The effect of quantitative easing on stock market liquidity

(1)

The Effect of Quantitative Easing on Stock Market Liquidity

Stan van Ruiten

Thesis supervisor: R.J. Döttling Master Thesis

MSc Finance, Quantitative Finance

University of Amsterdam, Amsterdam Business School, Amsterdam, the Netherlands.

Corresponding author. Grundel 9, 2377 BK Oude-Wetering, The Netherlands, E-mail: stan.vanruiten@student.uva.nl

Abstract

In this paper I look at the effect of quantitative easing (QE) by the European Central Bank on stock market liquidity. Liquidity in terms of price impact and turnover ratio were shown to be significantly affected by QE. Looking at the full sample, turnover ratios show to improve whereas price impact becomes larger as QE increases. Relative spreads showed to improve as quantitative easing increases in eight out of thirteen countries yet decrease in one. Price impact in two countries was shown to decrease as a result of QE whereas for one the opposite was discovered. Using differences-in-differences estimation showed that QE decreased stock market liquidity for companies whose bonds were purchased under CSPP compared to companies who were not targeted.

(2)

Statement of Originality

This document is written by student Stan van Ruiten who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

1. Introduction

Quantitative easing (QE) by central banks across the globe has been increasingly present in monetary policies after the 2007 financial crisis with several consequences. In this paper I look closely at the effect of quantitative easing by the European Central Bank (ECB) on stock market liquidity within the Eurozone. In this paper I pay special attention to the different policies that collectively form the expanded asset purchase programme (APP) in order to explain differences in the effect on liquidity between different countries and companies.

Quantitative easing is a monetary policy through which the ECB purchases all sorts of assets to inject large amounts of money into the economy in order to guard a stable inflation rate of around 2 per cent (European Central Bank, 2015). It is these large amounts of money that need to be reinvested. I argue that with interest rates being at all-time lows, some of this money will be reinvested in the stock market and in doing so will increase stock market liquidity. Using stock data from Thomson Reuters Datastream and data on quantitative easing provided by the ECB I show that, for the full sample, QE has no effect on bid-ask spreads yet seems to increase price impact and turnover ratios. This means that liquidity for the sample as a whole improves in terms of turnover ratios, however worsens in terms of price impact. Also the results show QE to significantly improve liquidity for some countries in the sample individually.

A prior study by Mishra, Parikh and Spahr in 2017 established a link between stock market liquidity and QE as performed by the Federal Reserve. They used changes in commercial bank credit as instrumental variable to show that QE improves stock market liquidity as long as it was accompanied by increases in commercial bank lending. They argue that increases in bank credit allow for increased funding to broker/dealers and market makers who in turn contribute to increased stock market liquidity. Support for this argument was

(4)

already provided by Gertler and Kiyotaki (2010) who account the effect on stock market liquidity by the Fed largely to its ability to improve commercial bank liquidity and lending to stock market participants.

Although other central banks might have had the same objective as the Fed, their policies do differ. This difference provides a research opportunity concerning the effect of QE on stock market liquidity. The setup of the programme of the European Central Bank (ECB) allows for further investigation of the way in which QE might affect stock market liquidity. The ECB started a new set of longer-term refinancing operations (LTRO), worth around 1,000 billion euros, at the end of 2011 (Gros, Alcidi & Giovannini, 2012). This programme

specifically aims to encourage banks lending activities to boost economic activity, which is the exact channel through which Mishra, Parikh and Spahr (2017) try to explain changes in stock market liquidity. Not until March 2015 did the ECB start with their expanded asset purchase programme (APP), which is more comparable to QE as it is performed by the Fed. This specific setup allows for the LTRO programme to be entered as an independent variable into the regressions in order to see whether QE affects stock market liquidity even with such a policy aimed to improve commercial bank lending employed prior to the actual asset

purchasing. Using asset purchasing under the APP programme as main independent variable for QE will then allow me to uncover the effect of QE on stock market liquidity unrelated to the enhancement of commercial bank lending through the LTRO programme.

In order to give an even clearer picture on the effect of the different policies employed I will also look at the difference of the effect on stock market liquidity between companies whose bonds are bought by the ECB its corporate sector purchase programme (CSPP) and those whose are not. Employing this differences-in-differences strategy showed that QE improved liquidity more for targeted companies. The next section will further explain the precise setup of the policies employed by the European Central Bank.

(5)

1.1. ECB policies

Quantitative easing as it is performed by the European Central Bank can be broken down to multiple different programs which are collectively called the expanded asset purchase programme (APP) (European Central Bank, 2018c). APP consists of four separate programmes that perform asset purchasing in different segments of the financial markets. By far the largest programme is called the public sector purchase programme (PSPP) which targets public sector securities like bonds issued by central governments, recognized agencies and international organisations (European Central Bank, 2015). In 2016 the ECB decided to also start purchasing investment-grade euro-denominated bonds within the eurozone. Central banks from six countries were appointed to execute this programme called the corporate sector purchase programme (CSPP), the appointed countries were Belgium, Germany, Spain, France, Italy and Finland. Under the programme bonds with credit assessments ranging from AA to BBB- were purchased, all with the aim to help inflation rates closer to 2 per cent (European Central Bank, 2016). Also there is the asset-backed securities purchase programme (ABSPP) which is much smaller but important as it aims to support the issuance of new securities (European Central Bank, 2014). Lastly there is the third covered bond purchase programme (CBPP3). As the name already indicates there have been two prior covered bond purchase programmes which both have been terminated, the first at the 30th of June 2010 when it reached a total amount of 60 billion euros (European Central Bank, 2010). The second programme started in November 2011, also being terminated a year later after it reached a nominal amount of 16.4 billion euros (European Central Bank, 2012). Both these programmes are to be dwarfed by APP as between March 2015 and March 2016 the combined asset

purchasing under PSPP, ABSPP and CBPP3 totalled around 60 billion euros each month (European Central Bank, 2015).

