HOW QUANTITATIVE EASING AFFECTS CORPORATE BOND YIELDS: AN EUROPEAN CASE

Transcription

1 HOW QUANTITATIVE EASING AFFECTS CORPORATE BOND YIELDS: AN EUROPEAN CASE by LUCA CARRIERI SUPERVISOR: prof. dr. FABIO CASTIGLIONESI CHAIRPERSON (SECOND READER): prof. dr. MICHEL R.R. VAN BREMEN

2 How Quantitative Easing Affects Corporate Bond Yields: An European Case Luca Carrieri ABSTRACT This paper examines the effects of Quantitative Easing on corporate bond yields. Through a high-frequency event-study based on two different Quantitative Easing programmes implemented by the European Central Bank, namely the Public Sector Purchase Programme and the Corporate Sector Purchase Programme, on the available population of corporate bond data for Germany and Italy, the paper finds that around the announcement date, there is a reduction of corporate bond yields, respectively for the two programmes, of around 11 and 8 basis points for both German and Italian bonds. The results are consistent across different categories of bonds, varying only in terms of the magnitude of changes. The programmes, therefore, seem to ease the financing conditions of the issuers.

3 INTRODUCTION In the last decades, the efficacy of textbook macroeconomics has been questioned by the depth and the persistency of financial crises, which led almost unanimously the governors of Central Banks to respond with non-conventional monetary policies. Among those, due to deflationary or disinflationary spirals, increased sovereign default risks and balance sheet recessions, it has become largely common for Central bankers to implement Quantitative Easing programmes as a response. While various aspects of the effects of Quantitative Easing on financial markets has been already investigated, such as whether it is a useful tool for alleviating the funding costs for sovereign issuance by lowering the yields, lower attention has been put into investigating whether those effects, if any, have a pass-through to the corporate sector. The aim of this paper is to investigate whether Quantitative Easing programmes may help the corporate sector as well by reducing their funding costs, more specifically, the investigation is undertaken on the effects of the recent Quantitative Easing programmes of the European Central Bank on selected Eurozone corporate bond yields. The first QE program analysed will be the Public Sector Purchase Programme, announced in January 2015 and started in March 2015, that added to asset-backed securities and covered bonds, the purchase of inflation-linked and floating rate securities issued by governments. The second QE program investigated will be the Corporate Sector Purchase Programme, announced in March 2016, and updated, with further specifics about what assets the programme would have included, in April 2016, based on the purchase for the first time by the European Central Bank of corporate bonds. Concerning the choice of two different Quantitative Easing programmes, the rationale behind it is based on the fact that different channels may be at work in affecting corporate bond interest rates, causing potentially different responses, so from a policy perspective it

4 may be useful to understand the reactions of the market to tailor, in the future, monetary policies according to the specific targets. The investigation will then be conducted on two countries, Germany and Italy. The reason for this choice relies firstly on the fact that the Eurozone is a way more heterogeneous aggregation compared to UK and Japan and to some extent also the US. For this reason, the analysis is not undertaken on a single country, rather on a twocountries comparison to test whether heterogeneous countries under the same monetary policy do respond differently. For a future research, the analysis may be then enlarged to a larger number of countries. The second point concerns the rationale for the specific countries chosen, Germany and Italy. The choice of Germany is based on consideration that it is the biggest economy of the Eurozone, with the largest financial markets, and it is widely considered as the benchmark for performances in the Eurozone case. The choice of Italy, on the other hand, is based on the fact that the country is the biggest economy (and third biggest economy overall of the Eurozone) and the largest (in size) financial market among the countries hit by the relatively recent sovereign debt crisis. This should reduce the risk that the effects may not be the result of a too small economy and too much undeveloped financial markets. The idea is thus having the possibility of comparing how a safe economy and a more volatile economy react to the same monetary policy strategies. Finally, the analysis will be based on responses of yields of all the population of corporate bond issued in the two countries for the period analysed, whose data are retrieved from Thomson Reuter Datastream. Conversely to the vast majority of the papers in the literature, corporate bonds will not be aggregated altogether, but rather they will be categorized in financial versus non-financial issued bonds, categorized for different (remained) maturity buckets and, for ratings.

