatility -0.512 (0.299) * Public Debt 0.005 (0.005) QE -0.049 (0.018) *** Robust standard errors in parentheses.1

***, **, * statistically significant at the 1, 5, 10 percent level respectively.

Table 2: OLS regression for Italy Coefficient

Constant 1.004 (0.485) ** Unemployment Gap 0.407 (0.077) *** Core CPI 0.201 (0.315) Inflation Disagreement -0.009 (0.009)

1_{In section 8, the Appendix, the result for the Breusch-Pagan / Cook Weisberg Test is shown. The}

null-hypothesis is rejected at the 5 percent significance level. Therefore, the OLS regression is performed with robust standard errors. Moreover, the normality assumption is fulfilled. For the OLS-regression in table 1 and 2, the assumption of no serial correlation is not met. The model includes time-series data which often exhibits autocorrelation. A possible solution could be to select a different functional form. However, that goes beyond the scope in this paper.

(15)

Volatility 1.243 (0.206) *** Public Debt 0.008 (0.008) QE -0.157 (0.036) *** Standard errors in parentheses.2

Table 3: Panel Data Regression with fixed effects

Constant 0.885 (0.187) *** Unemployment Gap 0.315 (0.037) *** Core CPI 0.010 (0.108) Inflation Disagreement -0.013 (0.004) *** Volatility 1.274 (0.145) *** Public Debt 0.006 (0.005) QE -0.130 (0.023) *** Standard errors in parentheses.3

2_{The OLS assumption of normality and homoskedasticity is fulfilled. See section 8 Appendix, figure 4 and 5.} 3_{The Wooldridge Test for Autocorrelation is shown in figure 7 of the Appendix. The null-hypothesis is not} rejected at the 5 percent significance level.

(16)

Discussion of Results

In Table 1 the results from the OLS regression for The Netherlands are reported. According to the results QE has the expected negative impact on the term premium of 10-year Dutch government bonds. The result for QE is statistically significant at the 1 percent level. The coefficient of -0.049 implies that a 1-percent-of-GDP increase in QE causes a decrease in the term premium by 4.9 basis points. The asset purchases of the ECB during the period March 2015 to March 2018 amount to roughly €106 billion, which is approximately 6.76 percent of the nominal GDP in The Netherlands during the same period. According to the estimates of table 1, these asset purchases translate to a decrease of 33.1 basis points in the term premium.

The control variables exhibit varying results. The variables core CPI inflation and public debt have the expected positive effect. When the core CPI inflation increases with 1 percentage point, the term premium increases with 32.1 basis points. This result is

statistically significant at the 1 percent level and therefore consistent with the expectation described in section 4. For the variable public debt, table 1 shows a coefficient of 0.005. With a p-value of 0.301, the coefficient is statistically insignificant. Thus, the null-hypothesis – according to which the coefficient should equal 0 – remains intact.

In contrast to the results of Gagnon et al. (2011), the variables unemployment gap, inflation disagreement and volatility have a negative effect on the term premium. The result for the unemployment gap is insignificant. Therefore, the coefficient cannot be interpreted. The inflation disagreement on the other hand, has a p-value of 0. This suggests a statistically significant result at the 1 percent level. This result is inconsistent with the expectation of economic theory, according to which an increase in the expected inflation causes an increase in the term premium through a rise in inflation uncertainty. A possible explanation for the contrasting result could be the difference in data specification. Gagnon et al. (2011) defined this variable as the five- to ten-year inflation expectations of the public. This paper uses a proxy which measures the one-year inflation expectations of the public. Since the proxy has a shorter timespan it could have influenced the estimate for inflation disagreement in a

different manner. The result for the volatility is in contrast with the expected positive relationship. With a p-value of 0.089 the coefficient is statistically significant at the 10 percent level. This contradicting result could be explained by a relatively more attractive Dutch government bond. Investors may have favored the volatility of yield on 10-year Dutch government bonds when compared to the volatility of Italian or Greek government bonds during the same period. Therefore, investors could have perceived the Dutch government

(17)

bond to be more valuable since the volatility of the Dutch government bonds increases by a relatively low amount. If the investors act upon this sentiment and purchase the Dutch government bonds, the price of those bonds will rise and the yield will subsequently decline.

Table 2 shows the results from the OLS regression for Italy. In this regression

analysis the QE estimate equals -0.157 and is statistically significant at the 1 percent level. A 1-percent-of-GDP increase in QE leads to a 15.7 basis point reduction in the term premium of 10-year Italian government bonds. For the period March 2015 to March 2018 the ECB’s asset purchases summed up to roughly €337.2 billion. This amount is equivalent to 6.9 percent of nominal GDP in Italy for the same period. Therefore, QE has lowered the term premium of 10-year Italian government bonds by 108.3 basis points during the period March 2015-2018.