(6)

2. Literature Review

In 2017 Christensen and Gillan established a link between liquidity and QE by

showing that QE decreases the liquidity premium in the Treasury inflation-protected securities (TIPS) and inflation swaps market as QE added a significant buyer to these markets. However this temporary effect was only found on the securities targeted by the purchasing programme of the Fed. Chordia, Sarkar and Subrahmanyam (2004) provided broader insights on the effect of monetary policy on liquidity in the stock and bond market. In their analysis they include a ratio of net borrowed reserves over total reserves as a measure of monetary tightness or looseness, where lower values for this ratio indicate lower monetary tightness. The results showed that monetary expansion, in the form of less strict reserve requirements, has a positive effect on liquidity of stock markets in periods of financial crises. Also they researched the effect of unexpected changes in the federal funds rate. Higher than expected changes in the federal funds rate were shown to decrease stock and bond market liquidity and vice versa for lower than expected changes. The effect of QE itself on stock market liquidity was not uncovered until 2017 when Mishra, Parikh and Spahr showed that QE by the Fed increased liquidity through the improvement of commercial bank lending. Supported by findings of Gertler and Kiyotaki (2010) they argue that increased commercial bank lending to stock market participants like market makers and dealer/brokers cause increases in stock market liquidity.

Since liquidity has been shown to have a number of different aspects which are not being captured by any single measure (Chordia, Roll & Subrahmanyam, 2000), I will be using different measures for liquidity. In this paper a similar approach towards liquidity will be used as Luo did in 2016 in his paper on stock market liquidity between 1973 and 2015. Liquidity will be represented by three different measures: price impact (Amihud, 2002), turnover ratio and relative bid-ask spreads. Price impact is a measure which indicates the effect of an order

(7)

on the price change that follows. In highly liquid markets this effect is smaller than in illiquid markets. Turnover ratio is a ratio which indicates how easy or difficult it is to trade a

particular stock on the stock market, higher values for this ratio means it is easier and thus indicate higher liquidity. Lastly relative bid-ask spread which is a relative measure of the difference between the price for which you can buy and sell your shares, lower relative spreads indicate higher liquidity.

3. Methodology

To be able to analyse the effect of QE on stock market liquidity ordinary least squares regression is used. As main dependent variable the different measures of liquidity are used. The main independent variable in the regression to be tested is the amount of assets purchased under APP per month per country divided by the market capitalization of that specific

country. This measure is a better representation of QE in our analysis as it accounts for the fact that an additional billion Euros invested in the German stock market might not have the same effect as that same billion Euros in the stock market of Lithuania.

Furthermore the model will be run with and without control variables which have been found to affect liquidity such as price, trading volume (Hasbrouck, 1991) and volatility

(Benston & Hagerman, 1974). Besides these a variable for the LTRO is included as well. Finally the model will also be run with and without time and firm fixed effects to correct for any factors that may influence liquidity within these factors. Also the regressions are

performed on subsamples of each country individually to account for any possible differences between different countries leading to the following model that is tested:

(1) 𝑆𝑡𝑜𝑐𝑘 𝑀𝑎𝑟𝑘𝑒𝑡 𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦_𝑓𝑡 = 𝛼_𝑡+ 𝛿_𝑓 + 𝛽₁𝑄𝐸_𝑐𝑡+ 𝛽₂𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 + 𝜀

Besides running model 1 a differences-in-differences estimation is performed as well. Creating a dummy variable for the firms whose bonds are bought under CSPP allows me to

(8)

analyse any potential differences of the effect of QE on stock market liquidity between targeted and non-targeted firms. Table 3 in appendix B shows that firms targeted by CSPP and firms that are not targeted by CSPP have significant differences in all liquidity measures and firm characteristics except for price. Relative spreads and price impact are lower for the targeted group indicating higher liquidity, whereas the turnover ratios indicate lower liquidity for the targeted group of firms. To see whether the effect of QE on stock market liquidity of the targeted firms is different I run the following model:

(2) 𝑆𝑡𝑜𝑐𝑘 𝑀𝑎𝑟𝑘𝑒𝑡 𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦_𝑓𝑡 =

𝛽1𝑄𝐸𝑐𝑡+ 𝛽2𝐶𝑆𝑃𝑃𝑑𝑢𝑚𝑚𝑦𝑓𝑡 + 𝛽3𝑄𝐸𝑐𝑡∗ 𝐶𝑆𝑃𝑃𝑑𝑢𝑚𝑚𝑦𝑓𝑡 + 𝛽4𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 + 𝜀

4. Data

Data for its policies are provided by the ECB and is used as a measure for QE in this study. First data for the public sector purchase programme (PSPP) is retrieved which consists of the monthly asset purchases per country under the programme. Unfortunately the data for the asset-backed securities purchase programme (ABSPP) (European Central Bank, 2018e) and the third covered bond purchase programme (CBPP3) (European Central Bank, 2018d) does not provide any information on the level of purchasing by each individual country. However since the amount of asset purchasing under PSPP is much larger than that of ABSPP and CBPP3 I will assume that the share of purchasing per country is similar to that under PSPP. Asset purchasing under CSPP is divided under central banks based in six different countries, Belgium, Germany, France, Spain, Finland and Italy (European Central Bank, 2016) The ECB provides information on which securities have been purchased under CSPP (European Central Bank, 2018f), allowing me to construct a dummy variable for all

companies whose bonds have been bought under CSPP in order to run a differences-in-differences estimation. Finally adding up the asset purchasing under these four programmes

(9)

gives the amount of asset purchasing per country under the expended asset purchase

programme (APP) on a monthly interval which will be used as main independent variable for QE. Greece is excluded from the sample since Greece is not included in this programme. Hereafter data for the longer term refinancing operations (LTRO) is collected, which also is provided by the ECB at a monthly interval (European Central Bank, 2018b). This data is used to construct a control variable for the effect of the ECB on stock market liquidity through commercial bank lending that was found by Mishra, Parikh and Spahr (2017).

