The implications of MIFID II on the quality of European Stock Markets

(1)

The implications of MIFID II on the quality of European

Stock Markets

Abstract: On the 3rd of January 2018, MiFID II entered into force in the European Union. MiFID II is a revised set of rules to improve investor protection and the functioning of capital markets in Europe. The goal of this revised legislation is to address the shortcomings of MiFID I because of changes in capital markets. This paper investigates the influence of the introduction of MiFID II on market quality parameters such as liquidity, volatility and trading volume, based on a model by Boneva, Linton and Vogt (2016). Through difference-in-difference and fixed effects regressions, MiFID II is assessed on the short-term effects for the Euro Stoxx 50 index. These regressions resulted in a temporary improvement in liquidity. Trading volume and volatility are not particularly affected by the introduction of the new legislation.

Keywords: MiFID II, Stock markets, Difference-in-difference, Fixed effects, Market quality

Author: Wessel Witmer, MSc.

Student number: s2176912

Supervisor: A.A. Tsvetkov, PhD

Date: 06.06.2018

(2)

2

1. Introduction

As of November 2007, the Markets in Financial Instruments Directive (MiFID) applies across the European Union. This directive provides a harmonised set of regulation in the EU with the purpose of increasing competition and investor protection in financial markets. MiFID introduced new electronic trading platforms, such as Systematic Internalisers (SI) and Multilateral Trading Facilities (MTF), which caused stock market fragmentation across Europe. However, the directive was introduced just before the start of the global financial crisis and lacked regulation regarding equities. Global policymakers felt the need for improved regulation to tackle ‘’under-regulated and opaque aspects of the financial system’’. On the 20th_of

October 2011, the European Commission introduced a proposal for the revision of MiFID in the form of a completely overhauled directive and new regulation: MiFID II and MiFIR1_{. The purpose of the regulation is to improve fairness and create safer,}

more transparent and more efficient markets. Moreover, investor protection is further increased by the introduction of stricter regulation for investment companies, banks, authorities and other financial institutions2_{. MiFID II and MiFIR were enforced on the}

3rd_{of January 2018. The renewed regulation has an impact on almost all aspects of}

trading within the EU. Due to technological developments, trading generally takes place on electronic trading venues and not over the phone. MiFID II stimulates electronic trading, but also imposes strict regulation for institutions regarding reporting information on trades. These rules applied directly after enforcement. The adoption of new regulation generally has an impact on financial markets, however, this is a side effect and not the purpose of new regulation. This paper focuses on these side effects by investigating changes in market quality after the introduction of MiFID II. In order to assess market quality, it is divided into three components: liquidity, volatility and trading volume. Literature review learns that these components are used the most when assessing market quality. Market depth, as a measurement of liquidity, is also widely used. However, market depth, which is the size of an order needed to change the market price, could not be measured due to a lack of data on bid and ask orders. Therefore, the quoted bid-ask spread is the only measurement that gives an indication on liquidity. The market quality components or parameters are described in the methodology section, where it will be explained why they are important and how they are calculated. In the results section, market quality is

1_{Markets in Financial Instruments Regulation}

2_{See the Q&A from ESMA on MiFID II and MiFIR for difference between taking ‘’reasonable steps’’ and}

(4)

4

assessed and based on that one could tell whether or not the introduction of MiFID II went smoothly.

