• No results found

Asset Purchase Programs and the Liquidity Effect on the Stock Market

N/A
N/A
Protected

Academic year: 2021

Share "Asset Purchase Programs and the Liquidity Effect on the Stock Market"

Copied!
74
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Asset Purchase Programs and the Liquidity

Effect on the Stock Market

Abstract:

This paper examines the extent to which the vast amounts of liquidity that flowed into the economy, resulting from large-scale asset purchase programs by the central bank, are reinvested in the stock market. When to a sufficient extent, this would potentially yield a separate effect on the price of stocks: the liquidity effect. The impact of asset purchase programs implemented by the Federal Reserve, the European Central Bank, the Bank of England and the Bank of Japan are investigated in this paper. The liquidity effect is measured by determining the relationship between the change in the amount of reserves resulting from these asset purchase programs and the developments on the stock markets. The existence of the liquidity effect would imply that an increase in the reserves leads to a rise in stock prices. The results indicate that a positive relationship between the amount of reserves and the stock markets exists in case of the United States and the United Kingdom. However, this relationship does not sustain when measured via alternative specifications, which casts these findings into doubt. This paper finds only weak indications for the liquidity effect concerning the stock market and these indications remain limited to the United States and the United Kingdom.

JEL codes: E52, E58, G11, G12

Keywords: asset purchase programs, unconventional monetary policy, stock market, liquidity effect

Name: R.F.J.M. Veraart

Supervisor: prof. dr. L.H. Hoogduin

Course: Master Thesis Economics, University of Groningen

Course code: EBM877A20

(2)

2

1. Introduction

During and in the aftermath of the financial and economic crises of 2008-2009, the central banks started to run out of their possibility to deploy their conventional monetary policy measures to comply with their mandates. Namely, as the economic tides began to worsen, the central banks had to engage in more accommodative monetary policy strategies. As a result, the central bank policy rates hit the zero or an effective lower bound1, which made them no longer able to affect the market interest rates in that particular way. Therefore, to further lower market interest rates, the policymakers had to resort to other measures, such as forward guidance2 and large-scale asset purchases (LSAPs). These policy measures are nowadays denominated as unconventional monetary policy.

LSAPs or asset purchase programs (APPs) aim at lowering market interest rates via the purchase of particular assets (predominantly government bonds) on a large scale, once the policy rate is at the effective lower bound. The central banks can finance these programs via increasing the amount of reserves on their balance sheets. Since the central bank represents an enormous buyer, the prices of the targeted assets will increase, while the yields will decrease. This decrease is supposed to push other interest rates downwards as well. In this fashion, by lowering market interest rates, the central bank aspires to steer inflation in the direction of its target level.

Given the extraordinary and in some respects unprecedented nature of this type of monetary policy, both inside and outside the academic sphere, the effectiveness of unconventional monetary policy has received a lot of attention and has been extensively debated upon (Romer & Romer, 2013; Blinder et al., 2017). Critics have expressed fears that the unconventional monetary policy measures will either be ineffective or will result in unintended and unwanted side effects. In particular, the enormous increase of the central bank balance sheets is expected to do more harm than good to the economy (Romer & Romer, 2013). Besides the concerns about the effectiveness of unconventional monetary policy in general, Galí & Gambetti (2015)

1 Since zero is not necessarily the lower bound, sometimes this level is also referred to as the effective lower

bound. The effective lower bound can be slightly above or below zero as well. In this paper, the terms zero and effective lower bound are used interchangeably.

2Forward guidance entails that the central bank attempts to influence market expectations by explicitly stating a

(3)

3

even state that since the crisis the consensus has become that central banks should monitor developments of asset prices and eventually react to these developments.

Plenty of research has yet been performed to assess the total impact of the unconventional monetary policy measures. In the literature, the relationship between monetary policy, both of a conventional and unconventional nature, and the stock market has been extensively addressed. In general, these studies3 find that monetary policy with an expansionary character has a positive effect on the stock market. These studies seem to have been predominantly interested in the impact of the monetary policy measurements at the time of announcement. However, resulting from the announced policy measures, at the actual impact vast amounts of liquidity flow into the economy due to the large-scale asset purchase programs. In the literature, this seems to have been left unaddressed regarding the stock market. Therefore, this paper aims to unravel the relationship between the increase in reserves on the central bank's balance sheet and the stock market.

A paper by Krogstrup, Reynard, and Sutter (2012) has looked into the impact of the amount of liquidity entering the economy, following from APPs, on the yield of 10-year Treasury bonds. Based on this development, they argue for the existence of a liquidity effect. The financing method of LSAPs, the increase in central bank reserves, is supposed to result in this effect. Namely, investors who have sold their assets to the central bank are now holding on to liquidity on which the return is equal to zero. Since they are most likely seeking higher yields, they are tempted to reinvest the revenues in other assets. When this occurs on a sufficiently large scale concerning a particular asset, this results in a separate effect on that asset which they have dubbed the liquidity effect4. Krogstrup et al. (2012) have attempted to measure this in case of

the yield of 10-year Treasury bonds and have found some indications that point towards the existence of this effect.

This paper aims to measure whether the liquidity effect comes into play regarding the stock market as well. Four central banks have been selected that are known to have engaged in asset

3 Bernanke & Blinder (2005) and Patelis (1997) among others have found this effect in the case of conventional

monetary policy and Haitsma, Unalmis & de Haan (2015) and Rogers, Scotti and Wright (2014), among others, in the case of unconventional monetary policy. A more extensive account of this relationship will be presented in section 4.2.

4 In previous literature, the liquidity effect is used to describe the negative impact of an increase in the money

supply on short-term interest rates, for which empirical support can be found. This effect is assumed to disappear when the short-term interest rates have reached the zero lower bound. However, Krogstrup et al. (2012)

(4)

4

purchase programs (APPs) on a large scale. These consist of the Federal Reserve, the European Central Bank, the Bank of England and the Bank of Japan. The monetary policy arrangements, as well as the APPs implemented by each central bank, will be discussed. In addition, the impact on the balance sheet of the central bank and the counterparties following from the APPs is presented. Furthermore, this paper elaborates upon the liquidity effect. On top of that, existing literature on portfolio reallocation during asset purchase programs will be presented. Moreover, by conducting a time series analysis, the relationship between the stock market index and the amount of reserves will be assessed empirically.

This paper endeavours to contribute to the existing literature to address the liquidity effect concerning the stock market. The contribution of looking into this effect for the stock market lies in numerous factors. First of all, this paper investigates the impact of LSAPs at the time of implementation, rather than at announcement. On top of that, whereas most papers are interested in the total effect of monetary policy, the scope of this paper is explicitly an (unintended) side-effect. Moreover, while most papers limit their focus to monetary policy implications of one central bank, this paper discusses monetary policy implemented by four central banks.