To look at any given effect on liquidity it is of course crucial to construct a measure for liquidity. First of all price impact as it has been discussed by Amihud (2002). This measure in particular is fairly easy to construct by dividing absolute stock return by the Euro volume traded. Both data on stock returns and Euro volume traded are retrieved at a weekly frequency from Thomson Reuters Datastream after which the measure price impact is

computed. Hereafter the variables turnover ratio and relative bid-ask are computed also using Thomson Reuters Datastream to collect data for turnover value, market capitalization and bid-ask spreads. To account for outliers in the data all ratios are winsorized. Price impact, relative bid-ask spread and turnover ratio at a 1, 0.5 and 0.1 per cent level respectively. After

computing all variables of importance the data is converted to monthly data to suit the data for the QE and LTRO programmes. Summary statistics for the liquidity measures are provided in the Appendix (See Appendix A, Tables 1 & 2). There is some variation in the number of observations for each country which is due to the following selection criteria. When retrieving the data only firms are included which are included in the leading stock market index for the country in which the firm is listed. Penny stocks are excluded to prevent the data from containing too many outliers. Also the firm must have been registered on a stock market for the entire observed period which ranges from January 2000 up to and including December 2017 and provide data for the necessary variables. This data is retrieved for all countries that

(10)

are part of the eurozone and fall under the regulation of the ECB except for Cyprus, Slovenia, Estonia and Malta which did not have more than five firms meeting these conditions.

To prevent biased results the hypotheses was tested using a measure for quantitative easing that is corrected for the market capitalization of the companies in that country. This measure is computed by dividing the amount of asset purchasing under QE per country by the market capitalization of the companies for that country. The reason for this decision being the data showing a high correlation of 0.8134 between the market capitalization of the stocks included for a country and the amount of assets purchased under QE in that country. Figure 1 shows the relative bid-ask spreads for the countries with the highest (France & Germany) and lowest (Latvia & Lithuania) market capitalization.It indicates market capitalization to be related to the relative spreads of companies, yet also being highly correlated to the amount of asset purchasing under QE.

Figure 1. Average Relative bid-ask Spread

Also the control variables are computed using data from Thomson Reuters Datastream. Both price and turnover volume are provided in the data and volatility is

(11)

the number of workdays in that year after which I take the square root of that number in its entirety.

5. Findings

In a first attempt the model was run on all three liquidity measures without control variables, time and firm fixed effects for the full sample. Also additional regressions were run with control variables and fixed effects. Table 1 below shows the results for the liquidity measure turnover ratio. QE does seem to have an effect on turnover ratios for all

specifications of the model. When control variables are included the results show that an increase in QE equal to one per cent of the country its market capitalization increases turnover ratios by around 0.00035, which for an average ratio of 0.0217 is an increase of 1.6 per cent.

Table 1. Regression Results for Turnover Ratio

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Running the same model for the measure price impact gives very different results in terms of economic effect. Where liquidity in terms of turnover ratios seem to improve as QE

(1) Turnover Ratio (2) Turnover Ratio (3) Turnover Ratio (4) Turnover Ratio Relative QE -0.0777*** 0.0416*** 0.0342*** 0.0355*** (0.0243) (0.0147) (0.0124) (0.0127) LTRO -0.000006** -0.000007** (0.000002) (0.0000026) Price -0.000003* -0.000003* (0.000001) (0.000001) Volatility 0.0379*** 0.0382*** (0.0089) (0.0089) Turnover Volume 0.189*** 0.191*** (0.0706) (0.0710) Constant 0.0216*** 0.0228*** 0.0222*** 0.0228*** (0.0017) (0.0033) (0.0032) (0.0032) Observations 41,819 41,819 39,320 39,320 Company FE NO NO NO YES

(12)

increases the opposite goes for liquidity in terms of price impact. Table 2 shows that an increase in QE increases price impact indicating lower liquidity. An increase in QE equal to one per cent of a country its market capitalization increases price impact by 11.7 per cent of the sample average in specification 4. When running the model for relative bid-ask spreads the results showed to be insignificant (See Appendix C, Table 4)

Table 2. Regression Results for Price Impact

Running the model on subsamples for all countries provided results which did in fact show that there were differences in the effect of QE on stock market liquidity between countries. Table 3 shows the results for the model with as dependent variable relative bid-ask spread for the separate countries with significant results. In six countries quantitative easing shows to decrease the relative bid-ask spread. An increase in asset purchasing equal to one percent of the country’s market capitalization is shown to decrease the relative spread by 0.31 percentage points in Belgium ranging up to 1.654 percentage points in Ireland. Lithuania is the only country for which the relative spreads increase as quantitative easing increases. The

(1) Price Impact (2) Price Impact (3) Price Impact (4) Price Impact Relative QE 0.000799 0.00155*** 0.00144*** 0.000923* (0.000503) (0.000581) (0.000534) (0.000541)

LTRO 4.93e-08 4.42e-08

(4.35e-08) (3.92e-08)

Price -7.29e-08* -7.91e-08*

(3.85e-08) (4.18e-08)

Volatility 0.00029*** 0.00018*

(0.000114) (9.17e-05)

Turnover Volume -0.00014*** -8.87e-06

(5.02e-05) (1.70e-05) Constant 0.000102*** 7.67e-05*** 7.08e-05*** 6.48e-05***

(2.08e-05) (1.66e-05) (1.48e-05) (5.47e-06)

Observations 41,532 41,532 39,036 39,036

Company FE NO NO NO YES

(13)

results for the LTRO programme do not show any significant results. One explanation for these results could be that the data which the ECB provides on their LTRO programme does not provide information on the countries to which the refinancing is offered.

