**Conventional and Unconventional: The Effect of the ** **ECB’s Monetary Policy Changes on Stock Market **

**Returns **

University of Amsterdam, Amsterdam Business School MSc Finance, Track: Quantitative Finance

Master Thesis

Student Name: Eline Heezen Student Number: 11061065 Thesis Supervisor: Mr T. Polat

July 2021

**Statement of Originality**

This document is written by Student Eline Heezen 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.

**Abstract **

In 2007 the ECB introduced an unconventional monetary policy for the first time.

Since then, the effect of this new type of policy has kept the financial world busy.

This paper examines the effect of all the (un)anticipated (un)conventional monetary policy changes of the European Central Bank (ECB) on market stock returns over the period May 2001 – May 2021. Conventional monetary policy changes are measured using data of futures markets and the ECB rates, and unconventional monetary policy changes are measured by using a spread in government bonds. The conducted event study method finds that the impact of unconventional monetary policy changes is significant and negative on the EURO STOXX 50 index, which is in line with the expectation. The presence of a crisis is also taken into account, as it is likely to interfere with investors’ expectations and actions. Nonetheless, the measures taken as a result of occurring crises did not lead to significant results, suggesting the presence of a crisis does not affect stock market returns with regard to monetary policies.

**Table of Contents **

**1. Introduction ... 5**

**2. Literature Review ... 9**

**3. Methodology ...13**

**4. Data and descriptive statistics ...19**

**5. Estimation Results ...25**

**6. Robustness Checks ...32**

**7. Conclusion ...39**

**References ...42**

**Appendix ...44**

**1. Introduction **

The first unconventional monetary policy was announced on 22-08-2007; it entailed the three-month Supplementary Long-Term Refinancing Operation, which was allotted one day later. Unconventional monetary polices were introduced to re- establish confidence in the European Central Bank (ECB) and restore financial stability (Kenourgios & Ntaikou, 2019). Many more unconventional monetary policies followed.

The ECB distinguishes itself from other central banks on the grounds that they introduced unconventional monetary polices at the very beginning of the crisis as well as using unconventional and conventional monetary policies at the same time for some periods (de Haan et al., 2020). While the ECB used both types of policies simultaneously at times, it is important to note the difference.

Joyce et al. (2012) state the difference between conventional and

unconventional monetary policies as follows: the main instrument of conventional monetary policy is setting a target short-term interest rate, and, when needed, the ECB provides extra short-term liquidity. The main goal of the short-term interest rate as an instrument is to drive the long-term rates in a comparable way. After the

ineffectiveness of short-term interest rate as an instrument was proven in the event that it reaches the zero lower bound, unconventional monetary policy was introduced.

This entails large amounts of asset purchases, with the prospect of this influencing more than just the short-term interest rate (Joyce et al., 2012).

Not only the distinction between conventional and unconventional is of importance, so is that of anticipated and unanticipated in monetary policies. Bernanke and Kuttner (2005) have proven this differentiation to be crucial. The basis for this is the efficient market hypothesis by Malkiel and Fama (1970), which states that solely interest rate surprises should affect stock prices and thus stock prices should not be affected by anticipated changes. Once more, signals would prevent anticipated changes to have an effect, as this signal would already be taken into account before the change actually occurs. Fausch and Sigonius (2018) conclude that if the

distinction is not made, this may lead to bias results.

When linking these different policies to the stock prices the following can be concluded. Based on previous papers, Wang and Mayes (2012) state that when there

is a positive surprise, normally this will provoke markets to believe there is more information at the central bank’s disposal than anticipated. Due to this adverse information, and the angst it brings with, the stock prices respond negatively (Wang

& Mayes, 2012). Thus unconventional policy, e.g. monetary easing, generally leads to a raise in stock prices (Haitsma et al., 2016). In crisis periods this might differ due to signalling, as investors might conclude that economic prospects are not positive and they adjust their investment accordingly (Kontonikas et al., 2013). As a result the stock prices may decrease rather than increase, as they would under normal circumstances.

While there have been multiple studies on the effect of monetary policy on stock markets, most of these focus on the effect in the United States, such as the studies by Bernanke and Kuttner (2005) and Rigobon and Sack (2004), or in a specific European country, such as the study of Germany by Fausch and Sigonius (2018). Hence a study regarding Europe would be of great contribution. Besides this, by prolonging the examined time period, it will also be interesting to see whether there are any apparent effects of the newest unconventional monetary policies of the ECB.

The most recent, and still ongoing, economic disruption in the Eurozone followed from the Covid-19 pandemic. The ECB responded with the introduction of the monetary policy Pandemic Emergency Purchase Programme (PEPP) in March 2020 (Castellarin, 2020). The European Central Bank (2021a) states PEPP is a non- standard monetary policy and a temporary asset purchase programme of both private and public sector securities. The envelope consisted of 750 billion euros, with December 2020 being stated as the original end date, although this has since been postponed. As PEPP was implemented so recently, the effects of the implementation are not yet clear.

This study follows Haitsma et al. (2016), which in turn pursues the event study method of Bernanke and Kuttner (2005). While this is the most commonly used approach, there is another by Rigobon and Sack (2004) who suggest the identification through heteroscedasticity approach. The different approaches results in mixed estimates, possibly due to the fact the latter approach relies on weaker assumptions and is robust to omitted variable bias and endogeneity (Fausch and Sigonius, 2018).

The goal of this paper is to examine the reaction of unanticipated (un)conventional

monetary policies on stock markets. This is due to the fact that anticipated monetary policy is expected to have zero effect on the grounds of the efficient market

hypothesis by Malkiel and Fama (1970), as the theory states that only surprises should affect the stock prices.

In this case, the effect of the monetary policies of the ECB on the stock market returns regarding the EURO STOXX 50 index is examined. These monetary policies entail the anticipated conventional policy, the unanticipated conventional policy and the unanticipated unconventional policy. The ECB has three key interest rates: the Deposit Facility rate, the Marginal Lending Facility rate and the Main Refinancing Operation rate; these are to be considered as three separate variables for the

anticipated conventional monetary policy. Accordingly, the model will have three separate regression analyses for the three key interest rates of the ECB; these will be explained further in the literature review.

This study also takes the presence of a crisis into account. As the time period examined ranges from 01-05-2001 – 01-05-2021, this presence of a crisis has been subdivided in the Subprime/Global Crisis, the Euro Crisis and the latest Covid-19 Crisis. Another control variable that will be included is the Morgan Stanley Capital International (MSCI) World Index excluding Europe. This is added to control for overall economic shifts in the rest of the world (Haitsma et al., 2016).

For all three regressions concerning the different rates of the ECB of all the estimates of the coefficients with regard to the monetary policies, only those of the unanticipated unconventional changes are significantly different from zero at a significance level of 1%. All six of the estimates regarding the coefficients of the conventional monetary policy changes are not significantly different from zero.

Therefore both the unanticipated and anticipates conventional monetary policies are not significantly associated with the stock returns, while the unanticipated

unconventional monetary policy is. These results lead to the assumption that only when a policy is both unanticipated and unconventional, the effect will significantly differ from zero.