5 These specifications will allow a deeper understanding of the mechanisms at work, avoiding compressing heterogeneous assets. Concerning the empirical strategy, in order to test whether any effect is present, I implement a high-frequency event-study to check whether on the event-window around the announcement days there is evidence of abnormal returns, following the example of Gagnon et al. (2011). There are two main advantages of a high-frequency event-study. Under the assumption of semi-strong market efficiency, the event-study allows to avoid aggregated data that may show no evidence of effects, even if there would have been any, simply by the fact that the market immediately incorporates the new information as they are publicly available. Moreover, a high-frequency event-study allows macroeconomic outlook to change very little except for the announcements themselves, making it easier to identify the eventual causality of any effect. The findings of this event-study highlight that the effects of the two Quantitative Easing programmes similarly reduce interest rates, but with different magnitude, both between each other and between the German corporate bond market and the Italian corporate bond market. The Public Sector Purchase Programme reduced on the overall, the German corporate yields of about 11 basis points, experiencing the largest reduction in investment grade bonds (by 12.5 basis points), while it had its lowest effect on non-financial institution issued bonds (by 5 basis points). On average, across the different categories, the programme seems to have been successful in reducing the interest rates of at least 11 basis points. The effects of the programme on the Italian corporate bonds market are similar, again with a slight difference in magnitude. Italian bonds overall experienced a drop of their interest rates of around 11 basis points, with the largest effect on bonds with very short-maturity (1 to 3 years remaining maturity bonds had a reduction of 12.7 basis points) and the lower effect on longer remained maturities bonds (more than 10 years

6 bonds had a drop of around 6.5 basis points). Concerning the Corporate Sector Purchase Programme, for German bonds, intermediate to longer maturity bonds and investment grade bonds once again are the ones benefitting mostly of the programme (respectively up to 9 and 9.5 bp reduction), and notably (since they were targeted directly) again the lowest reductions is shown in non-financial institutions issued bonds (yield reduced by 2.5 basis points). Concerning Italian bonds, the Corporate Sector Purchase Programme allowed non-financial institutions and short maturity bonds to have the largest decreases in interest rates (respectively of 8.5 and around 7.5). Interestingly, in the Italian case investment grade bonds experienced a lower reduction of their yields compared to the non-investment and not-rated bonds (of roughly 1 bp), even being the only eligible for the programme. These results, are mostly in line with other studies conducted in the literature, differing only in the magnitude of the effects. Most of other studies, as a matter of fact, show, as the present paper, a reduction in interest rates following Quantitative Easing announcements, however the majority of them are based on larger scale purchases compared to the amount purchased by the European Central Bank, so the reduction in yields tend to be generally larger for the other cases analysed. Gagnon et al. (2011), studying the effect of the Large Scale Asset Purchases (LSAP) implemented in US find that the LSAP program successfully reduced bond yields, with a larger effect on longer-term bonds compared to shorter-term bonds. They estimate that the program, which consisted in sovereign bond purchases for a total amount of $1.75 trillion, reduced the 10-year term premium between a range of 30 and 100 basis points, and that the effects spread across bonds that were not directly purchased, as the Baa corporate bond yield index dropped between 60 to 80 basis points. Hamilton and Wu (2012) study the second Quantitative Easing programme in US, finding that $400 billion asset

7 purchases would result in 14 basis point drop in the 10-year Treasury interest rate and 10 basis point reduction in the 6-month Treasury rate. Similarly, an UK-based study conducted by Breedon, et al. (2012) suggest that the Bank or England Quantitative Easing, of a magnitude of roughly 200 billion, caused a reduction of government bond yields (gilts) up to 50 basis points, while Joyce, et al. (2011) suggest that the programme depressed gilt yield up to 100 basis points. While the effects of the Eurozone Quantitative Easing are still dubious concerning their effect on inflation, they seem to ease the funding process for corporate bonds, particularly for high quality bonds and financial institutions related bonds. Concerning their maturities however, it seems to suggest that country specific differences creates a variation that allows virtuous countries, like Germany, to benefit for a longer outlook than countries where structural problems, like Italy, are in place. It must be recognized, however, that because of lack of data for specific categories of bonds, and some arbitrariness in the event-study design, there is the risk of either under or overestimation of the actual effects. This, hence, leaves the possibility of further research in enlarging the dataset, check the robustness of the results with different methodologies compared to the one employed in this case and/or test the methodology with different benchmarks. The rest of the thesis is organized as follows: The first part will be devoted to a literature review summarizing the effects of Quantitative Easing on bond yields. The second part will detail the two Quantitative Easing programmes chosen for the analysis and the expected effects deriving from these policies. The third part concerns the empirical investigation in its methodological approach, it will follow the data description, and finally the results of the analysis. The last part will conclude and address limitations and possible further improvements along the present research.

