The effects of the ECB's public and corporate bond purchases. An event study examining the German financial market in comparison to the G7-group

Table of Contents

Introduction

Literature review & background
Channels through which QE influences prices
Study layouts used in previous work
Findings of previous research
Spill-over effect on other assets
Positioning of this dissertation within the existing work
Quantitative Easing of the European Central Bank

Methodology

Data

Results
Government bonds
Corporate bonds
Currencies
Equity markets

Conclusion

References

Appendix

Abstract

This study quantifies the financial market impact of the ECB’s public and corporate bond purchase programs from January 2015 to June 2016. The conducted event study analyses eleven announcement events and their influence on government bonds, corporate bonds, exchange rates and equity markets. The focus lies on the German market, although all other G7 countries are included for reasons of comparison. These are the main findings: Government bond and corporate bond yields fell by 23bp and 32bp respectively over all events; the effect on the euro was negative but weak; and the euro equity markets were positively driven by 680bp. Furthermore, most of the effects occur on the day following the events. Finally, the news impact diminishes over time, causing an observable break between the public and corporate bond purchase program.

Introduction

Over the course of the last few years, mainly caused by the aftermath of the financial crisis from 2008/2009, the most important central banks in the world have followed an accommodative monetary policy. In a narrow sense, this initially included interest rate declines which mitigated the burden of heavily indebted governments and simplified financing for companies as well as consumers.

The lower the interest rates are, however, the less economic stimulus can be achieved by this approach, especially because zero interest is commonly seen as an important level not to cross. This raised the necessity of further measures. Among the different so-called unconventional methods, large-scale asset purchases (LSAP) have been applied by all major central banks. Through such programs, central banks intervene in the bond markets, causing an artificial demand - not only lowering the yields and consequently also the cost of capital for companies but also increasing the money supply. The main purpose behind this is reaching the usual inflation target of close to but below two percent. Such measures are often referred to as quantitative easing (QE).

Scientific research has intensively studied the consequences of such LSAP programs on the markets. The vast majority of research focussed on either the Fed policy or the Bank of England. Although the idea of quantitative easing was not new at this time, the United States of America were the first country which set up a QE program over the course of the last financial crisis. The United Kingdom and Japan followed. The European Central Bank (ECB) was later, therefore the euro zone is still a bit earlier in the interest cycle than the United States. The ECB started relieving the banks’ balance sheets at first, with the intention to stabilise at the point proven to be crucial in the financial crisis. That crucial point was when the covered bond purchases and asset-backed security purchases began in summer 2009. In spring 2015 the ECB opened a new chapter with the purchases of government and later also corporate bonds.

The record-low interest rates today and the continually record-breaking highs in stock indices in the last few years are undoubtedly not driven entirely by developments of the real economy but primarily by the artificial market distortion of the ECB’s interventions. To further concretise this relation, the goal of this study is the quantification of the financial market impact. Recognising the fact that the Western capital markets are today seen as relatively efficient, asset prices always reflect the available information. Derived from this thought, this study examines the asset price movements at the time of the news releases. There are undoubtedly longer term impacts that cannot be captured by this approach, but the idea is that initial responses are nonetheless very representative. The potential findings have undeniable practical implications for the investment industry. The financial markets and investors are enrapt by the ECB, hanging on every word leaving Mario Draghi’s lips - because the smallest nuances could move the markets.

The dissertation is organised as follows: After an overview of the existing literature on this topic and the ECB measures, I elucidate the event study methodology and my data set, before discussing the study results and finally concluding my findings.

Literature review & background

The existing literature on the effects of quantitative easing addresses a number of different programs, though the Fed’s QE program in the years 2008-2014 and the QE of the Bank of England from March 2009 are the two best researched programs. The financial crisis led to quantitative easing and the sudden appliance of multiple QE programs led to a jump in the academic interest in this field. Hence why most of the literature is current and up to date. Nonetheless the first appearance of QE was back in 1961 with Operation Twist in the United States. The Fed was not expanding its balance sheet at this time but swapping shorter bonds into longer ones. In more recent history, the Bank of Japan applied QE at the beginning of the millennium, attempting to escape many years of stagnation. Part of this program was the purchase of four hundred to one thousand two hundred billion. Yen of long-term Japanese government bonds. Finally, the Bank of Japan managed in 2005 to turn the consumer price index back into positive territory (Ugai, 2006).