The proceeding parts of the paper will be structured as following. Firstly, in the second section, the monetary policy arrangements and the implemented APPs will be discussed for each central bank. Moreover, the impact of these programs on the respective balance sheets will be presented. Thereafter, in the third section, the transmission channels of monetary policy to the stock market will be discussed, in particular the liquidity effect. Furthermore, section four comprises the literature on portfolio reallocation of different types of investors during asset purchase programs. On top of that, the existing literature concerning the stock market and monetary policy will be discussed. In the fifth section, the methodological framework to measure the liquidity effect is presented. Section six consists of a discussion of the data used in this paper. In the seventh section, the results will be presented and discussed. The eighth section will contain the limitations and suggestions for further research. The conclusions of this paper are presented in the ninth section.

(5)

5

the balance sheet of both the central bank and the counterparty (i.e. the party who sells assets to the central bank) following from asset purchase programs is shown.

2.1 Overview of monetary policy in countries with asset purchase programs

In this subsection, the operations of monetary policy in United States, the Euro Area, the United Kingdom and Japan will be presented. Moreover, the different APPs these countries have implemented will be described.

2.1.1 Overview of monetary policy in the United States

The central bank of the United States is more commonly known as the Federal Reserve (henceforth, the Fed). In the case of the Fed, the U.S. Congress has the power to determine its mandate. The Full Employment and Balanced Growth Act, also known as the Humphrey-Hawkins Full Employment Act, states that the mandate of the Fed consists of price stabilization, maximum employment, and moderate longer-term interest rates. The Federal Open Market Committee (FOMC) is appointed to set the policy of the Fed to act in accordance with its mandate. The Fed has several monetary policy tools available that enable the central bank to control short-term interest rates and to influence the availability as well as the cost of credit within the economy. In these ways, the Fed is supposed to be able to achieve its targets. Since the outburst of the economic and financial crises in 2007, the central bank was necessitated to invoke other, more unconventional, monetary policy measures (Romer & Romer, 2013).

Graph 1. The Effective Federal Funds Rate between 1-1-2000 and 1-1-2018. Data retrieved from

(6)

6

In the years before the crisis, adjusting the federal funds target rate was the most common monetary policy tool. By either selling or buying securities that were issued or backed by the U.S. government, the Fed was able to control the short-term money market interest rate. Thereby, it could ensure that the federal funds rate approached the target rate. A change in the federal funds rate will result in a shift in the market interest rates as well. Therefore, the Fed has the capacity to affect the real economy, albeit indirectly. Under normal circumstances, a lowering of the federal funds rate, will increase economic activity and thereby stimulate economic growth. By the time that the economic outlook started to worsen due to the outburst of the crisis in 2007, the FOMC regarded it necessary to decrease the federal funds rate continuously. Eventually, the federal funds rate hit a level where it could not be lowered any further, which is referred to as the zero or effective lower bound. Graph 1 clearly shows that the federal funds rate reached the zero lower bound at the end of 2008 and that it remained at that level for several years. As a consequence, the Fed had run out of the possibility to use the federal funds rate as a tool to implement more expansionary measures.

At the zero or effective lower bound, the central bank thus has to resort to other measures if it considers it necessary to push the market interest rates further downwards to ease monetary conditions. Therefore, in late 2008 the Fed announced the implementation of an unconventional monetary policy tool. Namely, the central bank announced that it would engage in large-scale asset purchases (LSAPs), which is more commonly referred to as Quantitative Easing (QE). The underlying thought of this policy move is that the purchase of longer-term government securities, funded by increasing the reserves on the liability side of the central bank's balance sheet, will result in downward pressure on longer-term interest rates. Basically, QE is intended to sort a similar effect as a decrease in the federal funds rate (on market interest rates in specific, and thus the economy in general), only accomplished through a different method. However, it requires the central bank to increase its balance sheet with enormous amounts of reserves, which results in a substantial flow of liquidity into the economy as well. In section 2.2, the impact on the balance sheet is discussed in more detail.

(7)

7

was used to purchase longer-term Treasury securities. This period was dubbed as the first round of QE. The Fed continued the purchasing of longer-term Treasury securities in the second round of QE. This took place between November 2010 and June 2011, where the Fed purchased more Treasury securities for the amount of 600 billion dollars. The period after the second round of QE, which lasted from September 2011 until December 2010, is referred to as Operation Twist5. In Operation Twist, the Fed engaged in a “maturity extension program” which consisted of the purchase of longer-term Treasury securities with maturities of 6 years or longer remaining at the value of 667 billion. Simultaneously, the Fed sold Treasury securities with maturities lasting less than three years for an equivalent amount of money. The fourth and final round of QE ran from September 2012 until October 2014. This round consisted of monthly purchases of long-term Treasury securities of 85 billion dollars and the monthly purchase of 40 billion dollars of MBS. In 2017, the Fed announced that it would start to reduce the size of its balance sheet. The reduction will take place gradually which means that for a while the balance sheet size will remain unusually large.

Graph 2. The total assets on the Fed balance sheet between 1-1-2004 and 1-1-2018. Data retrieved from https://fred.stlouisfed.org/.

Thus, invoked by the financial crises, the Fed has undertaken measures in the form of LSAPs which resulted in an enormous increase of its balance sheet. Since the crisis, other monetary policy measures, which are not mentioned in this overview, have been implemented and can

5 The FOMC previously engaged in an action in 1961 dubbed as ‘Operation Twist'. By flattening the yield curve,

(8)

8

also be denominated as unconventional monetary policy6. However, the reason that only the

implementation of LSAPs (or QE) is explicitly mentioned in this overview is that the focus of this paper will predominantly be on potential effects of the increase in central bank reserves on the stock market.

2.1.2 Overview of monetary policy in the Euro Area

The European Central Bank (ECB) is responsible for conducting monetary policy of all member states of the Economic and Monetary Union (EMU) that have adopted the Euro as a common currency. The ECB was established in June 1998 and became operative in January 1999 at the introduction of the euro. In the treaty of Maastricht, which legally arranges the establishment of the EMU (and thereby the ECB as well), the mandate of the ECB is enshrined. The mandate of the ECB is to maintain price stability. The quantitative definition of price stability by the ECB currently is to pursue an inflation rate of close to, but below two percent of the Harmonised Index of Consumer Prices (HICP). In the case of the ECB, the governing council is responsible for making decisions to act in accordance with its mandate. In normal times, the three conventional policy instruments available to the ECB are open market operations (OMO), standing facilities and minimum reserve requirements. However, invoked by the financial crisis, the ECB turned to other, more unconventional policy measures as well. These predominantly existed of several types of asset purchases programs as well as forward guidance. Before the crisis hit the Euro Area, the ECB mainly conducted monetary policy via the conventional monetary policy tools. An OMO boils down to the national central banks providing liquidity to the markets on a weekly basis with a maturity of a week as well. The ECB uses the standing facilities, consisting of the marginal lending and the deposit facility, to control the overnight liquidity market. The marginal lending facility enables counterparties to borrow liquidity if necessary at the ceiling rate and against appropriate collateral. The deposit facility allows counterparties to deposit excess liquidity at a specific rate, which determines the floor rate on the overnight market interest rate. By determining the minimum reserve requirements, the ECB can control money market interest rates. In a way, this allows the ECB to create or to expand a shortage in liquidity which forces the institutions to borrow liquidity. However, at the time that the crisis reached Europe as well, the conventional policy measures did no longer suffice. Consequently, the ECB was forced to come up with new policy measures