8 QE EFFECTS ON BOND YIELDS IN THE LITERATURE The effects of Quantitative Easing programmes on bond interest rates have been largely investigated in the recent literature for the US case. Gagnon et al. (2011) study the effect of the Large Scale Asset Purchases (LSAP) undertaken by the Federal Reserve on bonds yields of different maturity and characteristics. The program, that is referred to as QE1, was divided in two tranches, one in 2008 and another one in 2009, and included the purchase of housing agency debt, agency mortgage-backed securities and long term Treasury securities, for a total amount of $1.75 trillion. Using firstly a high-frequency (daily) event-study method and then a lower frequency (monthly) time-series analysis, the authors find that the LSAP program successfully reduced all the categories of bond yields investigated, with a larger effect on sovereign longer-term bonds compared to sovereign shorter-term bonds. They estimate that the QE1 program reduced the 10-year sovereign term premium between a range of 30 and 100 basis points, and the agency MBS of around 100 basis points. Additionally, the authors find that the effects spread across bonds that were not directly purchased, as the Baa corporate bond yield index dropped between 60 to 80 basis points. Supporting the described findings concerning QE1, D Amico, King (2013), using a panel of daily CUSIP-level data on the LSAP purchases and returns, investigate the effect of the variation in the supply of publicly available Treasury securities on sovereign yields with a regression analysis. Their estimate, based on $300 billion of asset purchased in 2009, led to a 4 basis point reduction in the purchased sector on the days the purchases occurred and up to a 50 basis point decline across the yield curve over the entire life of the program. Similarly to Gagnon and his co-authors, D Amico and King find that the effects are larger for bonds with 10 to 15 years of remaining maturity and smaller for bond with much different maturity, suggesting imperfect substitutability across assets.

9 A somewhat different contribution is given by Wright (2011), that in addition to testing whether the first tranches of the QE1 in 2008 had any effect on various sovereign and corporate bond yields, estimates the persistency of these effects, using a VAR approach. His findings suggest that while QE1 helped reducing bond yields of both directly purchased assets and of those with similar characteristics, however the effect for the private sector is smaller than for Treasuries, and these effects faded fast over the subsequent months. A more comprehensive investigation of the potential effects of Quantitative Easing has been done by Krishnamurty, Vissing-Jørgensen (2011). The authors, in addition to test the economic impact of asset purchases on bond yields, try to determine through which channels the effects were operating. The authors identify seven possible channels that may allow Quantitative Easing to affect the bond market : - Inflation channel; since the Quantitative Easing is an expansionary monetary policy, it may affect inflation expectations. If the programme is credible, it may reduce uncertainty about inflation and thus it would reduce the bonds interest rates by reducing the premium linked to inflation volatility. If the programme is not credible, however, may increase the uncertainty about inflation and/or the credibility concerning the efficacy of the monetary policies, in this case, the bond interest rates should increase - Signalling channel; unconventional monetary policies can lower the bond yields if they signal a credible commitment to keep interest rates low for a prolonged period of time. If the central bank weighs potential losses on its balance sheet in its objective function, by raising interest rates and holding long term assets, it would incur in such losses. Therefore, the credibility of the commitment depends on the size of the asset purchased. The signalling channel affects real bonds yields by

10 reducing the expected future policy rates, and the effect is expected to differ across different maturities. Shorter and intermediate maturities should experience a larger drop in yields compared to longer maturity bonds since the likelihood of interest rates increases is larger the longer the period considered - Portfolio rebalancing channel; in Quantitative Easing programmes, central banks purchase particular assets, reducing the private holding of those assets, in exchange for short-term forms of monetary base. A representative investor, as a consequence, experiences now a reduction in the total risk of his/her portfolio. If (s)he does not regard the additional money as a substitute for the assets sold, (s)he is now holding excess money balances. The reduction of portfolio risk makes room for new risk taking and so the excessive money holding will be converted in new risky assets that are a closer substitute of the assets initially sold. The sellers of these new risky assets sold, in turn, will now use the acquired cash to invest in more assets until the share of money relative to all assets is at the general desired equilibrium, clearing the market. At the equilibrium, asset prices are increased and their yields is lowered - Duration risk channel; it is linked to the premium investors require for holding longer maturity assets. By purchasing long term assets, central banks can reduce the duration risk and alter the yield curve, reducing-long maturity bond yields relative to short-maturity yields. With less duration risk to hold in aggregate, the required premium linked to it will be lower. The effect should be produced either because the investors most willing to bear this risk will be those holding the remaining longer maturity assets or if there is more homogeneity across investors in their attitude towards this risk, they should require a lower premium since they will hold smaller amount in their portfolios. The consequence of Quantitative Easing via

11 this channel will be that all long term asset yields should be reduced, particularly, the effect should be larger for assets with longer duration - Safety premium channel; this channel is linked to the preferred habitat theory of investors applied only to safe assets. As Krishnamurty, Vissing-Jorgensen (2012) show, there is evidence of clientele demand for long-term safe assets that has the effect of lower the yield of such assets. The evidence comes from the variation of the spread between Aaa bonds and Baa bonds as a response to the reduced supply of long-term Treasuries. When the supply of the latters is reduced, the availability of long-term safe asset is lower, and the spread between investment and non-investment grade bonds rises. As the authors argue, this premium is lower for very low levels of default risk and for larger supply of safe assets. - Default risk channel; lower grade bonds bear higher default risk compared to higher grade bonds so investors require an additional premium for bearing this risk. If the Central Bank provides stimulus to the economy and the risk aversion of the investors falls, the bond yields are reduced - Liquidity premium channel; investors typically require a premium for holding illiquid assets. In times of financial turmoil, this premium may become a substantial part of the required return to hold an asset, causing a substantial drop in its price. If a central bank steps into the market as a buyer, it makes it easier to sell assets, allowing investors to take larger positions in securities that are part of Quantitative Easing programmes, knowing that if needed they can resell them to the central bank. This increased trading opportunity has as the consequence of lowering the liquidity premium and thus, ceteris paribus, lower the bond yields. Using data from both QE1 and from the so-called QE2, the authors estimate, with a 2SLS regression analysis, that the effects are consistent with the other studies, with different magnitude between QE1 and QE2 and across different maturity and types of bonds.