In the matter of the effect of the presence of a crisis, neither the crisis dummies nor the interaction terms with the conventional policy changes are of any significance (with a maximum of 10%). It can therefore be concluded that, in times of crisis, conventional policy changes do not have a varying effect from time without crisis.

These results are in line with the results of previous papers, such as that of Haitsma et al. (2016). The limitations of this study include the somewhat difficult data collection via DataStream/Eikon. As a result, a shortcoming of this study entails not running empirical analyses on the different subgroups, e.g. industry portfolios as in Bernanke and Kuttner (2005). Another shortcoming is the result of not taking spillover effects from other central banks into account. Ricci (2015) shows that policies set by other central banks might affect the estimation results. In the case of the ECB he concludes that there is a positive reaction found in connection to policy changes of the U.S. Federal Reserve (Fed). However, it should be kept in mind that this study by Ricci (2015) is performed specially towards bank stock returns.

A contribution of this study includes the subdivision of the crisis period into multiple periods, as opposed to examining solely pre-crisis and crisis differences. This provides contrasting insights, as all the separate crisis periods in this study do not result in estimates that differ significantly from zero. Furthermore, the study contributes in the addition of the most recent data, including the Covid-19 time period. Finally, almost all studies perform analyses on the Fed and thus the United States, therefore performing the study on the ECB and thus Europe was of more value to contribute.

Robustness checks are performed with regard to the following four variables:

the FTSE 100 and the STOXX 600’s three subgroups: the Small Cap, the Mid Cap and the Large Cap, each separately as a dependent variable.

This introduction is followed by the literature review, in which previous studies and existing theories on the topic are discussed. The following methodology will discuss and explain the different variables and the model used. A more precise description of the collected data sample will subsequently be given. Thereafter, the results of the regression analyses will be interpreted and discussed. Finally, this will be followed by the robustness checks and the conclusion.

**2. Literature Review **

The first unconventional monetary policy, as previously mentioned, was the three- month Supplementary Long-Term Refinancing Operation which was announced on 22-08-2007. After this, multiple additional polices were introduced such as the Covered Bond Purchase Programme, the Securities Market Programme and Outright Monetary Transactions. Some of the newest policies are the PEPP (18-03-2020) and the Pandemic Emergency Longer-Term Refinancing Operations (30-04-2020). A complete list of announcements on these policies is available in Table A.1 in the Appendix.

The first implementation of monetary policy in the form of quantitative easing (QE) took place in January 2015. This Asset Purchase Programme was implemented due to a heavy drop in inflation rates and expectations, with the aim to re-establish price stability (Tuori, 2019; Urbschat & Watzka, 2020). QE is not exclusively

unconventional, it is also expansionary. Where expansionary policy is implemented to lower the ECB’s interest rate, the surprise aspect is implemented with the prospect of it influencing more than solely the short-term interest rate (Joyce et al., 2012).

Macroeconomic variables such as output, employment and inflation usually indicate the end goals of monetary policy. However, monetary policy instruments affect these variables indirectly at most (Bernanke & Kuttner, 2005). Bernanke and Kuttner (2005) state the most direct effects of the instruments are on the financial markets as a result of affected asset prices and returns via which policy makers try to achieve their end goals by reshaping economic behaviour. For this reason, it is essential to have knowledge on the relationship between monetary policy and asset prices in order to comprehend the transmission mechanism of monetary policy.

Furthermore, Bernanke and Kuttner (2005) provide good reasoning for the importance of obtaining quantitative estimates of the connections between monetary policy changes and stock prices. These monetary policy changes are conveyed to the stock market through multiple channels, including changes in the cost of capital and the values of private portfolios, also known as the wealth effect. Due to this rise in stock prices, other asset prices and the overall wealth, the risk of the creation of bubbles increases, which can have quite disastrous consequences for the economy (Huston & Spencer, 2018).

When the asset purchase programmes, in this case by the ECB, are too substantial or persistent, the creation of a bubble might be inevitable. The bubble is the result of uncontrollable increases in asset prices far beyond the desired scope (Huston & Spencer, 2018). Retrieving these quantitative estimations is of great use when considering future possible unconventional monetary policy programmes.

In the last two decades there have been a number of studies performed concerning the impact of monetary policy, usually focusing on the impact of unanticipated monetary policy changes. The majority of these studies analysed the relation between these monetary policy shocks and the stock market with regard to the Federal Reserve Policy, thus regarding the United States. These studies include those of Bernanke and Kuttner (2005), Kontonikas et al. (2013) and Rigobon and Sack (2004).

The predominant finding of these studies is that there is a significant reaction of asset markets due to monetary policy shocks (Fausch & Sigonius, 2018). With regard to the stock market, an unanticipated increase in the policy rate is linked with a decrease in stock prices and vice versa. Past studies suggest a relatively strong and consistent negative relationship between the ECB’s interest rates and the stock prices.

Bernanke and Kuttner (2005) find that a hypothetical unexpected 25-basis-point cut in the policy rate target, in this case that of the Federal funds, is linked with

approximately a 1% rise in broad stock indices on average. The results of Rigobon and Sack (2004) also indicate that a rise in short-term interest rates leads to a fall in stock prices.

On the other hand, Martin and Milas (2012) conclude that even if the first round of QE by central banks had positive effects, the potential following QE programmes will have little to no effect. They also state that such a large-scale immediate policy does more harm than a policy implemented more gradually. This study by Martin and Milas (2012) also concerns the Fed rather than the ECB. The main take away from previous literature is that the mixed results make it difficult to comprehend what should be taken into consideration from these studies when looking at the different possible monetary policies.

Although a greater number of studies examine the effects in the United States, there have been a few studies to examine the effects in Europe. An example of this is the study by Fausch and Sigonius (2018). This study analyses the effect of monetary

policy surprises by the ECB on the German stock market, and conclude the same negative relationship between the policy rate and the stock prices. They also find that a distinct and significant stock market response is solely observed in the case in which real interest rates are negative.

Furthermore, most studies regarding Europe examine one specific country and not Europe as a whole. Examples of this include Casiraghi et al. (2013), who studied the impact of unconventional monetary policy during the crisis in Italy; Doran et al.

(2013), who studied the effectiveness of the Securities Markets Programme in Ireland;

and Gibson et al. (2014), who studied the incentives of the Greek sovereign spreads.

The results of these studies are all mixed, making it difficult to draw any conclusions.

An important difference between the ECB and other central banks is the fact that the ECB introduced unconventional monetary policies very early on in the crisis (Haitsma et al., 2016). Other central banks only introduced these policies after hitting the zero lower bound. The ECB also even simultaneously practiced unconventional and conventional policies in some periods (de Haan et al., 2020).

When studying the impact crisis, Haitsma et al. (2016) find that the effect of the monetary policy of the ECB differs on the basis of it being a time of crisis or not.

They conclude that unconventional monetary policies impact the stock market significantly, while unanticipated conventional monetary policies do not in times of crisis, thus making the crisis an important aspect within this study.