Table 3. Regression Results for Relative Bid-Ask Spread

The model is estimated for Turnover ratio as well, which tries to capture how easy or hard it is to sell stocks of a certain company. Surprisingly the results show that QE has an effect on four countries, all of which show that the effect is in fact negative meaning that as QE as a percentage of market cap increases it becomes harder to sell shares, indicating lower liquidity which is opposite to the results for the full sample (See Appendix C, Table 5). Striking is the fact that Spain, Germany and France, whose relative spreads had shown to decrease because of QE, are now showing that their turnover ratios are actually decreasing because of it. Lastly the liquidity measure price impact is entered into the model to get a full

Sample Country

Lithuania Spain Germany Belgium Finland France Ireland Italy Relative QE 0.121*** -0.391*** -0.430*** -0.310** -1.277*** -0.413** -1.654*** -0.199***

(0.0316) (0.0906) (0.103) (0.116) (0.376) (0.161) (0.533) (0.0419) LTRO -1.6e-08 -9.5e-10 2.6e-10 -8.7e-10 -1.2e-09 -6.8e-10 8.5e-11 4.1e-09

(9.1e-09) (1.2e-09) (1.2e-09) (7.9e-10) (7.2e-10) (4.6e-10) (1.5e-08) (4.1e-09) Price -0.000970 -0.00001*** -0.00002*** -0.00002** -0.0001** -0.00002 -0.000006*** -0.00002 (0.00108) (0.0000008) (0.000002) (0.000008) (0.00004) (0.00001) (0.0000001) (0.00001) Volatility 0.0265 0.0109 0.0158*** 0.0032 -0.0085 0.0116 0.0289** 0.0018 (0.028) (0.0071) (0.0049) (0.0022) (0.0075) (0.0105) (0.0103) (0.0038) Turnover Volume -5.145 (3.538) -0.0043 (0.0032) -0.943** (0.405) 0.0097*** (0.0027) 0.0075** (0.0034) -0.103 (0.0932) -0.0802 (0.0687) -0.0011 (0.0008) Constant 0.0016 0.0067*** 0.0072*** 0.0054*** 0.0144*** 0.0055*** 0.0241*** 0.0065*** (0.0079) (0.0013) (0.0008) (0.001) (0.0028) (0.0016) (0.0046) (0.0011) Observations 784 3,830 5,075 2,233 3,846 6,273 3,043 3,512 R-squared 0.275 0.577 0.381 0.354 0.495 0.160 0.190 0.488

Company FE YES YES YES YES YES YES YES YES

Time FE YES YES YES YES YES YES YES YES

(14)

picture of the effects of QE on these different types of liquidity. QE seems to improve liquidity in Lithuania and Italy, whereas liquidity indicated by price impact worsens in Germany (See Appendix C, Table 6).

After running model 1 on the three liquidity measures nine countries provided significant results for QE of which six are responsible for the asset purchasing under CSPP. To investigate the role of purchases under CSPP a differences-in-differences estimation is performed. The ECB provides information and data on the bonds that are purchased under the programme (European Central Bank, 2018f). This data is used to construct a dummy variable which equals one for companies whose bonds have been purchased under CSPP since the start of the programme and zero otherwise. Running the estimation on the entire sample provides the results shown in Table 4.

Table 4. Differences-In-Differences Estimation

Relative Spread Turnover Ratio Price Impact

Relative QE 0.00079 0.0318*** 0.00095* (0.0151) (0.0121) (0.00055) LTRO 0.0000002 0.0000006* -6.83e-08 (0.000001) (0.000003) (4.44e-08) Price -0.000007*** -0.000003* -7.95e-08* (0.000001) (0.000001) (4.21e-08) Volatility 0.0121*** 0.0379*** 0.000178* (0.0037) (0.0088) (0.000091) Turnover Volume -0.0014 0.191*** -0.000007 (0.0015) (0.071) (0.000017) CSPP-Dummy 0.0026*** -0.0067* 0.000046** (0.0006) (0.0035) (0.000019) QE/CSPP Interaction -0.0563*** 0.278*** -0.00107** (0.0141) (0.0853) (0.000481) Constant 0.0085*** 0.0225*** 0.000067*** (0.0007) (0.0032) (0.0000056) Observations 38,398 39,320 39,036 R-squared 0.082 0.164 0.015

Company FE YES YES YES

(15)

First of all the results show that QE expressed as a percentage of market capitalization has no effect on relative bid-ask spreads in general, however there is an effect on Turnover ratios and price impact. Turnover ratios seem to increase as QE increases indicating improved liquidity, whereas the effect on price impact indicates lower liquidity. All of this is in line with the findings for the full sample regressions. However the main results are found in the row for the QE and CSPP-dummy interaction term. This variable captures the difference in the effect of QE on companies whose bonds are and are not purchased under CSPP. The results show to be significant for all three liquidity measures.

For companies whose bonds have been purchased under CSPP a one percentage point increase in asset purchasing as a percentage of market capitalization decreases relative spreads by 0.056 percentage points compared to non-targeted firms. This effect might seem small, however considering the average relative spread in this sample is 0.598 per cent means a decrease of 9.36 per cent. A one percentage point increase in QE also improves liquidity in terms of turnover ratio. This one percentage point increase leads to an increase of 0.00278 compared to companies whose bonds have not been purchased under CSPP, which for an average turnover ratio of 0.0217 is slightly larger than 12 per cent. Lastly price impact seems to improve as well, the same one percentage point increase is shown to lead to a decrease in price impact equal to 13.5 per cent of the sample average. Different specifications of the model can be found in appendix C Tables 7-9.