(9)

9

to comply with its mandate. In October 2008, the ECB announced a significant change in its refinancing policy, known as fixed rate full allotment. From that moment onwards, commercial banks could apply for unlimited liquidity provisions against appropriate collateral. In addition, the rate at which this liquidity will be provided became fixed whereas it used to be variable. At the time of announcement, the ECB expanded the list of acceptable collateral as well. Moreover, the ECB has conducted several rounds of different asset purchase programs as well. In 2009, the ECB launched the Covered Bond Purchase Program (CBPP1). A follow-up of this program, denominated as CBBP2, was initiated in November 2011. In the meantime, to make sure that monetary policy would be transmitted more smoothly, the Securities Market Program (SMP) was announced in May 2010. In order to remove the redenomination risk, the ECB announced the Outright Money Transaction in 2012. In spite of that, this program has never been applied to date. By June 2014, the ECB announced the implementation of several measures supposed to improve the functioning of monetary policy transmission. Still, it was not until January 2015 that the ECB actually engaged in asset purchase programs of government bonds on a large scale. The ECB initiated the purchase of bonds issued by euro area governments, which combined with the private sector programs consisted of the monthly amount of 60 billion euros up to March 2016. Between April 2016 and March 2016, the monthly purchases comprised 80 billion euros. From that date onwards, the amount was reduced back to 60 billion euros per month. The ECB decided to lower the monthly even further to 30 billion euros a month starting from January 2018.

All in all, the ECB has implemented several different purchase programs which served different means. Namely, some of these programs were mostly intended to improve the transmission of monetary policy. Only later on, when the ECB succeeded in improving the transmission of monetary policy, the central bank was able to implement programs that were directly aimed at acting in accordance with its targets. For that reason, these programs are referred to as QE conducted by the ECB as well.

2.1.3 Overview of monetary policy in the United Kingdom

(10)

10

employment. The actual decisions regarding monetary policy of the BoE are made by the Monetary Policy Committee. Its most popular policy tool is the Bank rate, which is the BoE’s official interest rate. Between October 2008 and March 2009, the BoE decided to lower the Bank rate from five percent to a half percent, which represents its effective lower bound. Therefore, in addition, to further ease monetary conditions the BoE decided upon the implementation of a QE program as well. The asset purchase program (APF) that was launched in March 2009 comprised of public sector purchases at the amount of 75 billion pounds. The assets purchased varied in lasting maturities between five and twenty-five years. Over time, the program was gradually expanded to 200 billion pounds by February 2010. The Monetary Policy Committee decided to relaunch the program in October 2011 leading to a total amount of 375 billion by the end of 2010. Similarly to the Fed, the BoE implemented QE relatively soon after it was faced with the crisis. The country has used two rounds of APPs. Thus far, the BoE has neither initiated nor announced the intention to unwind its balance sheet.

2.1.4 Overview of monetary policy in Japan

In the case of Japan, the Bank of Japan (BoJ) is responsible for conducting monetary policy. According to the Bank of Japan Act, monetary policy by the BoJ should be “aimed at achieving price stability, thereby contributing to the sound development of the national economy.’’ In January 2013, the BoJ has set this aim at a two percent rate of the annual change in the consumer price index (CPI). The Policy Board is in charge of deciding upon the monetary stance during Monetary Policy Meetings. The most common tool of the BoJ is the use of money market operations. The central bank can supply funds to financial institutions against sufficient collateral, which is referred to as a funds-supplying operation. Reversely, the BoJ can engage in so-called funds-absorbing operations, which it does by issuing and selling bills.

(11)

non-11

performing loans. The outright purchase programs consisted of monthly purchases of 400 billion yen.

In October 2008, the Bank of Japan announced that it would again lower its policy rate. Moreover, after the introduction of ‘Abenomics' in 2011, new rounds of QE were implemented. At the time of the implementation of these measurements, the quantitative definition of price stability was one percent inflation. The program was initially set at 35 trillion but was later on expanded to 101 trillion Yen. By January 2013, the quantitative definition of price stability changed from one to two percent. Where the annual purchases before this period comprised of 20 trillion, from this moment onwards, they were increased to 50 trillion a year. On 31 October 2014, the BoJ announced an additional expansion of the program amounting up to 80 trillion Yen of bonds purchases per year.

Thus, the BoJ was the first of the four central banks discussed in this paper to engage in asset purchase programs. Although the other countries also have had several rounds of QE, only Japan has experienced two distinct periods where it engaged in QE (at the beginning of this century and since ‘Abenomics’).

2.2 The balance sheet implications of asset purchase programs

In the previous section, the different asset purchase programs deployed by the four central banks have been presented. As has been mentioned, the central bank finances these programs by increasing the amount of reserves on its balance sheet. The overall mechanism of these asset purchase programs works the same, irrespective of the central bank and the type of assets that are purchased. In this section, an explanation is provided on how these APPs affect the balance sheets of the purchasers, the central banks, and the counterparties, predominantly banks or non-bank financial institutions. A similar explanation of the impact on the individual balance sheets (including stylized balance sheets) is presented by Christensen and Krogstrup (2018).

Figure 1. A highly simplified central bank balance sheet. *As described in section 2.1, the central bank can purchase various types of assets. Irrespective of the type of assets, the movements on the balance sheet are the same.

Central Bank

Assets Liabilities

(12)

12

When the central bank decides to purchase assets, these end up on the asset side of the central bank balance sheet. The central bank increases its reserves to finance these purchases made during the APPs. Thus, at the same time, the central bank reserves increase by an equal amount on the liability side of the balance sheet of the central bank. Figure 1 puts forward a stylized central bank balance sheet that shows the effect of asset purchases.

The central bank represents the buyer of the assets. The balance sheet of the seller is affected as well. As indicated, the sellers predominantly consist of commercial banks or non-bank financial institutions. Only banks can hold central bank reserves. Since reserves are used to finance the APPs, this means that if the central bank decides to purchase assets from non-bank financial institutions the balance sheet of a commercial bank is affected as well. Through Figure 2 and 3, it becomes clear what purchases from either commercial banks or non-banks financial firms entail for their balance sheets. If the central bank purchases directly from the commercial bank, only the non-bracketed arrows in Figure 2 are relevant. On the asset side of the balance sheet, the amount of assets decreases, and the amount of reserves increase.

Figure 2. A highly simplified commercial bank/ balance sheet. * As described in section 2.1, the central bank can purchase various types of assets. Irrespective of the type of assets, the movements on the balance sheet are the same.

If the central bank decides to purchase from a non-bank financial firm, Figure 3 and the bracketed arrows of Figure 2 come into play, while the non-bracketed arrows become irrelevant. Given the capacity of the commercial bank to hold reserves, these will go up while at the same time the amount of deposits increases on the liability side. In Figure 3, on the asset side, the amount of deposits increases while the amount of assets decreases.