In addition to the unconventional policies, the conventional policy of the ECB is the adjustment of their interest rates. The ECB has three key interest rates, namely the rates on the Deposit Facility (DF), the Marginal Lending Facility (MLF) and the Main Refinancing Operations (MRO).

Haitsma et al. (2016) state the main refinancing operations as the most prominent conventional monetary policy of the ECB. With these operations and its corresponding interest rate, the ECB supplies liquidity to financial institutions in return for collateral. The MRO rate entails both a fixed rate as well as a variable rate (European Central Bank, 2021b). The two facilities are used to make overnight deposits and obtain overnight liquidity (Haitsma et al., 2016). Through raising and decreasing these rates, the ECB affects both the interest rates and the liquidity of the market. As a conventional monetary measure the ECB cut the MRO, the DF and the MLF rates towards the effective lower bound during the financial crisis (Haitsma et al., 2016).

When conventional monetary policies were not longer sufficient in times of crisis, the ECB introduced multiple unconventional policies, such as the Long-Term Refinancing Operations (LTRO), the Securities Market Purchase Program (SMP) and QE (Haitsma et al., 2016). The frequency and height of the different rates and their changes are shown in Table A.2 in the Appendix.

An issue that arises with the study of this subject is reverse causality, as it is possible that interest rates not only change assets prices, but also vice versa. Both Bernanke and Kuttner (2005) and Rigobon and Sack (2003) study the occurrence of this and find that there are no explicit examples of situations in which this reverse causality takes place. The next section, the Methodology, further elaborates on endogeneity and omitted-variable bias, as this can also be an issue within this study.

This all leads to the following hypotheses that will be put to the test in this study:

Hypothesis 1: ‘Only unanticipated monetary policy changes affect stock returns.’

This first hypothesis concerns if the policy change is a surprise or not. As

aforementioned, it is expected that stock returns only react to surprises in monetary policy, as outlined in the efficient market hypothesis by Malkiel and Fama (1970).

This distinction has been proven to be important by Bernanke and Kuttner (2005) and Fausch and Sigonius (2018).

Hypothesis 2: ‘Only unanticipated unconventional monetary policy changes affect stock returns.’ This hypothesis involves more than the surprise element, as it also demands for the policy to be unconventional. Again, unconventional monetary policy was introduced after the ineffectiveness of short-term interest rate was proven when it reaches the zero lower bound (Joyce et al., 2012). It is therefore interesting to

examine whether the effect between the unanticipated policies differs.

Hypothesis 3: ‘The effect of conventional monetary policy changes is different during crises.’ Among others, Kontonikas et al. (2013) find a crisis has effect on the outcome of the analyses between monetary policy changes and stock returns. However, the size and significance of the results regarding the effect are still very mixed.

**3. Methodology **

Previous studies have shown that monetary policy has a significant effect on stock price movement, particularly when unanticipated. When attempting to directly

measure the reaction of the stock market on monetary policy announcements both the endogeneity problem and omitted-variables bias are likely to arise, making this measurement difficult and unsuitable. This is why multiple recent studies, such as Kuttner (2001), Bernanke and Kuttner (2005) and Fausch and Sigonius (2018), make use of the event study method, which is exercised to minimise these problematic side effects. Bernanke and Kuttner (2005) conclude that while the event study method is reliable, it is important to be mindful of the possible underestimation of the reaction to policy changes.

An event study uses high frequency data versus the weekly or monthly data otherwise usually used. This gives a more accurate and precise estimation. Due to the shortened period, the so-called event window, an attempt is made to dispose of the endogeneity problem and omitted-variables as much as possible. Kontonikas et al.

(2013) conclude that when daily data is used in an event study framework,

endogeneity becomes less of a concern. This is due to the fact that it is not probable that monetary policy is influenced by asset prices on the same day, so the reverse causality problem is minimised (Kontonikas et al., 2013).

It is also important to include conventional and unconventional monetary policy separately, as both could occur on the same day. This would mean that if solely unconventional monetary policy were to be included, the results would be biased as a result of omitted-variable bias. Ricci (2015) also finds that bank stocks responded more to unconventional than to conventional monetary policy changes.

In this study, the approach of Haitsma et al. (2016) is followed, which in turn follows both Kuttner (2001) and Bernanke and Kuttner (2005). These studies have proven it is essential to create a distinction between the anticipated and the unanticipated interest rate changes. The efficient market hypothesis by Malkiel and Fama (1970) is the basis for this theory, as it states that stock prices should only be affected by interest rate surprises and thus not by anticipated changes in monetary policy. Investors already

have information on the anticipated changes, and accordingly take this into account when valuating, even before the policy announcement. When the distinction is not made, this may result in biased results (Fausch & Sigonius, 2018).

The most frequently used approach in prior literature is to use market data of futures to measure unanticipated conventional monetary policy change. Bernanke and Kuttner (2005) make use of federal fund futures to obtain an appropriate estimate in the United States. Fausch and Sigonius (2018) make use of interest rate futures contracts in their study of Germany, as these are likely to be strongly shaped by the expectations of future policy rates on the market. The study on the use of a future rate as a predictor for the ECB’s policy rate by Bernoth and von Haagen (2004) finds that it is unbiased and has a high reliability when using the 3-month EURIBOR future rate. For this reason this rate will be used for the unanticipated conventional monetary policy change, and therefore also for the anticipated conventional monetary policy change.

For the unconventional monetary policy surprise, the study of Rogers et al.

(2014) uses the change in the spread between the 10-year German and Italian

government bond yields at the policy announcement time. Their reasoning for the use of this spread is that the unconventional monetary policies by the ECB were largely intended to lower sovereign spreads in the Eurozone. Along these lines, when the spread between the government bond yields rises following an unconventional monetary policy announcement, this suggests that the policy is tighter than predicted, and vice versa (Fausch & Sigonius, 2018).

As aforementioned, the study of Haitsma et al. (2016) presents empirical evidence that in times of crisis the ECB’s monetary policy has a different effect than when there is no crisis. Their study illustrates that the EURO STOXX 50 index is significantly affected by unconventional monetary policy in crisis. On the contrary, conventional monetary policy surprises have no effect on the stock market during times of crisis (Haitsma et al., 2016). As a result, it would be interesting and potentially necessary to include a crisis dummy within the study.

To further clarify, the formulas and estimations methods are given below.

For the unanticipated conventional component the following equation is used:

∆𝑟_{𝑡}^{𝑈𝐶} = 𝑓_{𝑠,𝑡}− 𝑓_{𝑠,𝑡−1} (1)

where ∆𝑟_{𝑡}^{𝑈𝐶} is the unanticipated interest rate change through conventional monetary
policy and 𝑓_{𝑠,𝑡}− 𝑓_{𝑠,𝑡−1} is the difference between the future spot rates, in this case the
3-month EURIBOR future rate. To calculate these future spot rates, the daily

settlement prices of the future are used. To transform the settlement price into the spot rate, the settlement price must be subtracted from 100. This is because the settlement price of the future is 100, when the yield of the future is 0%, thus giving the spot rate.