12 Additionally, they find evidence of the signalling channel on all the bonds, the safety premium channel on Treasuries, the inflation channel on inflation swap rates and TIPS, and default risk channel for corporate bonds. Finally, Hamilton and Wu (2012) provide an econometrical estimate of Treasury supply effects to a term structure model that finds its foundations in the preferred-habitat theory framework presented by Vayanos, Vila (2009). The authors find that $400 billion asset purchases would result in 14 basis point drop in the 10-year Treasury interest rate and 10 basis point reduction in the 6-month Treasury rate. A comparative work between the Federal Reserve and the Bank of England interventions is done by Christensen, Rudebusch (2012). They firstly compare the respective purchases concluding that they are roughly comparable in terms of relative size to the economy and to the stock of outstanding government debt. Then, using a dynamic term structure model, they decompose US and UK sovereign yields into an expectations component and a term premium component. They find that while US yields were responding mainly to variations in expectations, the UK yields were reduced mainly by the term premium component; leading the authors to conclude that the mechanisms at work crucially depend on financial market institutional structures or central banks communication policies. Breedon, et al. (2012) focus their analysis into the impact of the QE, undertaken by the Bank of England between March 2009 to February 2010, to asset prices. The authors generate a macro-finance yield curve to make a counterfactual path the yield curve would have followed without the QE program, in order to estimate the differential effects. In addition, they expand the effects of QE on a range of asset classes. Their finding suggest that the QE, of a magnitude of roughly 200 billion, caused a reduction of government bond yields (gilts) up to 50 basis points due to the portfolio rebalancing channel, while the pass through to other assets seemed limited.

13 While supporting the findings of Breedon, et al. (2012) concerning the transmission channels of the effects of the QE, Joyce, et al. (2011), however, implementing an eventstudy analysis, find that the program depressed gilt yields by up to 100 basis points. They further analyse the effects on other asset classes, finding that the effects were more meaningful for equities rather than for bonds or for the sterling. While Quantitative Easing programmes are a relatively recent monetary policy response for the US and UK case, this is not the case of Japan. The first Quantitative Easing program undertaken by the Bank of Japan was implemented in the year 2001 as a response to deflationary spirals hitting the economy. Ugai (2006) summarizes in a comprehensive survey the findings related to the effects of QE from 2001 to 2006 on bond yields, clustering them according to the transmission channel through which they operate. Baba, et.al (2005), and Oda, Ueda (2005) find that the signalling channel worked in lowering the yield curve by as much as 40 to 50 basis points for 3-year and 5-year bond yields. The portfolio rebalancing channel seemed to operate for Japanese government bonds in reducing yields in Takeda, et al. (2005), while in Kimura, Small (2006) meaningful reduction in yields were found in high grade corporate bonds and interestingly, effects on yields were detected also in stocks and low grade corporate bonds but increasing them, suggesting that the safety premium channel might have been at work in this case. A more recent investigation about Japanese QE is Lam (2011). The author studies the new asset purchase program introduced in early 2009, that included direct purchase of government and corporate bonds, with an event-study analysis based on asset returns and volatility. He finds that the program reduced 10-year government bond yields by 24 basis points and by 14 basis points the yields for the 2-year bond. The effect on investment grade corporate bonds was a reduction of 15 to 22 basis points.

14 In light of the discussed literature, the present paper steps in applying some of the methodologies employed to investigate the relationship between Quantitative Easing and corporate bond yields in the context of a European case, adding further detail in the analysis by testing the effects, if any, on each specific category of bonds. EUROPEAN CENTRAL BANK QUANTITATIVE EASING PROGRAMMES PUBLIC SECTOR PURCHASE PROGRAMME (PSPP) On January 22, 2015, the European Central Bank announced the public sector purchase programme (PSPP) as an expansion of the already existing asset purchase programme. The main characteristics of the PSPP consists in: - Expansion of asset class purchased, with the inclusion of bond issued by Eurozone central governments, agencies and European institutions - Combined monthly asset purchases to amount to 60 billions - The programme is intended to be carried out at least until September 2016 The decision taken by the Governing Council of the European Central Bank was aimed at fulfilling the mandate of price stability, for reaching the goal of a medium term inflation rate, measured by the HICP, of below but close to 2%. The programme, as a matter of fact, has been a response to actual and expected inflation rates at their historical lows, and encompasses the private sector asset purchases launched in 2014, which included the Asset-backed securities (ABS) purchase programme and the covered bond purchase programme (for operational and technical details of the programme, see Appendix A). The programme does not target specific durations of assets, the purchases are weighted by nominal outstanding amounts, with eligible remaining maturities at the time of purchase,