The movements in the balance sheets presented above show that as a result of the purchases by the central bank, highlight a few notable events. Namely, the total supply of particular assets available to the market decreases significantly. Moreover, the amount of reserves increases with an amount equal to the total purchases of assets by the central bank. Although commercial banks are the only counterparty capable of holding reserves, this does not mean that they

Commercial Bank

Assets Liabilities

↓ Assets* (↑) Deposits

(13)

13

exclusively can sell to the central bank. Albeit indirectly, the central bank can purchase from other parties as well while financing via reserves.

Figure 3. A highly simplified non-bank financial firm balance sheet. * As described in section 2.1, the central bank can purchase various types of assets. Irrespective of the type of assets, the movements on the balance sheet are the same.

3. The transmission of monetary policy to the stock market

In the previous section, the implications for the balance sheet of the central bank and the counterparties have been put forward. In this section, the transmission mechanisms of monetary policy to the stock market are presented. Moreover, we will discuss how these transmission mechanisms are related to the movements on the balance sheets.

Generally, most channels are capable of transmitting multiple types of monetary policy measures. However, a possible additional channel that should exclusively hold in the case of APPs, if existent, will be presented. This channel is referred to as the liquidity effect (Krogstrup, Reynard, and Sutter, 2012) or the reserve-induced portfolio balance channel (Christensen and Krogstrup, 2018). Concerning this channel, the substantial increase in the reserves resulting from the central bank plays an essential role. The next subsection contains a description of the transmission mechanisms that are supposed to hold for both conventional and unconventional monetary policy measures. In the subsection thereafter, the possible additional channel in case of APPs (i.e., the liquidity effect) will be presented.

3.1 The general transmission channels of monetary policy

According to the Federal Reserve (2016), there are two broad mechanisms through which monetary policy is assumed to affect the stock market. Namely, monetary policy (conventional and unconventional) is supposed to affect market interest rates (which serves as a means to an end). As a consequence, if the interest rates decrease, investors might be inclined to purchase stocks which leads to a price increase (The Federal Reserve, 2016). Secondly, as interest rates fall, investors might expect a more positive outlook for the economy. Since this is considered to be beneficial for the performance of firms, the firms will increase in value. For that reason, stocks are likely to increase in attractiveness as well. Thus, in the case of the APPs, they are

Non-Bank Financial Firm

Assets Liabilities

(14)

14

supposed to result in changes in market interest rates and in turn affect the stock market. Moreover, if investors perceive the announcement of APPs as a measure that will boost the economy, the stock market will rise as well. A paper written by Brouwer (2017) discusses the transmission of monetary policy to the stock market as well. Using the dividend discount model, he argues that monetary policy is likely to affect the stock market either via a change in real interest rate or expected future dividends.

In the literature, the effect of monetary policy on the stock market is split up into several channels. First of all, monetary policy announcements are always subject to interpretation by investors. Investors consider policy announcements to be signals concerning future monetary policy actions and central bank commitment to a particular monetary policy stance (Krishnamurthy & Vissing-Jorgensen, 2011). For example, a more restrictive announcement after a series of expansionary monetary policy announcements might imply more contractionary measures in the future as well. When the market adjusts the prices of the stocks following a monetary policy announcement, the expectations concerning future monetary policy will be incorporated in the current price. This transmission mechanism is dubbed as the signaling channel. The forward guidance program is a prime example of how central banks can influence the market via this channel.

In a similar vein, investors interpret every monetary policy announcement as an indication of the central bank’s outlook for the economy. Under normal circumstances, a monetary policy measure of expansionary nature will lead to a rise in stock prices7. However, if investors

conceive this to be a sign of a lack of confidence in the economy by the central bank, it might lead to a decrease in stock prices. This channel is often coined as the confidence channel. Via the confidence channel, monetary policy can be transmitted internationally as well. For instance, a prime example where the confidence channel was at work abroad, was the announcement of the Outright Money Transaction (OMT)8 by the ECB (Falagiarda et al., 2015). When applied to the different APPs, these could be at work via the confidence channel both in a positive and negative direction dependent upon the interpretation by the investors.

Where the previously mentioned transmission channels are predominantly related to the expectations of the market, most other channels are concerned with the actual impact of monetary policy on other variables such as market interest rates. For example, the portfolio

(15)

15

rebalancing channel is concerned with the difference in the attractiveness of various financial assets. If the yields of particular assets, such as government bonds, will go down, the attractiveness of these assets will decrease. As a consequence, investors will rebalance their portfolios and increase the relative weight of other assets such as stocks or corporate bonds, which will increase the price of these assets. In case of the APPs, this works in the following way. Namely, resulting from the asset purchases by the central bank, the total supply of particular assets significantly decreases. This drives up the price of the asset in question and thereby pushes down the yield. The decrease in supply and yield of the particular asset forces investors to rebalance their portfolios (Gagnon et al., 2011b). Namely, they will seek for similar investment opportunities only with higher returns. As for the confidence channel, the portfolio re-balancing channel does not have to be limited to country borders since investors can decide to purchase comparable assets abroad.

Stock prices are determined by the performances of the firms as well. This means that some other channels of monetary policy, such as the credit channel, affect the stock market as well. Bernanke & Gertler (1995) have presented an extensive account of the credit channel, which can be split up into the balance sheet channel and the bank lending channel. They state that a change in the external finance premium, which is the difference in costs between external and internal funding, results in an amplification of a change in the interest rates caused by monetary policy. According to Bernanke & Gertler (1995), two connections exist through which monetary policy can affect the external finance premium; the balance sheet channel and the bank lending channel. The balance sheet channel means that a change in the interest rate will change the financial position of a firm via its balance sheet and thereby it affects the external finance premium. An improved balance sheet will lead to an increase in the firm value and enables firms to invest more which might increase firm value as well. This will both lead to a rise in the stock price. A worsened balanced sheet, on the other hand, will probably lead to a decrease in the stock price. Namely, as investors perceive the cash flow of a firm to be riskier, they demand an increase in the equity premium, which will be discounted at a higher rate as well. This will probably lead to a decrease in the stock price.

(16)

16

well. When following the reasoning of the ECB9, the implementation of the APPs should result

in a lowering of market interest rates which makes loans more inexpensive. Consequently, an increase in lending and cheaper debt repayments should induce a boost in both consumption and investment. The boost in consumption and investment is supposed to be beneficial for the economy and thereby firm performances, which in turn should increase stock prices as well. Thus, according to this reasoning, these channels should hold for the APPs as well.

3.2 The liquidity effect or reserve-induced portfolio balance channel

When bearing in mind the implications of APPs, the channels described above are concerned with several things. Firstly, every announcement relating to APPs is subject to an interpretation by the market, which will lead to an effect on the stock market. Moreover, related to the movements on the balance sheet, the total supply of particular assets available to the market drops substantially. The reduced supply impacts the stock market via the portfolio rebalancing channel. Lastly, the potential implications of the asset purchase programs for the real economy are of importance as well. However, the substantial increase in the central bank reserves remains unaddressed by the channels described above. The question is whether this might entail something for the stock market and, if so, how this would be the case.