For this the following equation is used:

𝑓_{𝑠,𝑡} = 100 − 𝑠𝑒𝑡𝑡𝑙𝑒𝑚𝑒𝑛𝑡 𝑝𝑟𝑖𝑐𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑓𝑢𝑡𝑢𝑟𝑒 𝑎𝑡 𝑡𝑖𝑚𝑒 𝑡 (2)

For the anticipated conventional component of the regression, it should be taken into consideration that the ECB has not one but three key interest rates. This leads to slight differences in output.

For this the following equation is used:

∆𝑟_{𝑡}^{𝐴𝐶} = ∆𝑟_{𝑡}− ∆𝑟_{𝑡}^{𝑈𝐶} (3)

where ∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹}, ∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝐿𝐹} and ∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝑅𝑂} are the anticipated interest rate changes
through conventional monetary policy variable with regard the Deposit Facility,
Marginal Lending Facility and Main Refinancing Operations rates, respectively.

Furthermore, ∆𝑟_{𝑡}^{𝐷𝐹}, ∆𝑟_{𝑡}^{𝑀𝐿𝐹} and ∆𝑟_{𝑡}^{𝑀𝑅𝑂} are the corresponding actual interest rate
changes and ∆𝑟_{𝑡}^{𝑈𝐶} is the unanticipated interest rate change through conventional
monetary policy.

Next, for the unanticipated unconventional component the following equation is used:

∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡} = (𝑦_{𝑠,𝑡}^{𝐼} − 𝑦_{𝑠,𝑡}^{𝐺} ) − (𝑦_{𝑠,𝑡−1}^{𝐼} − 𝑦_{𝑠,𝑡−1}^{𝐺} ) (4)
where ∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡} is the spread between the 10-year bond yields of Italy and Germany
at time t and time t-1, which is used to proxy the unconventional monetary policy
shock. 𝑦_{𝑠}^{𝐼} and 𝑦_{𝑠}^{𝐺} are the 10-year bond yields of Italy and Germany respectively.

Lastly, to calculate the change in stock returns of the EURO STOXX the following equation is used:

∆𝑅_{𝑡}= 𝑙𝑛 𝑃_{𝑡}

𝑃_{𝑡−1} (5)

where ∆𝑅_{𝑡} is the change in the EURO STOXX’s return between time t and time t-1
and 𝑃_{𝑡} is the EURO STOXX’s closing stock price.

With these components, the following baseline model can be created:

∆𝑅_{𝑡} = 𝛼 + 𝛽_{1}∗ ∆𝑟_{𝑡}^{𝑈𝐶}+ 𝛽_{2}∗ ∆𝑟_{𝑡}^{𝐴𝐶} + 𝛽_{3}∗ ∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡}+ 𝜀_{𝑡} (6)
where ∆𝑅_{𝑡} is the change in the stock return from time t-1 to time t, 𝛼 is the constant,
𝛽_{1} is the coefficient related to the unanticipated conventional monetary policy change,
𝛽_{2} is the coefficient related to the anticipated conventional monetary policy change,
𝛽_{3} is the coefficient related to the unconventional monetary policy change and 𝜀_{𝑡} is
the error term which includes the impact of missing variables that influence the stock
return at time t.

The first and third beta in Eq. (6) regarding the unanticipated monetary policy changes, both conventional and unconventional, are expected to differ from zero and to be negative. This is due to the fact that these changes are unanticipated and

therefore a surprise, which results in a value different to zero. This follows from the efficient market hypothesis by Malkiel and Fama (1970), which states that only surprises should affect stock prices, and thus not anticipated changes. This leads to the expectation of the second beta, concerning anticipated conventional monetary policy change, to be zero. The reasoning behind the negative values for the first and third beta comes from previous studies, such as Bernanke and Kuttner (2005) and Fausch and Sigonius (2018), who find that policy changes are negatively related to stock prices.

For the control variable the study of follow Kuttner (2001) is followed. He uses the crisis dummy and the Morgan Stanley Capital International (MSCI) World Index excluding Europe. The latter is added to control for any economic shifts in the rest of the world (Haitsma et al., 2016).

For the crisis dummy it is important to determine the start time of the crisis. Ricci (2015) distinguishes three different time periods related to the crisis: the subprime crisis, the global financial crisis and the Euro sovereign debt crisis.

The following ranges of dates have been chosen on the basis of Ricci (2015):

• The Subprime and Global Crisis 01-06-2007 – 01-05-2010

• The Euro Sovereign Crisis 02-05-2010 – 30-06-2013

• The Covid-19 Crisis 18-03-2020 – present

The crisis dummies take value 1 during the specific crisis and 0 otherwise.

Haitsma et al. (2016) choose 22-08-2007 as the start of their crisis dummy. They split the time period they examine into two periods: pre-crisis and post-crisis with their study period running to 2015. However, further subdividing up the crisis periods seems to be reasonable when studying the Eurozone, especially now, with the current effects of Covid-19 still unclear.

For the Euro Sovereign Crisis, it was agreed upon to provide Greece bilateral loans of €80 billion on 2 May 2010, this makes this date a logical start of the time dummy (De Haan et al., 2020).

When including these crisis dummies -𝐶_{𝑡}^{𝑆𝐺}, 𝐶_{𝑡}^{𝐸}, 𝐶_{𝑡}^{𝐶}- as interaction terms with both
conventional policy terms, as control variables and the MSCI excluding Europe as
control variable in the model, the following occurs:

∆𝑅_{𝑡} = 𝛼 + 𝛽_{1}∗ ∆𝑟_{𝑡}^{𝑈𝐶}+ 𝛽_{2}∗ ∆𝑟_{𝑡}^{𝐴𝐶}+ 𝛽_{3}∗ ∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡}+ 𝛽_{4}∗ 𝐶_{𝑡}^{𝑆𝐺}𝑋∆𝑟_{𝑡}^{𝑈𝐶}
+ 𝛽_{5}∗ 𝐶_{𝑡}^{𝐸}𝑋∆𝑟_{𝑡}^{𝑈𝐶}+ 𝛽_{6}∗ 𝐶_{𝑡}^{𝐶}𝑋∆𝑟_{𝑡}^{𝑈𝐶}+ 𝛽_{7}∗ 𝐶_{𝑡}^{𝑆𝐺}𝑋∆𝑟_{𝑡}^{𝐴𝐶} + 𝛽_{8}

∗ 𝐶_{𝑡}^{𝐸}𝑋∆𝑟_{𝑡}^{𝐴𝐶} + 𝛽_{9}∗ 𝐶_{𝑡}^{𝐶}𝑋∆𝑟_{𝑡}^{𝐴𝐶}+ 𝛽_{10}∗ 𝐶_{𝑡}^{𝑆𝐺} + 𝛽_{11}∗ 𝐶_{𝑡}^{𝐸}
+ 𝛽_{12}∗ 𝐶_{𝑡}^{𝐶}+ 𝛽_{13}∗ ∆𝑅_{𝑡}𝑀𝑆𝐶𝐼 𝑒𝑥 𝐸𝑈𝑅𝑂+ 𝜀_{𝑡}