15 ranging from 2 to 30 years. The goal set by the European Central Bank is to be marketneutral also concerning the weighting the different maturity buckets for purchases, with the specification of some flexibility for taking into account the relative values of bonds and the liquidity of the different maturity segments. Furthermore, in principle, purchases of nominal marketable debt instruments at a negative yield to maturity are permissible as long as the yield is above the deposit facility rate. The asset purchases are conducted in the secondary market against central bank money, and are calculated at book value and do not include amortization, which may decrease or increase the value of holdings over time. The amortizations are calculated at quarterly basis and require the asset value to be revalued upward over time towards maturity for bonds whose price is below the face value, and downwards towards maturity for bonds whose price is above the face value. The Governing Council is responsible for the programme design features, making use of decentralised implementation to mobilise the resources. The share of purchases for National Central Banks in domestic markets is determined by ECB s capital keys, however, since Greece is not included in the programme, that implies that the Bundesbank has a share of purchases of 26,3% of the total purchases, while Banca d Italia retains a share of 18% of the total. Within these shares, the National Central Banks are allowed to choose between the purchase of central government securities and securities of certain agencies established in the respective jurisdiction. However, the European Central Bank has defined issue share limits, and issuer share limits. The issue share limits, that is, the maximum share of a single PSPP-eligible security that the Eurosystem will hold, is set at 25%, to be reviewed after six months; while the issuer share limits, that is, the maximum share of an issuer s outstanding securities that the Eurosystem will hold, is set at 33%. In both cases, the limits are set at nominal values and not market values.

16 Concerning hypothetical losses, it was deliberated that purchases of securities of European institutions (which will be 12% of the additional asset purchases, and which will be purchased by NCBs) will be subject to loss sharing. The rest of the NCBs additional asset purchases will not be subject to loss sharing. The ECB will hold 8% of the additional asset purchases. This implies that 20% of the additional asset purchases will be subject to a regime of risk sharing. 1 The ex-ante expected effects of the Public Sector Purchase Programme depends on the categories of bonds and by what channel (or channels ) may be working. The table below summarizes the possible scenarios. Channel Inflation Signalling Expected interest rate change / If the PSPP is considered to be credible in reaching its goal of support to the achievement of the 2% inflation rate mandate which would imply an increases of inflation the change should cause an increase in yields due to a higher premium for inflation If the PSPP reduces the uncertainty about inflation expectation, however, it should reduce the interest rate because of a lower compensation for the uncertainty. The net effect is thus uncertain exante If the PSPP is seen as a commitment to keep interest rates low for a prolonged period, one should expect lower interest rates, especially for shorter and to some extent to intermediate maturity bonds, while the effect should be lower or absent in bonds with long term maturities, since the longer the period, the more likely the policy interest rates will be reversed By rebalancing their portfolios, to substitute the asset purchased by the 1

17 Portfolio Rebalancing Duration Risk Safety Premium Default Risk Liquidity Premium # Central Bank, through the purchase of assets with similar characteristics of those just sold, the investors cause the yields of these other assets to drop, due to an increased demand which raises the prices If the target of purchased assets of the PSPP are long term bonds, this would cause the yield curve to flatten and it would reduce the duration risk. According to this channel, one should expect a reduction in interest rate, with larger magnitude for longer maturities. If the PSPP reduces the availability of safe assets in the market, investors that follow a preferred-habitat style of investing, will therefore shift part of their investments to private sector safe assets. If this channel is in place, one would expect therefore to see an increasing spread between investment and non-investment grade bond yields By purchasing public-sector assets, the European Central Bank reduces the sovereign component of the default risk premium paid by any bond, therefore one would expect to see a reduction in interest rates. Moreover, since the sovereign risk is larger for Italy than for Germany, one should expect to see a larger reduction in Italian bond compared to German bond yields Since the PSPP do not involve the purchase of any private-sector asset, the liquidity of the latter is not influenced, thus one should not expect any variation linked to this channel CORPORATE SECTOR PURCHASE PROGRAMME (CSPP) On March 10,2016, the European Central Bank announced a new quantitative easing programme, called Corporate Sector Purchase Programme (CSPP), which is added to the