A paper by Krogstrup, Reynard, and Sutter (2012) argues for the existence of a separate effect via the increase of reserves on the yield of 10-year Treasury bonds, which they have dubbed the liquidity effect. Christensen and Krogstrup (2018) describe the same effect, only referred to as the reserve-induced portfolio balance channel. Krogstrup et al. (2012) and Christensen and Krogstrup (2018) state that the existing literature, as for the channels presented in the previous section, predominantly focuses on the effects that monetary policy exerts either via the expectations of investors (i.e., the signaling and confidence channel) or changes in market interest rates. However, they argue that, in case of APPs, they expect an additional effect (i.e., the liquidity effect or reserve-induced portfolio balance channel) as a result of the increase in reserves held by counterparties. The theoretical underpinning of the liquidity effect is as follows. As shown in Figure 1, 2 and 3 on pages 11-13, when a central bank engages in APPs, the asset side of its balance sheet increases with the purchased assets. At the same time, the reserves increase with an equal amount on the liability side. Simultaneously, on the balance sheet of the counterparties, the amount of assets decreases, while the amount of reserves or

9 The ECB provides a seven-step reasoning on how QE is supposed to help the ECB to achieve its mandate via

(17)

17

deposits increases. As a consequence, they hold an amount of reserves on which the return is equal to zero. Therefore, holders of these reserves will be tempted to seek other investment opportunities with higher returns. If this happens on a sufficiently large scale concerning a particular asset, this will yield an effect on the price of this asset as well. Both Krogstrup et al. (2012) and Christensen and Krogstrup (2018) mention that such a channel is also argued for in an early work by Bernanke and Reinhart (2004). Moreover, they state that the liquidity effect is supposed to be independent of the type of asset purchased as its workings revolve around the increase in reserves and how these are reinvested. The scope of the research by Krogstrup et al. (2012) was the impact of the liquidity effect on the yield of 10-year Treasury bonds. Their empirical results bring forward some indications for the presence of the liquidity effect in case of the 10-year Treasury bonds. Moreover, they explicitly state that this effect is at work separately from and in addition to the portfolio rebalancing channel.

In line with the reasoning of Krogstrup et al. (2012), Christensen and Krogstrup (2018) state that the liquidity effect of LSAPs might spread to other financial assets as well. These findings put forward some interesting implications concerning this paper. Namely, assuming that a change in the yield of 10-year Treasury bonds affects the stock market, the LSAPs indeed sort an additional effect on the stock market price movements, albeit indirectly. Moreover, as the liability side effect might spread to various financial assets, the possibility exists that the liquidity effect will directly affect the stock market as well. Namely, if a substantial part of the reserves flows to the stock market, this will have its impact on the price.

(18)

18

4. Literature review

In this section, some literature will be presented on investor behaviour and flows of funds during asset purchase programs. Moreover, the existing stock of empirical research concerning the relationship between monetary policy and the stock market is presented.

4.1 Investor behavior and flows of Funds

In the literature, some theoretical and empirical work can be found, which is related to portfolio reallocation of investors induced by asset purchase programs. For example, the preferred-habitat theory, mentioned by Joyce (2012), Krogstrup et al. (2012) and Christensen and Krogstrup (2018) among others, helps to explain the behavior of investors. The theory states that investors prefer to hold a particular type of assets over other assets, which implies imperfect asset substitutability. This helps both explain why holders of particular assets require a premium when selling these to the central bank and what their investment strategy will be with the money retrieved from the sale of these assets. According to the preferred-habitat theory, investors might be inclined to reinvest their returns in comparable assets to the ones sold to the central bank. This theory would imply that investors are less likely to move to equity.

Some empirical studies have looked into the actual changes in the portfolio composition of counterparties involved in asset purchase programs conducted by central banks. These studies make use of Flow of Funds data to assess the changes in the portfolios of different types of investors. A paper by Carpenter, Demiralp, Ihrig, and Klee (2015) has looked into what types of investors sell assets to the Fed and what they do with their returns afterwards. They have found that “households” – which consists of hedge funds as well in their paper10 - form the

largest group from which the Fed buys the assets. The “households” seem to reallocate their portfolios to more risky assets, among which stocks. The methodology used by Carpenter et al. (2015) is to regress the Treasury holdings by the Fed on the nominal holdings of different types of investors. Thus, instead of using the increase in the amount of reserves following the APPs, they use the holdings by the Fed to measure the relationship with investor holdings. Based on their results, “households” have sold roughly sixty percent of the total assets purchased by the

10 The authors acknowledge that labelling hedge funds as households sounds confusing. However, Carpenter et

(19)

19

central bank during the first round of QE in the United States. This group of investors also seems to have reallocated towards equity, among other assets.

A study by Eksi et al. (2017) uses the same methodology to investigate the investment decisions of the households further, as they represent the largest seller to the Fed. In line with Carpenter et al. (2015), their results put forward that households increased the relative weight of equity as a result of LSAP conducted by the Fed. Saita and Hogen (2014) have looked into portfolio reallocation by investors in Japan. Overall, they find that investors decided to hold fewer government bonds and to increase the relative weights of equity in their portfolio. In particular, Saito and Hogen (2014) state that domestic banks and overseas investors seem to have favored equity over other types of assets. Koijen, Koulischer, Nguyen, and Yogo (2018) have looked into portfolio reallocation following from QE in the Euro Area. They indicate that reallocating to equity is limited exclusively to mutual funds.

(20)

20

The findings of these papers show that at least some of the returns following from asset purchases programs have been reinvested in equity. However, the results also put forward, although dependent on the country and the type of investors, in most instances investors did not reinvest in equity. Since the scope of these papers was limited to particular rounds of APPs and particular types of investors, it remains unclear how investors behaved during other rounds of QE. While in some instances (a part of) the reserves end up in equity, in the majority of the cases this does not seem to occur. The results of these papers suggest that, if existent, the liquidity effect probably will remain limited.

4.2 The relationship between monetary policy and the stock market

In the previous section, we have discussed the way in which monetary policy can reach the stock market. In the existing stock of literature, research can be found that have empirically assessed the relationship between monetary policy and the stock market. Concerning this paper, we are specifically interested in the relationship between APPs and the stock market. Nevertheless, a brief overview is provided of previously conducted research between monetary policy in general (both conventional and unconventional) and the stock market.