(7)

where ∆𝑅_{𝑡} is the change in the stock return, 𝛼 is the constant, 𝛽_{1} is the coefficient
related to the unanticipated conventional monetary policy change, 𝛽_{2} is the coefficient
related to the anticipated conventional monetary policy change, 𝛽_{3} is the coefficient
related to the unconventional monetary policy change. 𝛽_{4}, 𝛽_{5} and 𝛽_{6} are the

coefficients related to the interaction between the unanticipated conventional
monetary policy change and the three crisis dummies. 𝛽_{7}, 𝛽_{8} and 𝛽_{9} are the

coefficients related to the interaction between the anticipated conventional monetary
policy change and the three crisis dummies. 𝛽_{10}, 𝛽_{11} and 𝛽_{12} are the coefficients
related to the three crisis dummies being the Subprime Global Crisis, the Euro
Sovereign Crisis and the Covid-19 Crisis, respectively. 𝜀_{𝑡} is the error term which
includes the impact of missing variables that influence the stock return. Again,
because of the three different values of ∆𝑟_{𝑡}^{𝐴𝐶} and thus of the interaction terms

between the crises and ∆𝑟_{𝑡}^{𝐴𝐶}, three separate regressions will be run with regard to the
separate DF, MLF and MRO rates.

The first three betas predictions in Eq. (7) are expected to remain the same as in Eq.

(6). For the crisis dummies and their interaction terms with them, it is uncertain which sign and values are expected. As aforementioned, Kontonikas et al. (2013) state that in crisis periods the relation between stock prices and unanticipated policies, which is usually negative, might differ due to the signals received by investors on the

economic prospects. If they feel the monetary policy signals mediocre prospects, they will adjust their investment accordingly, and the relation between stock prices and policies may turn positive.

Naturally, the interaction terms with the anticipated policy is still expected to be zero, as still they should not affect stock prices even during crises. The MSCI excluding Europe is expected to be positive, as indices tend to be positively related to each other. The correlation between these stock returns can be found in Table A.3 in the Appendix.

**4. Data and descriptive statistics **

All data used has been collected from DataStream and from the online database of the ECB. The 10-year German and Italian government bond yields, the daily settlement prices of the 3-month EURIBOR future rate and the EURO STOXX’s stock close prices are all obtained from DataStream. While the key ECB interest rates, the MRO, the DF and the MLF, are all obtained via the website of the ECB. The total time period taken in this study covers the past 20 years, running from 01-05-2001 – 01-05- 2021.

In the Appendix, Table A.1 shows the list of the announcements regarding an unconventional policy change, as well as if it was solely an unconventional

announcement or also a conventional announcement as well. It illustrates that since the first introduction on 22-08-2007, unconventional monetary polices have been introduced quite often, which emphasises the importance the ECB gives to

implementing unconventional policies. It can also be noted from Table A.1 that the combination of unconventional and conventional monetary policies has been implemented a few times. During the time period 22-08-2007 – 10-12-2020, there were a total of 46 announcements on unconventional policy changes. On 5 of these days there was also a change in conventional policy, and thus a change in one of the key interest rates of the ECB. The importance of observing this simultaneous

implementation comes from the fact that the ECB distinguishes itself from other central banks when this is done (de Haan et al., 2020).

For this list of announcements, data was taken from Rogers et al. (2014), which provided the announcement up until April 2014. The list was further updated by hand for all announcements after this date, up to May 2021, using the data on the ECB website.

Table 1.1 shows the descriptive statistics of the retrieved data variables. This so- called ‘raw’ data, such as close prices, is directly collected from the databases of DataStream and the ECB.

Variable Mean Median SD Min

Quantiles

25^{th} percentile 75^{th} percentile Max Obs

Yield Italy 3.561 3.561 1.435 0.454 2.229 4.550 7.287 5,072

Yield Germany 2.270 2.270 1.772 -0.854 0.469 3.932 5.280 5,072

Futures Close 120.803 120.803 14.356 80.157 112.444 130.858 152.052 5,072

EURO STOXX Close 3,849.667 3,849.667 804.916 2,035.345 3,330.822 4,170.670 6,572.847 5,072

ECB DF 0.696 0.696 1.178 -0.500 -0.400 1.250 3.750 5,072

ECB MLF 2.053 2.053 1.692 0.250 0.250 3.250 5.750 5,072

ECB MRO 1.395 1.395 1.406 0.000 0.000 2.250 4.750 5,072

MSCI Close 179.564 179.564 57.570 83.428 135.280 212.635 364.130 5,072

*Note. This table gives an overview of all the collected data variables. SD represents standard deviation. Yield Italy represents the 10-year bond yield of Italy. Yield Germany *
represents the 10-year bond yield of Germany. Futures Close represents the daily settlement price of the 3-month EURIBOR future. EURO STOXX Close represents the
EURO STOXX's stock close price. ECB DF represents the ECB's deposit facility rate. ECB MLF represents the ECB's marginal lending facility rate. ECB MRO represents
the ECB's main refinancing rate. MSCI ex Euro Close represents the Morgan Stanley Capital International World Index excluding Europe close price. Source: DataStream
and the ECB.

as this is still big data with numbers that cannot be directly compared. This pertains to the Futures Close, EURO STOXX Close and the MSCI ex Europe Close.

For the Yields of Italy and Germany, the difference in the means and the medians can be observed with 3.562%, 4.036% and 2.271%, 2.524%, respectively.

Thus Italy has a higher value on average than Germany, while the yield of Germany has a little more variation with a standard deviation of 1.772% versus the standard deviation of 1.435% of Italy. For the three key interest rates, the average of the MLF rate is the highest (2.053%), followed by the MRO rate (1.396%) and the DF rate (0.697%).

Relative variables are created from the collected variables to enable comparison between the data. After creating these new variables, relations can be seen more clearly and some observations can be made. These created relative variables, which are used in Eq.(6) and Eq. (7), can be seen in Table 1.2.

Table 1.2 shows the descriptive statistics of the created data variables. The created variables are mostly relative variables, such as changes over time between time t-1 and time t. These variables are necessary to perform the desired regressions in this study. The calculation used to create these variables is shown in the

Methodology.

The means of the spread, the EURO STOXX’s and the MSCI excluding
Europe’s close-to-close returns as reported in Table 1.2, are all zero. Furthermore, all
three variables regarding the changes of the ECB’s key interest rates have a mean of
zero, though a difference in minimum and maximum can be seen; -1% and 0.5% for
the DF rate and -.75% and 0.25% for both the MLF and the MRO rate. The similarity
between the changes in the three anticipated conventional monetary policy interest
rates are clarified by the similarity in the changes of the ECB’s key interest rates, as
Eq. (3) deducts ∆𝑟𝑡𝑈𝐶 from the change in an ECB’s rate. With ∆𝑟𝑡𝑈𝐶used in all three
calculations and only slight differences in the ECB’s rates, this leads to minimal
variation between the different ∆𝑟_{𝑡}^{𝐴𝐶} values. Nevertheless, as they are all key interest
rates, it is of importance to perform regressions on them all.