18 existing Asset Purchase Programme. The main characteristics of the new programme consist in: - The Corporate Sector Purchase Programme is added to the existing Asset Purchase Programme and it is included in the combined monthly purchases - Combined monthly purchases with the previous Asset Purchase Programme increase from 60 billion to 80 billion - Investment-grade euro-denominated bonds issued by non-bank corporations established in the euro area will be included in the list of assets eligible for regular purchases According to the Governing Council of the European Central bank, the Corporate Sector Purchase Programme goal is to strengthen the pass-through of Eurosystem s asset purchases to the financing conditions of the real economy 2, and contributing to return the inflation rate to levels below, but close to, 2% in the medium term. The criteria for eligibility of bonds to the programme were not defined as of 10 March 2016, while the only specification provided was that securities issued by credit institutions and by entities with a parent company which belongs to a banking group will not be eligible. 3 Similarly, it was established to start the purchase at the end of the second quarter of 2016, postponing the official date to subsequent new announcements. On April 21, 2016, a new announcement was made, providing the technical details of the CSPP. The main information regarded: - The starting date of the programme, to be June The implementation, to be carried out by six national central banks on the behalf of the Eurosystem, coordinated by the ECB. The six national central banks are: Nationale Bank van België / Banque Nationale de Belgique, Deutsche Bundesbank, Banco de España, Banque de France, Banca d Italia, and Suomen Pankki/Finlands Ibid.

19 Bank. Each national central bank is responsible for purchases from issuers in a particular part of the euro area More information concerning the eligibility include that in order to qualify for purchase under the CSPP, debt instruments must have a minimum remaining maturity of six months and a maximum remaining maturity of less than 31 years, at the time of the purchase. It is not required any minimum issuance volume, allowing every corporate bond that falls into the eligibility requirements to be purchased. Purchases of eligible debt instruments with a negative yield to maturity are permissible as long as the yield to maturity is above the deposit facility rate at the time of purchase. The Eurosystem applies a maximum issue share limit of 70% per international security identification number (ISIN) on the basis of the outstanding amount, it also applies additional limits per issuer group, following a pre-defined benchmark. Concerning further information about issuers, the issuer of the debt instrument may not have a parent company which is subject to banking supervision outside the euro area; may not be a supervised entity of the European Central Bank or a member of a supervised group; may not be an investment firm (firms which are classified as MiFID investment firms from a legal perspective, are not eligible). It is added that it is not possible for debt nominated in euro but issued by non-euro area residents to be eligible for the programme. The purchases are conducted in both primary and secondary market but no primary market purchases has to involve debt issued by entities qualified as public undertakings. The purchases may include forms of international debt securities depending on whether the securities are issued in global bearer or global registered form. As for international debt securities issued in global registered form after 30 September 2010, they must be issued under the new safekeeping structure for international debt instruments. International debt securities issued in global bearer form after 31 December

20 2006 must be issued in the form of new global notes and are to be deposited with a common safekeeper which has been positively assessed in accordance with the Eurosystem User Assessment Framework. International debt instruments in individual note form issued after 30 September 2010 are not eligible. International debt securities have to fulfil these requirements regardless of whether they are bought in the primary or in the secondary market. Finally, the European Central Banks specifies that in the case of downgrading of an already purchased security, it is not required to sell the holdings of those securities; while principal payments on the securities purchased under the CSPP will be reinvested as they mature, for as long as necessary. (Further technical details of the programme are to be found in Appendix B) A similar analysis as the one for the PSPP concerning the ex-ante expected effects of the Quantitative Easing can be done for the Corporate Sector Purchase Programme. Channel Inflation Signalling Expected interest rate change / If the CSPP is considered to be credible in reaching its goal of support to the achievement of the 2% inflation rate mandate which would imply that expected inflation increases the change should cause an increase in yields due to a higher premium for inflation If the CSPP reduces the uncertainty about inflation expectation, it should reduce the interest rate because of a lower compensation for the uncertainty Once again, the net effect is uncertain If the CSPP is seen as a commitment to keep interest rates low for a prolonged period, one should expect lower interest rates, especially for shorter and to some extent to intermediate maturity bonds, while the effect should be lower or absent in

21 Portfolio Rebalancing Duration Risk Safety Premium Default Risk bonds with long term maturities, since the longer the period, the more likely the policy interest rates will be reversed By rebalancing their portfolios, to substitute the asset purchased by the Central Bank, through the purchase of assets with similar characteristics of those just sold, the investors cause the yields of these other assets to drop, due to an increased their demand which raises the prices. Compared to the PSPP, the CSPP can be expected to have a stronger portfolio rebalancing effect, since the substitutability of the asset purchased in this programme is larger than for the PSPP, which implied substituting public-sector with private-sector issued assets If investors expects the Central Bank to purchase a sufficient amount of medium to long-term assets, these would reduce the required duration risk premium to hold these assets If the CSPP reduces the availability of safe assets in the market, investors that follow a preferred-habitat style of investing, will therefore shift part of their investments to other types of private sector safe assets. If this channel is in place, one would expect therefore to see an increasing spread between investment and noninvestment grade bond yields. Moreover, since the programme involves direct purchase of corporate bonds, the effect can be expected to be larger, if any, compared to the correspondent PSPP one. By purchasing corporate-sector assets, the European Central Bank reduces the idiosyncratic component of the default risk premium paid by the purchased bond, therefore one would expect to see a reduction in interest rates. Since the CSPP involves only investment grade, nonfinancial institution issued bonds, one would expect this channel to operate, if so, only for these categories of bonds By purchasing corporate bonds, the European Central Bank will provide liquidity to the bond market in all its