Most of the results for the control variables are in line with the findings from Gagnon et al. (2011). As predicted, the variables unemployment gap, core CPI inflation, volatility and public debt all have a positive coefficient. From these variables the coefficient of the

unemployment gap as well as the volatility are statistically significant at the 1 percent level. For the unemployment gap a percentage point increase would result in a 40.7 basis point increase in the term premium. An increase in the unemployment gap gives a signal of increased uncertainty about the state of the economy. As investors perceive risk to be undesirable, a compensation is required for holding the increased risk. The 40.7 basis point increase in the term premium represents this compensation. For the volatility, a 1 percentage point increase in the monthly volatility of the yield on 10-year Italian government bonds causes an increase in the term premium of 124.3 basis points. Following the same reasoning as with the unemployment gap, a more volatile bond is considered to be risky and therefore requires additional compensation.

The inflation disagreement coefficient is negative and resembles the result of the inflation disagreement coefficient in the OLS regression for the Netherlands (table 1). However, the result for the OLS regression of Italy appears to be insignificant. Therefore, it can be assumed that the variable inflation disagreement does not affect the term premium of 10-year Italian government bonds.

Lastly, table 3 reports the results from the panel data regression. The result for the QE coefficient is highly significant and corresponds with the results found in table 1 and 2. The QE coefficient equals -0.130, which indicates a 13 basis point decline in the term premium when QE increases by 1-percent-of-GDP. For the control variables the results are similar to the expectations formed in section 3. The unemployment gap, core CPI inflation, volatility and public debt possess a positive coefficient. Of these variables the unemployment gap and

(18)

volatility have a significant positive coefficient. Moreover, a negative coefficient is estimated for the inflation disagreement. The result is significant, and the negative relationship seems to persist along all three regression analyses.

6. Conclusion

This paper has attempted to study the impact of QE on the term premium of 10-year Dutch government bonds. The findings from the OLS regression analyses in section 5 support the prediction that QE has a negative effect on the term premium. A 1-percent-of-GDP increase in QE impacted the term premium of Dutch and Italian government bonds by 4.9 and 15.7 basis points respectively. In addition, the results for QE appear to be statistically significant at the 1 percent level. Furthermore, the results for other explanatory variables were partially in line with the formed expectations. A notable and contrasting result was found for the control variable inflation disagreement. The variable exhibited a persistent negative relationship with the term premium, occurring through all of the OLS regression analyses. In conclusion, the ECB has proven to be successful at lowering the term premium of long-term government bonds.

However, this paper cannot account for the full extent of the QE effect as the program has not been concluded yet. A further study, once the ECB has exited QE, could provide a better representation as the effects on the term premium will be examined in its entirety. Another suggestion for future research would be to examine the effect of QE on the term premium for more countries. In this paper the scope for the panel data regression concerns two European countries. To measure the overall effect of QE more accurately an expansion of the observations from the panel data regression is recommended.

(19)

7. References

Brunner, K., & Meltzer, A. (1973). Mr. Hicks and the "Monetarists". Economica, 40(157), 44-59.

Walsh, C. E. (1998). The Natural Rate, NAIRU, and Monetary Policy. FRBSF Economic

Letter 98-28. Retrieved from

https://www.frbsf.org/economic- research/publications/economic-letter/1998/september/the-natural-rate-nairu-and-monetary-policy/#role

Christensen, J., & Rudebusch, G. (2012). The Response of Interest Rates to US and UK Quantitative Easing *. Economic Journal, 122(564), F385-F414.

Clouse, J., Henderson, D., Orphanides, A., Small, D., & Tinsley, P. (2000). Monetary Policy When the Nominal Short-Term Interest Rate is Zero. Federal Reserve Official.

Cœuré, B. (2015, May). How binding is the zero lower bound? Paper presented at the Removing the zero lower bound on interest rates conference, London. Retrieved from http://www.ecb.europa.eu/press/key/date/2015/html/sp150519.en.html

Draghi, M., & Constâncio, V. (2015, December). Introductory statement to the press

conference (with Q&A). Paper presented at The European Central Bank press conference,

Frankfurt am Main. Retrieved from

http://www.ecb.europa.eu/press/pressconf/2015/html/is151203.en.html#qa

Draghi, M., Constâncio, V., & Nowotny, E. (2016, June). Introductory statement to the press

Vienna. Retrieved from

http://www.ecb.europa.eu/press/pressconf/2016/html/is160602.en.html

Draghi, M., & Constâncio, V. (2016, December). Introductory statement to the press

Frankfurt am Main. Retrieved from

(20)

Draghi, M., & Constâncio, V. (2017, March). Introductory statement to the press conference

(with Q&A). Paper presented at The European Central Bank press conference, Frankfurt am

Main. Retrieved from http://www.ecb.europa.eu/press/pressconf/2017/html/is170309.en.html

Draghi, M., & de Guindos, L. (2018, June). Introductory statement to the press conference

(with Q&A). Paper presented at The European Central Bank press conference, Riga.