Variable Mean Median SD Min

Quantiles

25^{th} percentile 75^{th} percentile Max Obs

∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡} 0.000 0.000 0.068 -0.775 -0.018 0.016 0.706 5,071

∆𝑅_{𝑡} 0.000 0.000 0.016 -0.131 -0.007 0.008 0.128 5,071

∆𝑟_{𝑡}^{𝑈𝐶} -0.007 -0.007 0.726 -4.926 -0.408 0.397 3.391 5,071

∆𝑟_{𝑡}^{𝐸𝐶𝐵,𝐷𝐹} -0.001 -0.001 0.031 -1.000 0.000 0.000 0.500 5,071

∆𝑟_{𝑡}^{𝐸𝐶𝐵,𝑀𝐿𝐹} -0.001 -0.001 0.029 -0.750 0.000 0.000 0.250 5,071

∆𝑟_{𝑡}^{𝐸𝐶𝐵,𝑀𝑅𝑂} -0.001 -0.001 0.028 -0.750 0.000 0.000 0.250 5,071

∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} 0.006 0.006 0.727 -3.391 -0.397 0.408 4.926 5,071

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝐿𝐹} 0.006 0.006 0.727 -3.391 -0.398 0.406 4.926 5,071

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝑅𝑂} 0.006 0.006 0.727 -3.391 -0.397 0.408 4.926 5,071

∆𝑅_{𝑡}𝑀𝑆𝐶𝐼 𝑒𝑥 𝐸𝑈𝑅𝑂 0.000 0.000 0.010 -0.100 -0.004 0.005 0.086 5,071

*Note. This table gives an overview of all the created data variables. SD represents standard deviation. ∆𝑠𝑝𝑟𝑒𝑎𝑑*𝑡 represents the spread between the 10-year bond yields of
Italy and Germany, respectively. ∆𝑅_{𝑡} represents the EURO STOXX's close-to-close return. ∆𝑟_{𝑡}^{𝑈𝐶} represents the unanticipated conventional interest rate change; the change in
future spot rates is used for this. ∆𝑟_{𝑡}^{𝐸𝐶𝐵,𝐷𝐹} represents the change in the ECB's deposit facility rate. ∆𝑟_{𝑡}^{𝐸𝐶𝐵,𝑀𝐿𝐹} represents the change in the ECB's marginal lending facility rate.

∆𝑟_{𝑡}^{𝐸𝐶𝐵,𝑀𝑅𝑂} represents the change in the ECB's main refinancing operations rate. ∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} represents the anticipated conventional interest rate change regarding the DF rate.

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝐿𝐹}represents the anticipated conventional interest rate change regarding the MLF rate. ∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝑅𝑂} represents the anticipated conventional interest rate change regarding
the MRO rate. ∆𝑅_{𝑡}𝑀𝑆𝐶𝐼 𝑒𝑥 𝐸𝑈𝑅𝑂 represents the Morgan Stanley Capital International World Index excluding Europe close-to-close return.

standard deviation is clearly visible as the lines in Fig. 1.2 show overall more variation in highs and lows than Fig. 1.1. Table 1.2 illustrates this through the

standard deviation values of 0.016% of ∆𝑅_{𝑡} (Fig. 1.1) and 0.726% of ∆𝑟_{𝑡}^{𝑈𝐶} (Fig. 1.2).

Fig. 1.3 shows all three of the ECB’s key interest rates in one figure. It is apparent that generally, the rates change at the same time by the same amount, with few exceptions. Table A.2 in the Appendix further illustrates this by displaying the frequency and height of the rate changes of all three key interest rates separately. The table also illustrates that from 2008 onwards, all rates lower and remain at a level lower than previously.

Fig. 1.4 displays the 10-year yield of Italy and Germany as well as the spread between these two yields. The red dashed line represents the start of the Euro Sovereign Crisis on 02-05-2010, as chosen in this paper due to this being the date it was agreed upon to provide Greece bilateral loans of €80 billion (De Haan et al., 2020). The figure shows the spread to be near zero up to 2008, while since circa 2008 forward the spread starts to vary more. The height of movement in the spread was during the Euro Sovereign Crisis (02-05-2010 – 30-06-2013, as taken in this paper).

**5. Estimation Results **

This section includes the estimates of both regression analyses and provides a description of the results. First, the baseline regression, Eq. (6) is observed and interpreted for all the separate regression regarding the three different ECB rates.

Thereafter the full model regression, Eq. (7), including interaction terms and control variable is examined and interpreted, once more for all three regressions regarding the different ECB rates. Robust standard errors are used in all six regression analyses.

The results regarding the estimates of the coefficients of the baseline regressions, Eq.

(6), for the three key interest rates of the ECB are given in Table 2.1. The regression includes a constant (𝛼) and the (un)anticipated (un)conventional variables as

discussed in the Methodology.

Table 2.1 illustrates that five values have significant values with a maximum
of 10%. The betas of the anticipated conventional monetary policy changes of the
MLF and MRO take on the values of 0.0190% with a significance level of 5% and
0.0174% with a significance level of 10%, respectively. The ∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} variable does
not have a significant beta in this analysis. These betas of the anticipated conventional
variable were not expected to be significantly different from zero, as this does not
concern a surprise policy. As aforementioned, according to the efficient market
hypothesis by Malkiel and Fama (1970), only surprises should affect stock prices.

Therefore the betas of the anticipated conventional monetary policy change of the MLF and MRO being significant is not in line with the existing theory. This

significance in columns (2) and (3) is also not in line with the first hypothesis, which states that only unanticipated monetary policy changes affect stock returns.

However, none of the betas with regard to the unanticipated conventional monetary policy are of significance in this analysis. These results are primarily in line with the results of Fausch and Sigonius (2018) and Haitsma et al. (2016). However, due to this insignificance, a side note needs to be placed on hypothesis 1 as the

unanticipated conventional policy change is thus not significantly different from zero, and only the unconventional variable is.

The ∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡} variable betas are significant at a 1% level (p<0.01) for all
three regressions and thus all three key interest rates. The significance and the value

of approximately -0.0771% for all three variables, which relate to unanticipated unconventional monetary policy, is in line with previous studies. Bernanke and Kuttner (2005), Rigobon and Sack (2004) and Fausch and Sigonius (2018) all conclude negative relations. These results are in line with the second hypothesis, which states that only unanticipated unconventional monetary policy changes affect stock returns.

Hypothesis 3 does not apply to these first three regression analyses, Eq. (6), as the crisis dummies are not yet added to the model. Thus only column (1), regarding the DF rate, is in line with the two applicable hypotheses in this analysis, with the side note of the unanticipated conventional policy change not being significant.

Despite this, it is likely for the values and significances of both the alpha and the betas to change in the subsequent regression, which includes interaction terms and control variables. The latter regression model, Eq. (7) will in all probability give a more true representation of the real coefficients than the estimates of this baseline regression model, Eq. (6).