5.1. Robustness Checks

To ensure the robustness of the results for different time periods I performed the main estimation for smaller time windows whilst still including the full period in which QE was performed. The results for the measures price impact and relative spread show to remain similar in economic magnitude and statistical significance when I decrease the sample period to 2008-17 (See appendix D, Tables 10-17). For the measure turnover ratio the interaction

(16)

term between the CSPP dummy variable and QE variable shows to be no longer significant after the sample is decreased to 2004 up to 2017 (See Appendix D, Table 13). This means that for this particular sample the effect of QE on turnover ratios is not significantly different for companies which are targeted by CSPP compared to those that are not.

6. Conclusion

This paper looked at the effect of the quantitating easing as performed by the

European Central Bank on stock market liquidity. When running model 1 on the full sample I found that quantitative easing improved liquidity in terms of turnover ratios however

worsened it in terms of price impact. For the liquidity measure relative bid-ask spread no significant results were found.

Looking more closely at the effect on each country separately I found some

contradicting results. For nine out of thirteen countries significant results were reported for at least one liquidity measure. Relative spreads were shown to decrease as quantitative easing increases for six countries. The only exception was Lithuania which showed a positive effect of QE on relative spreads. For the liquidity measure turnover ratio four countries reported significant results, all indicating that QE had a negative effect meaning liquidity decreases. The results for price impact showed to be mixed looking at the economic effect. The price impact for Lithuania and Italy showed to decrease as QE increases, indicating higher liquidity. However the results for Germany were contradicting as it showed an increase of price impact. This shows that QE affects liquidity differently for different countries, which in turn could be an explanation as to why the results differ from the findings for the full sample regressions.

Though the economic effect indicated by the coefficients were contradicting at times a pattern can be found looking at the specific countries that reported significant results. Six out of nine countries that reported significant results are responsible for the asset purchasing

(17)

under CSPP. These countries being Belgium, France, Finland, Italy, Spain and Germany. This result gave additional reasons to have a closer look at the CSPP and its relation towards stock market liquidity. Since the ECB provides data on the exact bonds that are purchased under CSPP it was possible to construct a dummy variable for the companies which had been targeted by CSPP and perform a differences-in-differences estimation (European Central Bank, 2018f).

It is through this estimation that I found QE to have an effect for all liquidity measures for companies that have been targeted by CSPP compared to companies that have not. The estimation on the full sample showed that a one per cent increase of QE as a percentage of that country its market capitalization decreases relative spreads by 0.056 percentage points for companies targeted by CSPP. Also the same increase in QE increases turnover ratios by 0.00278 and decreases price impact by 0.00107, which means increased liquidity for all measures by 9.36, 12.8 and 13.5 per cent respectively.

Now the question arises as to why there is such a significant increase in the liquidity of stocks for companies whose bonds have been purchased under CSPP. One possible

explanation could be an increase in the amount of trust investors have in a company. As the ECB decides to purchase bonds from a company these companies have to meet all

requirements set by the ECB itself. As investors now know that these companies meet the requirements of the ECB they might have increased trust in the ability of the company to repay its debt in the future and might see it as a safer investment than before. This in turn might attract additional investors in the stock simultaneously improving its liquidity.

Suggestions for future research are mainly focussed on the LTRO programme for which barely any significant results were found on liquidity. Yet previous literature describes channels through which this programme could potentially improve stock market liquidity (Gertler & Kiyotaki, 2010; Mishra, Parikh & Spahr, 2017). A possible explanation for this

(18)

result could be the fact that the data on LTRO by the ECB does not provide information on which banks and countries this funding is going to. Therefore any research providing more insights in the exact workings of the Longer Term Refinancing Operations will not be limited in its usefulness as an insight in itself but will allow future research to discover its effects on other economic factors as well.

(19)

7. References

Amihud, Y. (2002). Illiquidity and stock returns: cross-section and time-series effects. Journal of financial markets, 5(1), 31-56.

Benston, G. J., & Hagerman, R. L. (1974). Determinants of bid-asked spreads in the over-the-counter market. Journal of Financial Economics, 1(4), 353-364.

Chordia, T., Roll, R., & Subrahmanyam, A. (2000). Commonality in liquidity. Journal of

financial economics, 56(1), 3-28.

Chordia, T., Sarkar, A., & Subrahmanyam, A. (2004). An empirical analysis of stock and bond market liquidity. The Review of Financial Studies, 18(1), 85-129.

Christensen, J. H., & Gillan, J. M. (2017, September). Does quantitative easing affect market liquidity?. Federal Reserve Bank of San Francisco.

Datastream. (2018) Thomson Reuters Datastream. [Online]. Available at: Subscription Service (Accessed: March 2018)

European Central Bank (2010, June 40). Covered bond purchase programme completed [Press release]. Retrieved from

https://www.ecb.europa.eu/press/pr/date/2010/html/pr100630.en.html

European Central Bank (2012, October 31). Ending of covered bond purchase programme 2

(CBPP2) [Press release]. Retrieved from

https://www.ecb.europa.eu/press/pr/date/2012/html/pr121031_1.en.html

European Central Bank (2014). Decision of the European Central Bank of 19 November 2014 on the implementation of the asset-backed securities purchase programme

(20)

European Central Bank (2015). Decision (EU) 2015/774 of the European Central Bank of 4 March 2015 on a secondary markets public sector asset purchase programme

(ECB/2015/10).