Quantitative easing follows three main objectives: First, the central bank wants to commit to an environment of low interest rates in the mid-term which shall enhance companies’ investments and consumer spending. Second, the accompanying effect of deleveraging banks’ balance sheets is supposed to allow them to grant more credits. Third, higher asset prices on the financial markets boost the net worth of consumers and shall be indirectly expressed in higher consumption (Steeley, 2015). Furthermore, a potential weakening of the national currency boosts the export level of the economy, especially for economies with as high a current account surplus as Germany. A weak currency also has the tendency to further support the stock market because the market appears cheaper to foreign currency investors in this case.

Upon reviewing at the existing literature, it makes sense to structure the already conducted research precisely in order to acknowledge all different aspects.

Channels through which QE influences prices

So first of all, one has to work out how exactly quantitative easing influences the asset prices. The existing research has identified a number of different so-called transmission channels to classify the effects. Although some authors like Krishnamurthy & Vissing-Jorgensen (2011) decided to break down the effects in many different categories (including a safety channel, inflation channel and MBS risk-premium channel), the majority of research works only with three main channels. Those are the liquidity effect, the signalling effect and the portfolio balance effect (PB).

The first effect is the liquidity effect. The central bank conducting QE buys bonds on the market and puts them on its balance sheet. The bond selling investors are consequently holding more liquidity afterwards and the liquidity premium falls, at least on liquid bonds. The prices of bonds like treasuries partly consist of a liquidity premium, accounting for their high liquidity. This is especially true in times of market distress. When the overall liquidity in the market increases due to asset purchases, the liquidity premium on these bonds falls and this channel consequently implies rising yields (Krishnamurthy & Vissing-Jorgensen, 2011). So, whereas the signalling effect is supposed to lower yields at the announcement, the liquidity effect is said to re-increase them at the time of the actual purchase.

The signalling effect describes the commitment of the central bank to ongoing low interest rates. It is also in the self-interest of the central banks to keep interest low after purchasing assets because they would otherwise suffer from large losses on their balance sheet positions and the state might consequently be obliged to recapitalize the central bank - a prospect which should be avoided by any means (Bauer & Rudebusch, 2014). The sentiment of corporates and consumers will also be invigorated because they are motivated to spend more when the risk of deflation diminishes (Hendrickson & Beckworth, 2013). The whole signalling aspect can also be described as the creation of confidence (Lim, et al., 2014) and a good measure of it can, for example, be the market’s implied volatility.

The portfolio balance effect is caused by the fact that in order to motivate investors to sell their holdings in a bond, the expected return for the future must be lower. The central bank is causing demand, the price rises and the yield falls. When different asset classes are relatively comparable regarding a potential substitution, the effects on bonds can spill over to other asset classes. The best substitute are other bond segments, so mainly corporate bonds. Another aspect is therefore the increased risk appetite (the yields for the same level of risk are now lower) which causes a shift towards longer durations and lower credit quality. Most research has in common that it attributes the largest part of the price effects to the portfolio balance effect (Gagnon et al., 2011; Hamilton & Wu, 2012).

The liquidity effect does not materialise before the actual purchase actions and signalling no longer plays an important role because when the purchases finally take place, they were widely anticipated. On the other hand, portfolio balanced effects need some time in order to feed through to other market segments. This leads to the conclusion that an event study layout analysing announcement and not purchase dates must necessarily capture mostly signalling effects (see also Breedon,et al., 2012).

Study layouts used in previous work

Most researchers conducted event studies of either the announcements or the actual purchases (e.g. Gagnon, et al., 2011; Meaning & Zhu, 2011; Joyce, et al., 2011). In their event study, Joyce et al. (2011) summed up the changes and compared them to a control group consisting of other G8-economies. They also tried to adjust the impact of different announcements depending on their amount of new information/ surprise factor (they used Reuters polls of expected purchase amounts). This approach allowed them to measure the yield change per billion pounds of unexpected purchase volume. An OLS regression leads to the result of 0.6 basis points per unanticipated billion.

Breedon et al. (2012) applied a combined approach. They estimated the portfolio balance effect with a term structure model with macroeconomic variables as input, and finally compared the constructed term structure to the real one. They assign the delta to the portfolio balance effect. Bernanke et al. (2004) pursue a comparable approach. In order to measure the liquidity effect, they also conducted an event study. Unlike many other researchers, they do not focus on the signalling effect of the announcements but instead on the liquidity effect of the bond purchase operations. That is an important difference in the event day definition.