Table 2.2 gives the estimates of the coefficients of the full regression analysis, including the interaction terms and the control variables as discussed in the

Methodology. Once more, the regression is done three times, each one executed with regard to the different ECB rates.

The results show the estimates of the coefficients of the ∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡} variable
and the ∆𝑅_{𝑡}𝑀𝑆𝐶𝐼 𝑒𝑥 𝐸𝑈𝑅𝑂 variable to be the only estimates to have significance at a 1%

level (p<0.01). Nevertheless these are both significant in all three the columns.

The estimate of the beta regarding the unanticipated unconventional variable
(∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡}) is approximately -0.040% for all three key interest rates. The significant
negative association to the stock prices, albeit small, follows the expectation of both
the efficient market hypothesis by Malkiel and Fama (1970) and the results of
previous studies such as Fausch and Sigonius (2018).

Furthermore, both the unanticipated and anticipated conventional estimates are now both not significantly different from zero, where the latter was of some

significance in the previous analysis. The anticipated conventional estimates now follow the expectation of the efficient market hypothesis by Malkiel and Fama (1970), which argues that anticipated changes should not associate with stock prices.

**Table 2.1 **

*Baseline Regression Analysis: Types of Monetary Policy *

Dependent Variable: ∆𝑅_{𝑡}

(1) (2) (3)

DF MLF MRO

∆𝑟_{𝑡}^{𝑈𝐶} 0.0026 0.0107 0.0091

(0.0092) (0.0096) (0.0106)

∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡} -0.0771*** -0.0771*** -0.0771***

(0.0056) (0.0056) (0.0056)

∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} 0.0110

(0.0092)

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝐿𝐹} 0.0190**

(0.0096)

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝑅𝑂} 0.0174*

(0.0106)

Constant 0.0001 0.0001 0.0001

(0.0002) (0.0002) (0.0002)

Observations 5,071 5,071 5,071

R-squared 0.2862 0.2869 0.2866

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

*Note. This table reports the empirical link between the EURO STOXX's close-to-close returns and the *
(un)anticipated (un)conventional monetary policy changes without control. ∆𝑅𝑡 represents the EURO
STOXX's close-to-close return and is the dependent variable. ∆𝑟_{𝑡}^{𝑈𝐶} represents the unanticipated interest
rate change through conventional monetary policy; the change in future spot rates is used for this.

∆𝑠𝑝𝑟𝑒𝑎𝑑𝑡 represents the spread between the 10-year bond yields of Italy and Germany, respectively.

∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} represents the anticipated interest rate change through conventional monetary policy regarding
the deposit facility rate. ∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝐿𝐹}represents the anticipated interest rate change through conventional
monetary policy regarding the marginal lending facility rate. ∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝑅𝑂} represents the anticipated
interest rate change through conventional monetary policy regarding the main refinancing rate.

The estimates of unanticipated convectional changes were expected to be significantly different from zero with a negative relation to the stock prices. However, this does not seem to be the case when interpreting the results. While not in line with the

expectations through theory, it is in line which previous studies such as Fausch and Sigonius (2018) and Haitsma et al. (2016).

The combination of these results is completely in line with hypothesis 2, as only unanticipated unconventional monetary policy changes affect stock returns and neither of the conventional monetary policy changes do. However, as with the previous regression analyses, a side note has to be added to the first hypothesis, as again the variable concerning the unanticipated conventional policy change is not significant.

Now that the crisis dummies and interaction variables are added into the model, these additional variables can be interpreted with regard to the third hypothesis. However, none of the estimates regarding any of the crisis periods is significant (at a maximum level of 10%). This leads to the conclusion that the results are not in line with the third hypothesis, thus meaning that a crisis does not affect stock prices and does not change the effect conventional monetary policy changes have on stock prices.

Lastly, the control variable with respect to the change in MSCI excluding Europe is highly significant with a level of 1% (p<0.01) for all regression analyses.

The beta estimates of this variable are approximately 1.00% for all three ECB rates.

During the regression analyses it becomes clear that the interaction term regarding the
anticipated conventional policy rate and the Covid crisis dummy was biased within all
the previous three regressions. To prevent omitted-variable bias, the ∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} X

Covid, ∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝐿𝐹} X Covid and ∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝑅𝑂} X Covid variables were removed. The logic
for this omitted-variable bias can be explained by looking at the data and Eq. (3),
which is used to calculate the anticipated conventional policy rate change.

The formula in Eq. (3) that is used, ∆𝑟_{𝑡}^{𝐴𝐶} = ∆𝑟_{𝑡}− ∆𝑟_{𝑡}^{𝑈𝐶}, includes the change
in the actual interest rate. This actual interest rate is the ECB interest rate, which in
the regression entails the DF, MLF and MRO rate. When observing the data of the
ECB interest rates, it is evident these rates do not chance in the Covid crisis time
period, and thus ∆𝑟_{𝑡} is zero (this can also be seen in Fig. 1.3). As a result the formula

**Table 2.2 **

*Full Model Regression Analysis: Types of Monetary Policy including Control *
Dependent Variable: ∆𝑅_{𝑡}

(1) (2) (3)

DF MLF MRO

∆𝑟_{𝑡}^{𝑈𝐶} 0.0003 0.0013 0.0009

(0.0085) (0.0079) (0.0084)

∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡} -0.0403*** -0.0403*** -0.0404***

(0.0028) (0.0028) (0.0028) Subprime Global Crisis Dummy 0.0001 0.0001 0.0001

(0.0005) (0.0005) (0.0005)

Euro Crisis Dummy -0.0000 -0.0000 -0.0000

(0.0003) (0.0003) (0.0003)

Covid Dummy -0.0002 -0.0002 -0.0002

(0.0007) (0.0007) (0.0007)

∆𝑟_{𝑡}^{𝑈𝐶} X Subprime Global Crisis 0.0126 -0.0007 0.0017
(0.0101) (0.0121) (0.0114)

∆𝑟_{𝑡}^{𝑈𝐶} X Euro Crisis -0.0264 -0.0179 -0.0231

(0.0223) (0.0149) (0.0194)

∆𝑟_{𝑡}^{𝑈𝐶} X Covid -0.0012 -0.0012 -0.0012

(0.0014) (0.0014) (0.0014)

∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} 0.0049

(0.0085)

∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} X Subprime Global Crisis 0.0147
(0.0102)

∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} X Euro Crisis -0.0242
(0.0223)

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝐿𝐹} 0.0059

(0.0079)

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝐿𝐹} X Subprime Global Crisis 0.0014
(0.0121)

Dependent Variable: ∆𝑅_{𝑡}

(1) (2) (3)

DF MLF MRO

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝐿𝐹} X Euro Crisis -0.0157

(0.0150)

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝑅𝑂} 0.0055

(0.0084)

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝑅𝑂} X Subprime Global
Crisis

0.0037

(0.0114)

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝑅𝑂} X Euro Crisis -0.0209

(0.0194)

∆𝑅_{𝑡}𝑀𝑆𝐶𝐼 𝑒𝑥 𝐸𝑈𝑅𝑂 1.0017*** 1.0005*** 1.0006***

(0.0284) (0.0285) (0.0285)