European Central Bank (2016, April 21). ECB announces details of the corporate sector purchase programme (CSPP) [Press release]. Retrieved from

https://www.ecb.europa.eu/press/pr/date/2016/html/pr160421_1.en.html

European Central Bank. (2018a). History of cumulative purchase breakdowns under the PSPP [Data file]. Retrieved from

https://www.ecb.europa.eu/mopo/implement/omt/html/index.en.html

European Central Bank. (2018b). History of all ECB open market operations [Data file]. Retrieved from

https://www.ecb.europa.eu/mopo/implement/omo/html/top_history.en.html

European Central Bank. (2018c) History of cumulative purchases breakdowns under the CSPP [Data file]. Retrieved from

https://www.ecb.europa.eu/mopo/implement/omt/html/index.en.html#cspp

European Central Bank. (2018d) History of cumulative purchases breakdowns under the ABSPP [Data file]. Retrieved from

European Central Bank. (2018e) History of cumulative purchases breakdowns under the CBPP3 [Data file]. Retrieved from

European Central Bank. (2018f) List of securities held under the CSPP [Data file]. Retrieved from https://www.ecb.europa.eu/mopo/implement/omt/html/index.en.html#cspp

(21)

European Central Bank. (2018g). Asset purchase programmes. European Central Bank. Retrieved from https://www.ecb.europa.eu/mopo/implement/omt/html/index.en.html

Gertler, M., & Kiyotaki, N. (2010). Financial intermediation and credit policy in business cycle analysis. In Handbook of monetary economics (Vol. 3, pp. 547-599). Elsevier.

Gros, D., Alcidi, C., & Giovannini, A. (2012). Central banks in times of crisis: The FED vs. the ECB.

Hasbrouck, J. (1991). Measuring the information content of stock trades. The Journal of

Finance, 46(1), 179-207.

Krishnamurthy, A., & Vissing-Jorgensen, A. (2011). The effects of quantitative easing on

interest rates: channels and implications for policy (No. w17555). National Bureau of

Economic Research.

Luo, Y. (2016). A study of stock market liquidity from 1973 to 2015 (Doctoral dissertation, Massachusetts Institute of Technology).

Mishra, A. K., Parikh, B., & Spahr, R. W. (2017). Individual Stock Market Liquidity, Financial Crisis and Quantitative Easing.

(22)

8. Appendix Appendix A – Data

Table 1. Mean and Standard Deviation (in parentheses) of Liquidity Measures

Table 2. Mean and Standard Deviation (in parentheses) of Liquidity Measure Sample Country

Italy Latvia Lithuania Luxembourg Portugal Spain

The Netherlands Relative Spread 0.0041 0.0221 0.0188 0.0189 0.0067 0.0046 0.0017 (0.0084) (0.0194) (0.0199) (0.0199) (0.0086) (0.0093) (0.0026) Turnover Ratio 0.0347 0.0014 0.0047 0.0064 0.0121 0.0368 0.032 (0.0536) (0.0036) (0.0158) (0.0286) (0.0262) (0.0449) (0.0332) Price Impact 0.000001 0.0012 0.0007 0.00035 0.00006 0.000005 0.0000008 (0.000004) (0.0013) (0.0011) (0.0007) (0.00027) (0.00009) (0.000004) Observations 3768 917 1090 1302 2104 4120 4123 Sample Country

Austria Belgium Finland France Germany Ireland Relative Spread 0.0068 0.0031 0.0039 0.0015 0.0042 0.0192 (0.0122) (0.0044) (0.0072) (0.0042) (0.0045) (0.0204) Turnover Ratio 0.0136 0.0229 0.0313 0.0302 0.0009 0.0086 (0.0226) (0.0511) (0.0435) (0.0341) (0.0027) (0.0254) Price Impact 0.00003 0.000003 0.000002 0.000005 0.000026 0.00022 (0.00014) (0.000009) (0.000012) (0.000005) (0.00011) (0.00063) Observations 2585 2387 4112 6727 5425 3356

(23)

THE EFFECT OF QUANTITATIVE EASING ON STOCK MARKET LIQUIDITY 23

Appendix B – Methodology

Table 3. Mean, (Standard Deviation) and p-value of difference

p-value of Non-CSPP CSPP difference Relative Spread 0.00614 0.00159 0.0000*** (0.01174) (0.00234) Turnover Ratio 0.02183 0.01856 0.0011*** (0.03777) (0.02915) Price Impact 0.00008 0.000002 0.0000*** (0.00041) (0.000012) Volatility 0.10969 0.08485 0.0000*** (0.07946) (0.04946) Turnover Volume 19566.1 22826.4 0.0786* (70002) (54057.1) price 43.08 49.07 0.2669 (205.79) (47.88) Observations 40557 1459

(24)

Appendix C – Findings

Table 4. Regression Results for Relative Spread and Price Impact for Full Sample

Relative Spread Relative Spread Relative Spread Relative Spread

Relative QE -0.0206 0.0014 0.0045 -0.0005 (0.0145) (0.0151) (0.0146) (0.0151) LTRO 0.000002 0.000002 (0.0000018) (0.0000018) Price -0.0000065*** -0.000007*** (0.000001) (0.000001) Volatility 0.0144*** 0.0120*** (0.0039) (0.0037) Turnover Volume -0.0041* -0.0015 (0.0021) (0.0015) Constant 0.0069*** 0.0082*** 0.009*** 0.0085*** Relative QE (0.00062) (0.00082) (0.0008) (0.00065) Observations 40,900 40,900 38,398 38,398 Company FE NO NO NO YES

(25)

Table 5. Regression Results for Turnover Ratio Sample Country

Spain Germany Portugal France

Relative QE -2.988*** -0.432*** -1.776*** -3.308***

(0.872) (0.135) (0.397) (1.055)

LTRO 1.75e-08 4.83e-10 -5.61e-09 -1.02e-08

(1.39e-08) (3.17e-10) (3.19e-09) (7.37e-09)

Price -0.000038** -0.0000075* -0.00094** -0.0000918 (0.000013) (0.0000036) (0.00037) (0.000068) Volatility 0.0154 0.0071** 0.0432*** 0.0373 (0.0330) (0.0029) (0.0102) (0.0266) Turnover Volume 0.187*** 0.755* 0.241 1.081*** (0.0479) (0.388) (0.150) (0.242) Constant 0.0602*** 0.0047*** 0.0354*** 0.0349*** (0.0149) (0.0013) (0.0067) (0.0085) Observations 3,848 5,100 1,970 6,324 R-squared 0.280 0.340 0.381 0.323