22 Liquidity Premium segments, therefore one would expect to see a general reduction in bond yields due to a lower liquidity premium EMPIRICAL PART METHODOLOGY The investigation of the effects of Quantitative Easing on Eurozone corporate bond yields is undertaken on German and Italian assets, with a high-frequency event-study. The choice of the high-frequency event-study is based on the fact that such technique measures changes in asset prices (or returns) in a narrow window of time around events in financial markets. The preference of a high-frequency event-study on a low-frequency event-study is given by the fact that bonds may be sensible to various macroeconomic variables and since these latter can vary dramatically quarterly or even weekly, allowing for a more diluted frequency may cause multiple macro events to affect their yields, making more complicated to disentangle the effects of Quantitative Easing alone. The first step in implementing the event-study is identifying the events of interest and the timing of the events. Concerning the events of interest - the Public Sector Purchase Programme and the Corporate Sector Purchase Programme - the choice fell on the announcement of such programmes rather than their actual implementation. Under the hypothesis of semi-strong market efficiency and rational expectations, around a major event date bonds should react to incorporate shortly the new publicly available information such that the actual event occurring the market will not show further significant responses, if any.

23 Concerning the timing of the events, an estimation window and an event window need to be defined. The estimation window is used to calculate the normal returns and their standard deviation that will be used as benchmark for calculating the presence of abnormal returns at the event window. Following MacKinlay(1990), the estimation window employed consists of 120 days prior the event window. The event window contains n-days around the event date, before it in order to capture eventual effects if the news becomes unofficially available in advance, and after it, in order to allow the market to fully incorporate the new information in the asset pricing. The choice for the length of the event window is of 2-day around the events day. The rationale behind it relies on one hand on the assumption of semi-strong market efficiency, on the other hand on allowing macroeconomic outlook to change very little except for the announcements themselves. The second step for the implementation of the event-study is to specify the benchmark model to describe the normal bond returns behaviour, which is, the returns onewould expect in normal circumstances without an event. While the majority of empirical studies propose a mean-adjusted return method, which simply averages the returns observed out of the number of days that constitutes the estimation window, in this event-study I will determine normal returns as derived by the CAPM. Defining the normal returns as NR, these are calculated as where is the risk free rate, estimated as the returns of the 10 years German government bond, which is the ratio between the covariance of bond returns and the market returns, divided by the variance of the market returns, estimated over the estimation period of 120 days, and constitutes the market returns, which will be proxied by the Barclays Euro

24 Government Float Adjusted Bond Index, that is constituted by the Euro area government bonds and thus provide an approximation of the Eurozone bond market. While other methodologies, like the mean-adjusted method, may suggest a closer link with realized returns, the advantage of the CAPM is that it also take into account that the pricing of the bonds over the estimation window may not be fully efficient, thus calculating the normal returns with this methodology may allow to capture as well whether the Quantitative Easing programmes are affecting efficient pricing of the bonds, and not just the realized pricing. In addition, since the goal is to investigate the effect of Quantitative Easing on the corporate bond market as a whole, I apply the Capital Asset Pricing Model method to homogenous classifications of bonds, defined by : rating, remained maturity, type of issuer. Concerning ratings, two buckets are made: one including all the investment grade bonds (defined as up to BBB- for S&P and Fitch or Baa3 for Moody s) one including all the non-investment grade bonds (below the thresholds defined above) and the non-rated bonds The latter might be of interest since the included non-rated bonds may represent an intermediate category between the rated bonds. As a matter of fact, multiple reasons may explain why some bonds are not rated. Some, for example, may find useful not rate their bonds if they are aware of likely poor grading, so in order not to give that signal to the market and having an higher funding, they may prefer to exploit uncertainty if they believe that the latter can provide a lower cost of funding. A complete opposite situation, however, might be in place. Very small issuers may find too costly to rate their bonds even though they might rate as investment grade, so they end up not having a rating. Generally, to this

25 category belongs especially small issuers, typically small financial institutions and the majority of their financiers are usually non-institutional investors. Concerning the maturity of bonds, the distinction is based not on maturity as defined by the difference between expiration date and the issuing date, but on remaining maturity, calculated as the difference between the event day and the issuing date. The rationale behind this choice is due to the consideration that an investor would not treat a 30-year bond as a long term asset at one month from the expiration date. The remaining maturity is divided into maturity buckets. These are therefore: one for short-term investing, defined as a remained maturity of less than three years one for medium-term investing, defined as a remained maturity of less than five years and more or equal to three years one for long-term investing, defined as a remained maturity of less than ten years and more or equal to 5 years one for very long-term investing, defines as a remained maturity of more or equal to 10 years Since the bond market segment investigated concerns corporate bonds and not include sovereign bonds, the choice for the bucket of short-term investing is not based on a remaining maturity of less than one year but it is assumed that the category is still homogenous to anything below three years. Moreover, since long run effects will be larger the longer the remained maturity, the long-term investing is divided into two different subcategories, as it is assumed that a change in yield for a 5-year maturity bond will not be homogenous with a change in the yield of a 30-year maturity bond. Concerning the type of issuer, there are two buckets: one for bonds issued by financial institutions one for bonds issued by non-financial institutions