Sellin (2001) and Thorbecke (1997) indicate that early works on the matter were inconclusive concerning whether monetary policy indeed sorts an effect on the stock market. Nonetheless, later works virtually unambiguously show that there is, in fact, a relationship between monetary policy and the stock market. Several studies performed by Thorbecke (1997), Rigobon & Sack (2004), Ehrman & Fratzscher (2004) and Bernanke & Kuttner (2005) among many others have found that conventional monetary policy of an expansionary nature (i.e., a decrease in the interest rate) has a positive effect on the stock market. These papers looked into the stock market reactions of the Wilshire 5000 (Rigobon & Sack, 2004), Nasdaq (Rigobon & Sack, 2004), Dow Jones Industrial Average (Thorbecke, 1997; Rigobon & Sack, 2004) and the S&P 500 (Thorbecke, 1997; Rigobon & Sack, 2004; Ehrman & Fratzscher, 2004; Bernanke & Kuttner, 2005). Additionally, some of these studies (Thorbecke, 1997; Erhman & Fratzscher, 2004; Bernanke & Kuttner, 2005) have found that dependent on different factors, such as the type of industry, firm size or financial characteristics, stock prices show different responses. Thus, in general, in the case of conventional monetary policy, a decrease in the interest rate, will lead to an increase in stock prices.

(21)

21

500 found a positive relationship between unconventional monetary policy and the stock market. Even more, some of the studies have extended the scope of their research by drawing comparisons with the effect of conventional monetary policy on the stock market. For example, a study by Haitsma, Unalmis & de Haan (2015) has found that unconventional monetary surprises of expansionary nature by the ECB have a positive and more profound effect on stock prices of the EURO STOXX 50 index than conventional monetary policy surprises. In a similar vein, a previous study conducted by Brouwer (2017) has looked into the impact of both conventional and unconventional monetary policy measures on the Dutch stock market. His results put forward that only in case of unconventional monetary policy a statistically significant effect can be detected. A study by Rogers, Scotti, and Wright (2014) indicates, similar to the study of Haitsma et al. (2015) that unconventional monetary policy of an expansionary nature has a positive effect on the stock market response in the U.S. and the Euro area. In contrast to what Haitsma et al. (2015) find for the Euro Area, Rogers et al. (2014) state that the effects of conventional monetary policy in the United States have a more substantial impact on stock prices than unconventional monetary policy. On the other hand, a paper by Eksi, Kamil & Tas (2017) finds that compared to conventional monetary policy, the effect of unconventional monetary policy is even seven times larger in the case of the S&P 500.

Besides the effects of unconventional and conventional monetary policy on asset markets, the study of Rogers et al. (2014) has also looked into the asymmetry in effects of unconventional monetary policy. They have found that in the case of the U.S. and the U.K. the positive effect of an unconventional monetary policy expansion on the stock market is more significant than the negative effect of monetary policy tightening. Nonetheless, in general, unconventional monetary policy appears to show a similar effect on the stock market as conventional monetary policy.

(22)

22

Swanson (2005), which relies upon the assumption that the impact of forward guidance is the same before and during the effective lower bound period. This allows determining the individual effects of both policy measures. In contrast with the findings of Rogers et al. (2014), the results of Swanson (2015) point towards significant effects of both forward guidance and LSAPs. Both Rogers et al. (2014) and Swanson (2015) acknowledge that it is complicated to separately measure the impact of the different types of monetary policy since the announcements have often coincided. Moreover, an LSAP announcement is sometimes a forward guidance announcement as well. However, the distinct characters of the policy measures make it a worthwhile endeavor.

The overview of the existing empirical stock of literature on the relationship between monetary policy and the stock market brings several things to light. In general, both conventional and unconventional monetary policy measures of an expansionary nature lead to a positive impact on the stock market. In spite of the plethora of research on the matter, the majority of the papers on unconventional monetary policy do not seem to make a distinction between the different types of policy measures. The papers that have done so do not break the impact of APPs further down as is proposed in this paper. In addition, the papers in question are only concerned with the effects of monetary policy at announcement, whereas according to the theory the liquidity effect or reserve-induced portfolio balance channel should materialize at implementation. In particular, the liquidity effect is concerned with the role of reserves rather than the impact of monetary policy on other variables that might influence the stock market. Thus, when endeavoring to test the liquidity effect, a methodology different from the prevailing methodologies in the existing stock of literature on the relationship between monetary policy and the stock market has to be developed. The methodology that will be used to measure the liquidity effect is inspired by the work of Krogstrup et al. (2012) and will be presented in the fifth section of this paper.

5. Methodology

(23)

23

additional adjustment that will be applied to all regressions is discussed. This adjustment allows measuring the impact of the amount of reserves per separate round of asset purchase program. In the third and final subsection, the hypotheses are discussed.

5.1 Measuring the liquidity effect

In section 3.2, it is already stipulated that Krogstrup et al. (2012) have come up with a methodology to measure the liquidity effect with respect to the yields of 10-year Treasury bonds. They make use of a time-series approach which is built upon earlier work by Gagnon, Raskin, Remache, and Sack (2011a, 2011b). In this paper, a similar time-series approach will be used as well. However, rather than looking at the impact of the liquidity effect on the yields of 10-year Treasury bonds, we are interested in the effect on the returns of the relevant stock indices. A regression similar to the one specified by Krogstrup et al. (2012) should enable us to measure the liquidity effect concerning the stock market.

Firstly, to do so, the dependent variable will have to be changed to the returns on the stock indices. In principle, the regression will be the same for the separate stock indices as we attempt to measure the same effect for every index. However, based upon country and time-specific characteristics, adjustments in the regression will be made if necessary. Nevertheless, the baseline specification of the regression is as follows:

∆𝑦𝑖𝑡 = 𝛼𝑖 + 𝛽𝑖1∆𝑅𝑖𝑡 + 𝛾𝑖𝑋𝑡 + 𝜀𝑖𝑡 (1)

In equation (1), ∆𝑦𝑖𝑡 denotes either the monthly or weekly change in the returns in country i at

time t of the stock index transformed into logs. The variable ∆𝑅𝑖𝑡 indicates the log-transformed change in the amount of reserves held by the central bank of country i at time t as a result of the particular APP, either on a monthly or weekly basis. The variable 𝜀𝑖𝑡 represents the error term of country i at time t and 𝛼𝑖 represents the constant. The coefficient 𝛽𝑖1 indicates for

country i the effect of a change in the reserves resulting from APPs on the returns of the stock index.

The vector 𝑋𝑡 comprises of several control variables that are expected to affect the returns of

(24)

24

are included. In addition, to control for circumstances that are not captured by the other control variables, a zero lower bound dummy will be added. This dummy takes a value of one during the entire zero lower bound period and zero otherwise. The duration of this period for each country is specified in the following subsections.

Although in this paper we are concerned with the liquidity effect, the stock market is likely still affected via other monetary policy channels as well. Therefore, for reasons of robustness, to control for the announcement effects of monetary policy measures, event dummy variables will be included. The monetary policy announcement dummy captures every important announcement related to APPs. Since the announcement effects do not have to be limited to country borders, both domestic and foreign monetary policy announcement dummies are constructed, which is done as follows. Every month or week in which the central bank has made an important announcement related to APPs, the dummy variable takes a value of one and zero otherwise. Appendix A contains tables with the relevant announcements for each central bank. This process is performed for each central bank, which results in four separate monetary policy announcement dummies.

Finally, a remark has to be made concerning the use of the amount of reserves in this paper as the independent variable. Namely, the capacity to hold reserves is limited to commercial banks and the central bank itself. However, the findings presented in section 4.1 show that the central banks have purchased assets from other parties than commercial banks as well. Therefore, the the amount of reserves serves as a proxy for the liquidity that ends up in the hands of the investors resulting from the asset purchase programs.