Company FE YES YES YES YES

Time FE YES YES YES YES

(26)

Table 6. Regression Results for Price Impact Sample Country

Lithuania Germany Italy

Relative QE -0.0281*** 0.0015* -0.000065*

(0.00385) (0.00076) (0.000032)

LTRO -8.96e-10 5.76e-11* 0

(9.74e-10) (0) (0)

Price 0.00008* 4.96e-08 -1.18e-08

(0.000038) (6.99e-08) (7.37e-09) Volatility 0.0024** -0.000063 0.000004 (0.0010) (0.000056) (0.000003) Turnover Volume -0.755* 0.0015 0.000001 (0.357) (0.0023) (0.0000008) Constant 0.0036*** -0.000002 0.000001*** (0.00007) (0.000012) (0.0000003) Observations 902 5,075 3,520 R-squared 0.300 0.064 0.099

Time FE YES YES YES

(27)

Table 7. Regression Results Without Control Variables and Fixed Effects

Table 8. Regression Results Without Control Variables, With Time Fixed Effects

Relative QE -0.0192 -0.0695*** 0.000824 (0.0147) (0.0231) (0.000514) CSPP-Dummy -0.000491*** -0.00368*** -3.52e-06 (0.000186) (0.00132) (2.73e-06) QE/CSPP Interaction -0.0295* -0.184 -0.000763 (0.0158) (0.115) (0.000499) Constant 0.00698*** 0.0217*** 0.000103*** (0.000615) (0.00169) (2.08e-05) Observations 40,900 41,819 41,532 Company FE NO NO NO Time FE NO NO NO

Relative QE 0.00215 0.0391*** 0.00157*** (0.0151) (0.0142) (0.000583) CSPP-Dummy 0.00107*** -0.00434*** 1.47e-05 (0.000325) (0.00149) (1.04e-05) QE/CSPP Interaction -0.0221 0.157* -0.000967* (0.0163) (0.0928) (0.000575) Constant 0.00817*** 0.0228*** 7.64e-05*** (0.000815) (0.00331) (1.65e-05) Observations 40,900 41,819 41,532 Company FE NO NO NO

(28)

Table 9. Regression Results With Control Variables and Time Fixed Effects

Relative QE -0.0189 -0.0633*** 0.000706

(0.0145) (0.0206) (0.000471)

LTRO -5.00e-06*** 1.62e-07 4.25e-09

(1.69e-06) (4.90e-06) (5.25e-08)

Price -7.29e-06*** -4.34e-07 -8.73e-08*

(1.66e-06) (1.57e-06) (4.69e-08)

Volatility 0.0111*** 0.0283*** 0.000158***

(0.00191) (0.00478) (5.76e-05)

Turnover Volume -0.00412** 0.190*** -4.20e-05**

(0.00193) (0.0733) (1.75e-05) CSPP-Dummy -0.000312 -0.0168*** 2.72e-06 (0.000343) (0.00302) (5.15e-06) QE/CSPP Interaction -0.0280* 0.355*** -0.000746* (0.0156) (0.0835) (0.000449) Constant 0.00631*** 0.0146*** 9.02e-05*** (0.000569) (0.00143) (1.83e-05) Observations 38,398 39,320 39,036 Company FE NO NO NO

(29)

Appendix D – Robustness Checks Table 10. Regression Results for period 2001-17

Relative QE 0.000648 0.0307** 0.000953*

(0.0151) (0.0120) (0.000542)

LTRO 2.11e-06 -6.22e-06** 4.31e-08

(1.79e-06) (2.44e-06) (3.92e-08)

Price -6.87e-06*** -2.78e-06* -7.94e-08*

(1.27e-06) (1.52e-06) (4.17e-08)

Volatility 0.0132*** 0.0355*** 0.000155*

(0.00382) (0.00854) (8.16e-05)

Turnover Volume -0.00218 0.187*** -1.05e-05

(0.00159) (0.0703) (1.64e-05) CSPP-dummy 0.00263*** -0.00598* 4.33e-05** (0.000622) (0.00337) (1.93e-05) QE/CSPP interaction -0.0596*** 0.227*** -0.00104** (0.0144) (0.0849) (0.000476) Constant 0.00997*** 0.0240*** 7.24e-05*** (0.000705) (0.00318) (1.38e-05) Observations 36,751 37,341 37,232 R-squared 0.081 0.168 0.014

(30)

Table 11. Regression Results for period 2002-17

Relative QE 0.000261 0.0294** 0.000953*

(0.0151) (0.0120) (0.000541)

LTRO 2.16e-06 -6.37e-06*** 4.43e-08

(1.78e-06) (2.34e-06) (3.90e-08)

Price -6.77e-06*** -2.82e-06* -7.89e-08*

(1.21e-06) (1.47e-06) (4.10e-08)

Volatility 0.0120*** 0.0323*** 0.000130*

(0.00411) (0.00838) (7.68e-05)

(0.00158) (0.0691) (1.51e-05) CSPP-dummy 0.00247*** -0.00499 4.15e-05** (0.000586) (0.00328) (1.88e-05) QE/CSPP interaction -0.0578*** 0.179** -0.00103** (0.0139) (0.0849) (0.000475) Constant 0.0126*** 0.0113*** 0.000115*** (0.00129) (0.00187) (2.16e-05) Observations 34,820 35,317 35,216 R-squared 0.070 0.171 0.014

(31)

Relative QE 0.000404 0.0289** 0.000930*

(0.0151) (0.0120) (0.000547)