Constant -0.0000 -0.0000 -0.0000

(0.0002) (0.0002) (0.0002)

Observations 5,071 5,071 5,071

R-squared 0.6494 0.6486 0.6487

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

*Note. This table reports the empirical link between the EURO STOXX's close-to-close returns and the *
(un)anticipated (un)conventional monetary policy changes as well as interaction terms with multiple
crisis periods and control variables including the crisis dummies and the MSCI ex Europe close-to-
close returns. ∆𝑅𝑡 represents the EURO STOXX's close-to-close return and is the dependent

variable. ∆𝑟𝑡𝑈𝐶 represents the unanticipated interest rate change through conventional monetary policy;

the change in future spot rates is used for this. ∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡} represents the spread between the 10-year
bond yields of Italy and Germany, respectively. The crisis dummies entail the Subprime Global (01-06-
2007 – 01-05-2010), the Euro Sovereign (02-05-2010 – 30-06-2010) and the Covid-19 Crisis (18-03-
2020 – present), which take value 1 when the crisis is present. ∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} represents the anticipated
interest rate change through conventional monetary policy regarding the DF rate. ∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝐿𝐹}represents
the anticipated interest rate change through conventional monetary policy regarding the MLF
rate. ∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝑅𝑂} represents the anticipated interest rate change through conventional monetary policy
regarding the MRO rate. ∆𝑅_{𝑡}𝑀𝑆𝐶𝐼 𝑒𝑥 𝐸𝑈𝑅𝑂 represents the MSCI Index excl. Europe close-to-close return.

is now ∆𝑟_{𝑡}^{𝐴𝐶} = 0 − ∆𝑟_{𝑡}^{𝑈𝐶} = − ∆𝑟_{𝑡}^{𝑈𝐶} and the interaction term therefore becomes

− ∆𝑟_{𝑡}^{𝑈𝐶} X Covid, which is in the negative version of the already included interaction
term ∆𝑟_{𝑡}^{𝑈𝐶} X Covid, included in all three regressions.

Consequently, the correlation between these independent variables in these
three regression analyses leads to omitted-variable bias. These three interaction term
variables regarding the Covid-19 dummy have therefore been removed from the
regressions. The correlation between ∆𝑟_{𝑡}^{𝑈𝐶} and ∆𝑟_{𝑡}^{𝐴𝐶} for all three of the ECB rates
can be found in Table A.4 in the Appendix.

**6. Robustness Checks **

To test for robustness, additional regression analyses will be performed with related dependent variables. The additional dependent variables include the Financial Times Stock Exchange 100 index (FTSE 100) and the STOXX Europe 600. The latter is subdivided into three sections based on their market capitalisation, resulting in the EURO STOXX Small Cap, Mid Cap and Large Cap, each containing 200

components. A correlation matrix of these additional dependent variables with the EURO STOXX 50 index and the MSCI ex Europe index can be seen in Table A.3 in the Appendix.

All additional data was collected from DataStream, which led to a minor limitation. DataStream allows this data to be downloaded up until 20 years prior. Due to data collection taking place in 16-06-2021, data was collected from 16-06-2001 – 01-05-2021 (versus the main study 01-05-2001 – 01-05-2021). However, this effect is expected to be negligible as the difference is very minor. The number of observations is now 4,968 rather than the 5,071 observations before.

Table 3.1 shows the descriptive statistics of the four additional dependent variables. For all four the variables, both the mean and median have a value of zero. It may be observed that all values of the statistics are very similar, which is a logical consequence of the high correlation between the variables (Table A.3)

First, a regression analysis is done with regard to the FTSE 100 as dependent variable.

This regression analysis is done, as before, with respect to all three key interest rate of the ECB. The results of the estimates can be seen in Table 3.2.

Few things differ from the previous regression, which concerned the EURO
STOXX 50 as dependent variable. The estimates of the unanticipated unconventional
variable ∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡} and the control variable ∆𝑅_{𝑡}𝑀𝑆𝐶𝐼 𝑒𝑥 𝐸𝑈𝑅𝑂 are both still significant at
a level of 1%; this corresponds to the preceding results. However, when looking at
column (1), regarding the DF rate, both the estimates of the unanticipated and the
anticipated conventional monetary policy changes interacted with the Subprime
Global Crisis dummy are significant at a 1% level and have a positive value. For the
MLF rate, column (2), none of the other variables have significant estimates with a
maximum level of 10%. For the MRO rate, column (3), the same is true as with the

Variable Mean Median SD Min

Quantiles

25^{th} percentile 75^{th} percentile Max Obs

FTSE 100 0.000 0.000 0.014 -0.126 -0.006 0.007 0.118 4,968

EURO STOXX Small 0.000 0.000 0.014 -0.119 -0.006 0.007 0.099 4,968

EURO STOXX Mid 0.000 0.000 0.013 -0.127 -0.006 0.007 0.096 4,968

EURO STOXX Large 0.000 0.000 0.015 -0.130 -0.007 0.008 0.123 4,968

*Note. This table gives an overview of all the additional dependent variables. SD represents standard deviation. FTSE 100 Returns represents the FTSE’s close-to-close return. *

EURO STOXX Small represents the EURO STOXX's small cap close-to-close return. EURO STOXX Mid represents the EURO STOXX's mid cap close-to-close return.

EURO STOXX Large represents the EURO STOXX's large cap close-to-close return.

Dependent Variable: FTSE 100 Returns

(1) (2) (3)

DF MLF MRO

∆𝑟_{𝑡}^{𝑈𝐶} -0.0005 0.0010 0.0012

(0.0078) (0.0072) (0.0077)

∆𝑠𝑝𝑟𝑒𝑎𝑑_{𝑡} -0.0195*** -0.0196*** -0.0196***

(0.0023) (0.0023) (0.0023)

Subprime Global Crisis Dummy 0.0003 0.0002 0.0003

(0.0005) (0.0005) (0.0005)

Euro Crisis Dummy 0.0002 0.0002 0.0002

(0.0003) (0.0003) (0.0003)

Covid Dummy -0.0001 -0.0001 -0.0001

(0.0006) (0.0006) (0.0006)

∆𝑟_{𝑡}^{𝑈𝐶} X Subprime Global Crisis 0.0257*** 0.0143 0.0220**

(0.0087) (0.0117) (0.0095)

∆𝑟_{𝑡}^{𝑈𝐶} X Euro Crisis 0.0043 0.0013 0.0024

(0.0149) (0.0101) (0.0130)

∆𝑟_{𝑡}^{𝑈𝐶} X Covid -0.0014 -0.0014 -0.0014

(0.0013) (0.0013) (0.0013)

∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} 0.0029

(0.0078)

∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} X Subprime Global Crisis 0.0261***

(0.0088)

∆𝑟_{𝑡}^{𝐴𝐶,𝐷𝐹} X Euro Crisis 0.0024
(0.0149)

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝐿𝐹} 0.0044

(0.0072)

∆𝑟_{𝑡}^{𝐴𝐶,𝑀𝐿𝐹} X Subprime Global Crisis 0.0146
(0.0117)