Moreover, it should be noted that mutations in the amount of reserves do not exclusively follow from LSAPs. The amount of reserves may alter for other reasons as well. This could invalidate the choice to use the amount of reserves as a proxy for the size of LSAPs. The balance sheet items of the central banks are carefully selected to reduce this issue to a minimum. In the following subsections, the regression and the time period for every country are presented and motivated.

5.1.1 The United States

(25)

25

of the regression by Krogstrup et al. (2012). In this way, a different effect following from the amount of reserves between the two distinct periods is possible. This leads to the following specification of the regression for the United States:

∆𝑦𝑡= 𝛼 + 𝛽1𝐷𝑡𝑛𝑜𝑛−𝑍𝐿𝐵∆𝑅𝑡+ 𝛽2𝐷𝑡𝑍𝐿𝐵∆𝑅𝑡−1+ 𝛾𝑋𝑡 + 𝜀𝑡 (2)

Other than the inclusion of the dummy variables related to the zero lower bound no further adjustments are required. Obviously, the zero lower bound period has to be defined. Again, we will follow Krogstrup et al. (2012) who indicate that this started in mid-December 2008. In this paper, the end of the zero lower bound is defined as the moment where the federal funds rate equals or surpasses the most recent lever prior to the zero lower bound. This means that in this paper July 2017 marks the end of the zero lower bound. Concerning the time span of the monthly frequency data, the same starting point as Krogstrup et al. (2012) is adopted, which is February 1990. Due to the different moments of writing, we are able to extend the time period compared to Krogstrup et al. (2012). In this paper, the last period included will be January 2018. The data sample of weekly frequency runs from March 1990 up to April 2018.

5.1.2 Euro Area

Similarly to the Fed and the BoE, when the policy rate had reached the zero lower bound, the ECB started to implement unconventional monetary policy measures. Therefore, as for the United States, in the Euro Area we will control for the period before and after the policy rate has hit the zero lower bound. For the Euro area, the regression looks as follows:

∆𝑦𝑡= 𝛼 + 𝛽1𝐷𝑡𝑝𝑟𝑒−𝑍𝐿𝐵∆𝑅𝑡+ 𝛽2𝐷𝑡𝑍𝐿𝐵∆𝑅

𝑡−1+ 𝛾𝑋𝑡 + 𝜀𝑡 (3)

Other than that, no further adjustments are considered to be necessary. The zero lower bound period is assumed to run from June 2014 up until the moment of writing. It has to be noted that the ECB came into existence in June 1998 and did not become operable until January 1999. Moreover, since the first difference is used of the reserves, the starting point is February 1999 for the monthly data. In case of the weekly frequency, the starting point is January 8 in 1999. The ending point of the sample is January 2018 for the monthly frequency and April 2018 for the weekly data.

5.1.3 The United Kingdom

(26)

26

area. Thus, as for regression (2) and (3), a pre-zero lower bound and zero lower bound dummy is included. The regression for the United Kingdom looks as follows:

∆𝑦𝑡= 𝛼 + 𝛽1𝐷𝑡𝑝𝑟𝑒−𝑍𝐿𝐵∆𝑅𝑡+ 𝛽2𝐷𝑡𝑍𝐿𝐵∆𝑅𝑡−1+ 𝛾𝑋𝑡 + 𝜀𝑡 (4)

Since the Bank of England lowered the Bank rate to its effective lower bound in March 2009, that moment will be used as the starting point of the zero lower bound. At the moment of writing, we still assume the Bank rate to be at the effective lower bound. Due to limited data availability on reserves and the VIX, the time span will be from March 1996 up until December 2017 using monthly data. For the weekly frequency data, the sample runs from December 1996

up to December 2017. The regressions are conducted for both frequencies.

5.1.4 Japan

In this paper, Japan stands out since it already has implemented Quantitative Easing at the beginning of this century. Moreover, there have been two distinct rounds; at the beginning of the century and the start of ‘Abenomics’. Similarly, the relevant domestic and foreign event dummy’s will be added to the regression. Moreover, the zero lower bound dummy will be separately incorporated as well. This leads to the following regression:

∆𝑦𝑡= 𝛼 + 𝛽1𝐷𝑡𝑛𝑜𝑛−𝑍𝐿𝐵∆𝑅𝑡+ 𝛽2𝐷𝑡𝑍𝐿𝐵∆𝑅𝑡−1+ 𝛾𝑋𝑡 + 𝜀𝑡 (5)

In the case of Japan, it has hit the zero lower bound already in the spring of 1999 and remained at that level up until 2006. In October 2008, the BoJ announced that it would lower its policy rate again. The rate has remained relatively flat up until now. For Japan, the period will run from February 1998 up to January 2018. The later starting point is motivated by limited data availability of Japan’s policy rate. For the BoJ holds that they only publish balance sheet figures on a monthly basis, which limits the data sample to be available on a monthly frequency.

5.2 Measuring the liquidity effect per asset purchase program

(27)

27

take a value of one during the program and zero otherwise. An additional dummy variable is constructed when no asset purchase program was in place.

Concerning equations (2) to (5), this means that 𝛽1𝐷𝑡𝑝𝑟𝑒 𝑍𝐿𝐵∆𝑅𝑡 or 𝛽1𝐷𝑡𝑛𝑜𝑛− 𝑍𝐿𝐵∆𝑅𝑡 and

𝛽2𝐷𝑡𝑍𝐿𝐵∆𝑅

𝑡 will be dropped. Instead, the change in the amount of reserves multiplied with

either a QE or a non-QE dummy will be added to the regressions. For each round of QE, a separate dummy will be constructed. In this way, it is possible to measure whether the relationship between the amount of reserves and the stock market differs per distinct asset purchase program. In case of the EU, a separate zero lower bound non-QE dummy will be included. This contrasts with the methodology for the other countries. This is explained by the fact that for a relatively long period that the ECB’s policy rate was at the effective lower bound, while it engaged in asset purchase programs which cannot be classified as QE. Therefore, in the case of the Euro Area, an additional control variable is included to allow for a difference in effect between the pre-zero lower bound and the zero lower bound.

5.3 Hypotheses

(28)

28

If the investment strategies can differ over time and between types of investors, this implies that differences across countries might exist as well. Namely, the counterparties from whom the central bank purchases can vary in composition. For example, Saito and Hogen (2014) indicate as well that some investors (e.g., insurance companies and pension funds) preferred to hold on the government bonds, whereas domestic banks were more willing to sell to the central bank. If a particular type of investors sells relatively more to the central bank in one country than another, this might result in a different effect on the stock market (dependent on investor preferences).

All in all, for various reasons the effect of reserves on the stock market is expected to differ both over time and across countries. That is to say, in some occasions the liquidity effect might be at work, while in other occasions this is not the case. Nonetheless, the liquidity effect entails that a positive relationship is expected between the amount of reserves and the stock market. For every country, the same regression is used to perform multiple regressions. First of all, a regression will be performed covering the entire sample. The use of pre- or non-zero lower bound and zero lower bound dummies allows for accounting for the difference in effects between the two periods.