LTRO 2.15e-06 -6.18e-06*** 4.35e-08

(1.78e-06) (2.24e-06) (3.88e-08)

Price -6.66e-06*** -2.95e-06* -7.51e-08*

(1.14e-06) (1.50e-06) (3.92e-08)

Volatility 0.0116** 0.0319*** 0.000116

(0.00447) (0.00845) (8.02e-05)

(0.00142) (0.0697) (1.38e-05) CSPP-dummy 0.00235*** -0.00466 4.02e-05** (0.000544) (0.00325) (1.85e-05) QE/CSPP interaction -0.0519*** 0.157* -0.00104** (0.0134) (0.0880) (0.000475) Constant 0.00883*** 0.00996*** 0.000129*** (0.000539) (0.00141) (2.99e-05) Observations 32,678 33,092 32,999 R-squared 0.061 0.179 0.013

(32)

Relative QE 0.00166 0.0296** 0.000920*

(0.0148) (0.0115) (0.000556)

LTRO 2.14e-06 -5.82e-06*** 4.16e-08

(1.79e-06) (2.12e-06) (3.86e-08)

Price -6.33e-06*** -2.78e-06* -7.17e-08*

(1.07e-06) (1.47e-06) (3.81e-08)

Volatility 0.0127*** 0.0360*** 0.000112

(0.00490) (0.00921) (8.99e-05)

(0.00142) (0.0706) (1.41e-05) CSPP-dummy 0.00241*** -0.00404 3.80e-05** (0.000510) (0.00316) (1.84e-05) QE/CSPP interaction -0.0580*** 0.123 -0.00103** (0.0129) (0.0900) (0.000484) Constant 0.00819*** 0.0129*** 0.000131*** (0.000777) (0.00198) (2.83e-05) Observations 30,485 30,834 30,754 R-squared 0.057 0.187 0.013

(33)

Relative QE 0.00253 0.0263** 0.000882

(0.0146) (0.0112) (0.000538)

LTRO 2.16e-06 -5.25e-06*** 4.29e-08

(1.79e-06) (2.00e-06) (3.90e-08)

Price -6.07e-06*** -2.80e-06* -6.78e-08*

(1.02e-06) (1.47e-06) (3.66e-08)

Volatility 0.0130*** 0.0347*** 0.000133

(0.00485) (0.00909) (9.40e-05)

Turnover Volume -0.00216 0.186** -1.88e-05

(0.00148) (0.0720) (1.42e-05) CSPP-dummy 0.00251*** -0.00358 3.77e-05** (0.000493) (0.00302) (1.81e-05) QE/CSPP interaction -0.0636*** 0.0892 -0.00101** (0.0127) (0.0915) (0.000470) Constant 0.00553*** 0.0235*** 0.000101*** (0.000606) (0.00213) (2.34e-05) Observations 28,596 28,876 28,801 R-squared 0.057 0.205 0.013

(34)

Relative QE 0.00152 0.0227** 0.000835

(0.0147) (0.0105) (0.000515)

LTRO 2.07e-06 -4.72e-06** 4.20e-08

(1.78e-06) (1.90e-06) (3.88e-08)

Price -5.96e-06*** -2.99e-06* -6.77e-08*

(1.03e-06) (1.55e-06) (3.71e-08)

Volatility 0.0138*** 0.0337*** 0.000145

(0.00477) (0.00904) (8.89e-05)

(0.00150) (0.0762) (1.50e-05) CSPP-dummy 0.00266*** -0.00308 3.83e-05** (0.000481) (0.00287) (1.75e-05) QE/CSPP interaction -0.0691*** 0.0635 -0.000973** (0.0129) (0.0921) (0.000447) Constant 0.00395*** 0.0217*** 5.27e-05*** (0.000653) (0.00289) (1.94e-05) Observations 26,652 26,875 26,801 R-squared 0.059 0.222 0.013

(35)

Relative QE -0.000764 0.0209** 0.000721

(0.0148) (0.00946) (0.000490)

LTRO 2.08e-06 -4.47e-06** 4.33e-08

(1.76e-06) (1.82e-06) (3.87e-08)

Price -6.37e-06*** -3.97e-06** -7.87e-08*

(1.15e-06) (1.92e-06) (4.31e-08)

Volatility 0.0158*** 0.0305*** 0.000144*

(0.00452) (0.00860) (8.50e-05)

Turnover Volume -0.00280* 0.192** -1.34e-05

(0.00150) (0.0773) (1.55e-05) CSPP-dummy 0.00286*** -0.00299 3.94e-05** (0.000470) (0.00279) (1.73e-05) QE/CSPP interaction -0.0750*** 0.0472 -0.000874** (0.0131) (0.0963) (0.000425) Constant 0.00320*** 0.0265*** 2.40e-06 (0.000622) (0.00242) (1.92e-05) Observations 24,452 24,663 24,592 R-squared 0.061 0.241 0.012

(36)

Relative QE -0.00407 0.0164* 0.000489

(0.0147) (0.00849) (0.000466)

LTRO 2.15e-06 -4.37e-06** 4.49e-08

(1.75e-06) (1.73e-06) (3.87e-08)

Price -1.13e-05*** -5.62e-06** -1.72e-07*

(2.22e-06) (2.85e-06) (9.41e-08)

Volatility 0.0147*** 0.0250*** 0.000129*

(0.00455) (0.00830) (7.70e-05)

(0.00196) (0.0902) (2.29e-05) CSPP-dummy 0.00313*** -0.00360 4.18e-05** (0.000488) (0.00266) (1.73e-05) QE/CSPP interaction -0.0808*** 0.0503 -0.000712* (0.0133) (0.0941) (0.000408) Constant 0.00486*** 0.0151*** 8.78e-05*** (0.000509) (0.00186) (2.22e-05) Observations 22,466 22,648 22,584 R-squared 0.056 0.263 0.010