Retrieved from http://www.ecb.europa.eu/press/pressconf/2018/html/ecb.is180614.en.html

Eggertsson, G. B., & Woodford, M. (2003). The zero bound on interest rates and optimal monetary policy. Brookings Papers on Economic Activity, (1), 139.

Gagnon, J. E., Raskin, M., Remache, J., & Sack, B. (2011). Large-scale asset purchases by the Federal Reserve did they work? Economic Policy Review,17(1), 41-59.

Gil-Alana, L. A., & Moreno, A. (2012). Uncovering the US term premium: An alternative route. Journal of Banking and Finance, 36(4), 1181-1193.

Joyce, M., Miles, D., Scott, A., & Vayanos, D. (2012). Quantitative Easing and

Unconventional Monetary Policy – an Introduction*. Economic Journal, 122(564), F271-F288.

Joyce, M., Mclaren, N., & Young, C. (2012). Quantitative easing in the United Kingdom evidence from financial markets on QE1 and QE2. Oxford Review of Economic Policy, 28(4), 671-701.

Joyce, M., Tong, M., & Woods, R. (2011). The United Kingdom's quantitative easing policy design, operation and impact. Quarterly Bulletin // Bank of England, 51(3), 200-211.

Kim, D., & Wright, J. (2005). An arbitrage-free three-factor term structure model and the recent behavior of long-term yields and distant-horizon forward rates. Federal Reserve

(21)

Krishnamurthy, A., & Vissing-Jorgensen, A. (2011). The Effects of Quantitative Easing on Interest Rates: Channels and Implications for Policy. Brookings Papers On Economic

Activity, 215-287.

Mankiw, N., Reis, R., & Wolfers, J. (2004). Disagreement about inflation expectations. Nber

Macroeconomics Annual 2003, 18, 209.

Modigliani, F., & Sutch, R. (1966). Innovations in Interest Rate Policy. The American

Economic Review, 56(1/2), 178-197.

Oda, N., & Ueda, K. (2007). The effects of the bank of Japan's zero interest rate commitment and quantitative monetary easing on the yield curve: A macro-finance approach. Japanese

Economic Review, 58(3), 303-328.

Osterrieder, D., & Schotman, P. (2017). The Volatility of Long-Term Bond Returns:

Persistent Interest Shocks and Time-Varying Risk Premiums. The Review of Economics and

Statistics, 99(5), 884-895.

The European Central Bank. (2015, January). ECB announces expanded asset purchase

programme [Press release]. Retrieved from

http://www.ecb.europa.eu/press/pr/date/2015/html/pr150122_1.en.html

The European Central Bank. (2016, March). Monetary policy decisions [Press release]. Retrieved from http://www.ecb.europa.eu/press/pr/date/2016/html/pr160310.en.html

The European Central Bank. (2017, October). Monetary policy decisions [Press release]. Retrieved from https://www.ecb.europa.eu/press/pr/date/2017/html/ecb.mp171026.en.html

Tobin, J., Karcken, J. H., & Brunner, K. (1961). Money, capital, and other stores of value. The American Economic Review, 51(2), 26-37.

(22)

8. Appendix

The Netherlands – OLS assumptions Figure 1: Jarque-Bera Normality Test

Figure 2: Breusch-Pagan / Cook Weisberg Test

Figure 3: Breusch-Godfrey LM Test

Italy – OLS assumptions

Figure 4: Jarque-Bera Normality Test

Figure 5: Breusch-Pagan / Cook Weisberg Test Jarque-Bera test for Ho: normality:

Jarque-Bera normality test: .0409 Chi(2) .9798

Prob > chi2 = 0.0463 chi2(1) = 3.97

Variables: fitted values of Termpremium Ho: Constant variance

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity

H0: no serial correlation

1 137.761 1 0.0000 lags(p) chi2 df Prob > chi2 Breusch-Godfrey LM test for autocorrelation

Jarque-Bera test for Ho: normality:

Jarque-Bera normality test: 3.68 Chi(2) .1588

Prob > chi2 = 0.8336 chi2(1) = 0.04

Variables: fitted values of Termpremium Ho: Constant variance

(23)

Figure 6: Breusch-Godfrey LM Test

Panel data regression assumptions

Figure 7: Wooldridge Test for Autocorrelation

H0: no serial correlation

1 129.262 1 0.0000 lags(p) chi2 df Prob > chi2 Breusch-Godfrey LM test for autocorrelation

Prob > F = 0.0766 F( 1, 1) = 68.490 H0: no first-order autocorrelation

The impact of quantitative easing by the ECB on the 10-year term premium of Dutch government bonds

Faculty of Economics & Business