6. Data

In this section, the data that will be used for the empirical analyses is discussed. The first subsection will elaborate upon the data in general. In the subsection thereafter, the descriptive statistics will be discussed in more detail. The final subsection will address the relevant statistical assumptions.

6.1 General data descriptions

(29)

29

The monthly or weekly returns of the stock index represents the dependent varaiable for each area. The stock indices used in this paper are the S&P 500, EUROSTOXX50, NIKKEI225 and the FTSE100 for respectively the United States, the Euro Area, Japan and the United Kingdom. In case of the United States, the S&P500 is preferred over other stock indices as it is considered to be the most representative stock index. The EUROSTOXX50 is used in case of the Euro Area. This index consists of what are considered to be the fifty most important shares of the Euro Area. In a similar vein, The FTSE100 is the national equivalent of the EUROSTOXX50 for the United Kingdom, only consisting of hundred stocks rather than fifty. For that reason, that index is used for the United Kingdom. The choice for NIKKEI225 is motivated by the fact that it is the leading and most representative index of Japan. The data of the stock indices of the United States, the Euro Area, and Japan is retrieved from Yahoo!Finance and that of the United Kingdom is obtained via Datastream. The data was available in daily frequency. However, this paper uses both the monthly and the weekly (apart from Japan) change in the stock index. For all indices holds that the daily data was matched with either the weekly or monthly announcement dates of balance sheet statements of the central banks.

(30)

30

Base,' which is motivated by an official statement11 on asset purchases by the BoJ. The BoJ

publishes data on this balance sheet item only on a monthly basis. In the absence of monthly publications, from the weekly data the last announcement in a particular month was used for the following month (e.g., an announcement on the 27th of a particular month was used as the level of reserves for the month thereafter).

The policy rates of the four central banks are retrieved from various sources. In case of the Fed, the ECB and the BoE, they originate from the database of BIS. The data on the respective policy rates has been based on different rates over time. For the United Kingdom, this means that the data on the policy rate comprises three different rates12. For the ECB holds that up to January 2000 and after October 2008, the fixed rate for liquidity provision is used. In between these periods, the minimum bid rate for liquidity provision is used. In case of the Fed, for the entire period, the mid-point of the Federal Funds Target Rate (FFTR) has been used. Data on the policy rate is retrieved from the database of the Bank of Japan, which is the Uncollateralized Overnight Call Rate (UOCR).

The inflation rates for the four areas are retrieved from the database of BIS as well. These are available at a monthly frequency. This gives rise to an issue since the analyses will not only be performed with a monthly frequency, but with a weekly frequency (apart from Japan) as well. This means that the inflation rate has to be converted. The conversion has been done in this paper by linearly interpolating the monthly inflation rates.

The data on the VIX in case of the S&P500 is retrieved from the database of the Federal Reserve. For the other three indices, the relevant VIX-indicators are gathered via Datastream. The data was available in a daily frequency and to a limited extent in a monthly frequency as well. By matching the dates of the week to the daily dates, the data has been converted to a weekly frequency. Concerning the monthly frequency, for the months where data was missing, the first day of the month indicated in the daily data set was used.

The data on the unemployment rate is gathered via Datastream for all four countries and was available on a monthly frequency. In case of the United Kingdom, unemployment rates are published at the end of each month. For the other three areas holds that the rates are published

11. The official statement by the BoJ can be found here:

https://www.boj.or.jp/en/announcements/release_2013/k130404a.pdf

12 The three different rates are respectively the Bank of England band 1 dealing rate, the repo rate and the Bank

(31)

31

in the mid of each month. Either way, the unemployment rate used in a particular month was used as the rate belonging to the following month. Unemployment has been linearly interpolated, as for inflation, to use the data on a weekly basis.

As indicated in the methodology, announcement dummies are used to control for announcement effects of the asset purchase programs. The relevant events for every individual country are listed in Appendix A. The relevant data is gathered from overviews presented in previous papers and where necessary expanded by retrieving additional information from central bank press releases and statements available on the central bank websites.

6.2 Descriptive statistics

The descriptive statistics are provided in Appendix B in Tables B1-B4 for the monthly frequency, and Tables B5-B7 for the weekly frequency. Since the zero lower bound marks the start of the period in which the central banks decided to introduce more unconventional monetary policy measures, the descriptive statistics of the pre- or non-period and the ZLB-period have been incorporated separately.

(32)

32

Furthermore, when looking at the stock indices, it is striking that for the total period the average value of the EUROSTOXX50 has a negative sign based on a weekly frequency and extremely close to zero based on a monthly frequency. However, the decomposition into two periods shows that in the period since the effective lower bound, the average return is positive instead of negative for the weekly data. In addition, for the monthly data holds that the average return increases severely. A similar development is visible for the United Kingdom when looking at the mean returns of the FTSE100. Before the zero lower bound the return was below or extremely close to zero, dependent on the frequency of the data. During the zero lower bound period, the returns have become relatively high. When looking into the United States, it is clearly visible that the average return is far larger since the ZLB-period. The average return of the NIKKEI225 is relatively low for the entire sample range.

Concerning the reserves, the descriptive statistics show that the average amount of reserves is enormous. The mean values are substantially larger during the ZLB-period for all central banks. Naturally, this follows from the fact that the central banks engaged in the APPs after the policy rat reached the effective lower bound. Outside this period, the reserves take a significantly lower value on average.

Lastly, the descriptive statistics of the policy rates have to be discussed. In line with the development of the average amount of reserves, a stark difference is noticeable between the pre-ZLB and the ZLB-period. By definition, the average rates should take a value of zero or very close to zero once the respective policy rates are at the effective lower bound. The descriptive statistics show that this indeed is the case.

6.3 Assumptions

Referenties

GERELATEERDE DOCUMENTEN

The non-normal incidence of thin-film guided, in-plane unguided optical waves on straight, possibly composite slab waveguide facets is considered.. The quasi-analytical,

Although mobile devises like smartphones with GPS become increasingly important, roadside devices might remain the main source of information for traffic management, because

However, the non-serious trailer also (unintentionally) provided an imposed goal (i.e., ‘designed to calm and soothe’). Moreover, no significant differences were found between

Of the economic schools based on rational expectations only New Keynesian models, using a slow moving variable, show real effects on output after a monetary shock.. The New

• Figure D24: BVAR- Model with Sims-Zha (Normal Wishart) prior (euro area) Figure D1 is displayed on the next page... The blue line represents the posterior median responses. The

The coefficient γ is the main interest of this paper since it is the coefficient of the sum of real unit labor costs and bank lending rates and thus it will determine whether or

While simulations starting with ordered proteins at every intermediate distance between the free protein and the fibril generally lead to a monotonic free energy profile,

To investigate the mechanism of charge transport in more detail and to determine values of Ea and other transport parameters, we carried out temperature dependent J(V)