The advantage of term structure models might be seen in the fact that they account for longer price changes and not just small event windows and announcement effects. Breedon et al. (2012) chose an interesting approach insofar as they separated the liquidity and the PB effect in a very clear war. Regarding the type of data which previous researchers looked at, there is a broad variety given. Joyce et al (2011) analysed gilt yields and OIS-spreads for the bond market. They furthermore looked at liquidity, represented by market turnover and bid-ask-spreads. Krishnamurthy & Vissing-Jorgensen (2011) also looked at treasury yields, agency yields and MBS yields as well as credit default swaps.

In order to see the results in comparison, a control group is needed. Breedon et al. (2012) compared against the hypothesis of an average zero price change under normal conditions and against an interpolated price movement based on their yield curve model from their portfolio balance analysis. What is surprising is that they never recognize the signalling aspect. As already said, Joyce et al. (2011) compared to other G8 group members which are unlikely to be influenced by overseas events.

Findings of previous research

Let us take a closer look at the actual findings: Joyce et al. (2011) found that across all maturities, the average yield decline was around one hundred basis points. The liquidity improved considerably, the market turnover slightly increased whereas the average bid-ask- spread narrowed significantly from 1.4 to 0.9 bp across all maturities. Using a comparable study layout, Meaning & Zhu (2011) measured a cumulated yield decline for ten-year government bonds of -80 bp in the US and -50 bp in the UK. Staying in the same country, Breedon et al. (2012) find that the bond prices of UK gilts are also being influenced by the actual purchases. They already begin rising two weeks before the purchase, overall by approx. 0.6% but decline to an even lower level in the twenty days after the purchase operation which can be explained by the liquidity effect. They also explain this behaviour by the inertia of the government bond market. Analysing QE1 of the Fed, also Krishnamurthy & Vissing-Jorgensen (2011) found a significant but diminishing yield decline of a cumulated -74 bp for ten-year US-treasuries. Interestingly, the maturity reactions were inconsistent in this case with thirty-year bonds reacting to a lesser extent than shorter treasuries. And finally D’Amico and King (2010) measured an overall yield decline of roughly -50 bp in the US. Considering all the work, the true value is likely to be somewhere between fifty and one hundred basis points.

An important side aspect is whether bond purchases are offset by new bond emissions. Some papers account for this difference, though this seems to be rather secondary overall.

Spill-over effect on other assets

The impact on other assets than government bonds comes with a delayed effect. Currency prices and corporate bonds are the closest in reaction (see also Joyce et al., 2011), stocks might follow later due to time-consuming portfolio balancing.

LSAP have an impact on other assets than government bonds through mainly two channels. The first is the portfolio rebalancing/substitution of the sold bonds because holding cash is not a good substitute. The second are the implications of such a program and what they suggest about the state of the economy (e.g. employment or corporate earnings). So it implies a macroeconomic outlook too.

The most obvious idea is to look upon government bonds which were not included in the purchases, as Breedon et al. (2012) did, for example. They found spill-over effects to unbought maturities of government bonds, different rating classes and swaps, so basically the overall government bond market. That is very unsurprising regarding the fact that within this segment, these bonds are almost perfect substitutes for each other.

The closest asset class to government bonds are obviously corporate bonds. Joyce et al. (2011) found that corporate bond yields fell, as expected, around 70 bp over the course of all six events, non-investment grade yields fell by up to even 150 bp and also their spreads narrowed tremendously (75 bp). Meaning & Zhu (2011) measured a very comparable -63 bp for US corporate bonds and -52 bp for the UK.

An interesting aspect mentioned by Krishnamurthy & Vissing-Jorgensen (2011) is the safety aspect. They say that treasuries and corporate bonds have different price drivers. Investors pay a safety premium on the former and when the central bank is reducing the supply of those safe bonds, the safety premium rises. This effect is not however transferrable to the lower-rated corporate segment because the safety is not given there.

Regarding the effect on different corporate bonds maturities, Krishnamurthy & Vissing- Jorgensen (2011) discovered that longer termed bonds seem to be preferred by investors as substitutes, so their yield should fall the most. The reason behind this phenomenon is the fact that investors who want to maintain their current yield have to choose longer and therefore usually (in case of an increasing term structure) higher yielding papers.

[...]