26 The separation by type of issuer is based on the idea of comparing the pass-through from the financial sector to the real economy, which gives an indication about eventual presence of frictions in a country s market, and thus, its efficiency. Each return of the bond is calculated as the variation of yield to maturity, that is, the interest rate that equalize the price of the bond with its future cash flows. The final step of the event-study is to compute the abnormal returns over the eventwindow and testing their significance. The abnormal returns, as previously discussed, are simply defined as the difference between realized returns and normal returns, that is Since the event window chosen for the event study consists of multiple days, before testing the statistical significance of the abnormal returns, I first compute the cumulative abnormal returns as the sum of the abnormal returns over the event-window, formally and then, since there are n different bonds analysed, I compute the cumulative average abnormal returns, defined as the average of the cumulative abnormal returns, that is Finally, I test the significance of the cumulative average abnormal returns. The null hypothesis is that there are not abnormal returns and this will be two-sided tested. Since the population variance of the returns is unknown, I will test through a sample variance which needs to be estimated. In order to determine the proper test, several considerations first needs to be done, based on the fact that the standard hypothesis

27 testing is done by assuming that the variable tested, in this case the abnormal returns, is independent, identically and normally distributed, such that the hypothesis testing will follow a standard normal distribution with average 0 and variance equal to 1. In order to account for the distribution of abnormal returns over the event-window, possible event-induced variance, serial correlation and cross-sectional correlation, the cumulative abnormal returns are adjusted in order to proxy the standard normal distribution (for the detail of the adjustments, see Appendix C). DATA The data employed are retrieved from Thomson Reuter Datastream, and consist of all the available German and Italian corporate bonds time series price data outstanding at the date of 17/06/2016. Specifically, for Germany, the data consist of 7039 bonds, 6921 of which belong to financial institutions and 118 to non-financial institutions. Similarly, for Italy, the data are composed of 1773 bonds, 1733 of which were issued by financial institutions and 40 of them by non-financial institutions. Since not all the bonds outstanding at the date of 17/06/2016 were already issued at the time of the Quantitative Easing implementations, the total amount of data differs between the first and the second Quantitative Easing analysed. While almost all the category buckets used in the analysis contain enough data to approximate the Central Limit theorem (which is usually set at 30 minimum observations), the number of bonds for Italian non-financial institutions bonds for the Public Sector Purchase Programme reaches only 23 observations, not allowing a proper statistical interpretation of the results. Similarly, the comparison between categories may be impaired in some cases by the large difference of observations.

28 Concerning Germany, in addition to the disproportion between financial versus nonfinancial institutions, it is important to note that the maturity bucket for over 10 years remaining maturity is significantly lower compared to the other buckets (250 observations compared to at least 1400 observations for all the other categories). Italian bonds show the same limitations of German bonds, and additionally only around a 100 bonds are rated as investment grade out of the total number of bonds. These limitations therefore suggest caution in the interpretation of the results, not just in terms of statistical significance but potentially even in their overall effects. RESULTS The table below displays the effects of the two Quantitative Easing programmes, on German and Italian corporate bond interest rates, showing that these seem to have been generally beneficial in reducing the funding costs for corporates, with changes of rates with direction and magnitude comparable to the ones of similar programmes conducted by other Central Banks.

29

30 PUBLIC SECTOR PURCHASE PROGRAMME Chart 1 Public Sector Purchase Programme Overall -1,0-3,0-5,0-7,0 Germany Italy -9,0-11,0-13,0-11,3-10,9 The findings for the Public Sector Purchase Programme (PSPP), which involved the purchasing of sovereign bonds by the European Central Bank, suggests that effects of the programme spread beyond the securities targeted, affecting the corporate sector as well, as displayed in chart 1. The chart shows the cumulative interest rate variation over the 2- days event window on the overall sample of German and Italian corporate bonds. As it can be seen, the PSPP helped reducing both German and Italian corporate bonds interest rates, with a similar magnitude of around 11 basis points, where the slight difference is in favour of German bonds. This difference, however, is unlikely to signal any specific difference between the two countries because of its very small magnitude, and the fact that German available bonds analysed were roughly three and half times more than the Italian bonds available. However, while the overall effects of the programme seems to be of equal scale between German and Italian bonds, dissecting them into more homogeneous categories, their magnitude varies.