Literature review does not provide a comprehensive definition of market quality. Sellberg (2011) defines market quality as “the prospects for a market participant to successfully match his/her order at a competitive price on a given trading venue”, but Stuchfield et al. (1999) doubt the existence of a precise definition. They state that you can sometimes view market quality from a cost-perspective, and sometimes from a liquidity-perspective. Market participants want to trade against low costs without having any impact on the market, but they also demand quick execution of orders and flawless information streams from the market operators. For the purpose of this paper, liquidity is the most important factor determining market quality. Liquidity is the degree to which a financial instrument can be bought or sold without affecting its price. A significant determinant of the liquidity of stocks are tick sizes. Tick sizes are described as the minimum increments by which the price of a financial instrument can move on exchanges. MiFID II provides a new regime for tick sizes that stock exchanges are required to implement. Article 49 of MiFID II is the basis for the new tick size requirements in the EU. These requirements form the basis for a tick size table as described in Appendix A. The tick size table is determined on the basis of a so-called liquidity band combined with the price of an instrument. Liquidity bands are determined based on the average number of orders relating to a specific financial instrument per day. After determining the price and the liquidity band of a stock, a minimum tick size emerges from the table. The table shows that stocks with a larger amount of average trades per day have a lower minimum tick size. Next to that, a lower price implies a lower minimum tick size. Previous literature on tick sizes reveals that a change in a tick size regime has a big influence on the liquidity of markets. According to the model of Seppi (1997), a decrease in the minimum tick size will constitute a decrease in the cumulative depth of the limit order book for different kinds of investors, which is negatively correlated with liquidity. Harris (1994) reaches an equivalent conclusion: a reduction in tick sizes reduces market liquidity. Next to these theoretical models, there is a wide range of empirical literature on tick size changes in the past decades.

(5)

5

focused on the relationship between market quality and fragmentation after the first directive. The literature overview generally consists of long-term studies and theoretical models for market quality impact. This study differs because of a shorter time frame. Another difference is the research subject. Where most previous studies focused on trading venues, this study focuses on a stock index traded on multiple venues. Here, the index is Euro Stoxx 50. This index consists of the 50 most important blue-chip stocks traded in the Eurozone. Through difference-in-difference and fixed effects estimations, it is clear that the introduction of MiFID II had a short-lasting effect on liquidity. Volatility and trading volume are barely affected. The effects on liquidity fade away during the month of January 2018.

New market regulation has the potential to impact market participants severely. Investors could have to deal with uncertainty and institutions may have to deal with operational problems. Investors try to forecast the possible impact of new regulation. The results of this paper could be helpful in estimating some of the market quality parameters investors are interested in. For institutions who must anticipate on trading behaviour, the results may tell something about possible problems regarding trading capacity or minimum tick sizes. However, the results are limited to a certain time period and are therefore only suitable to anticipating events regarding a short time period.

The next section of this paper describes the literature review which is used as background for this paper. Literature review consists of: (1) background information on MiFID and MiFID II, (2) prior research on MiFID and market quality, (3) theoretical implications on tick size changes, (4) empirical implications on tick size changes. Section 3 describes the methodology, research question, sub-questions and their hypotheses. Section 4 describes the data that is used to measure the effects of the regulatory change. Section 5 explains the results after measuring the relevant dimensions. Section 6 concludes the paper.

2. Literature review

2.1 Background of MiFID and MiFID II

The European Securities and Markets Authority (ESMA) enforced MiFID on the 1st_{of November 2007 across} _{the EU. MiFID’s purpose was to improve}

competitiveness by creating a single market for investment services and activities. Furthermore, it would ensure protection for investors and financial instruments on a high level. The revision of MiFID resulted in a new legislative proposal on the 20th_of

(6)

6

(MiFIR). Both were published on the 12th_{of June 2014}3_{. Next to the directive and the}

regulation, ESMA published a Q&A on the 10th_{of November 2017.}

Gillet et al. (2017) wrote a chapter in the book ‘’Financial Regulation in the EU: From Resilience to Growth’’ about challenges and implications of MiFID and MiFID II regarding the efficiency of markets. They assessed MiFID after 2011 and identify potential challenges for the revised directive. The challenges and contributions of MiFID are separated into four pillars. For this paper, the pillar ‘’transparency and quality of the markets’’ is important. It gives a good overview of the implications of MiFID by referrals to several authors who performed empirical analyses on the topic of market quality in both the US and Europe.

2.2 MiFID and market quality

One of these authors is Gresse (2011) who studied global and local liquidity before and after the introduction of MiFID. Global liquidity refers to liquidity on all markets whereas local liquidity focused on stocks traded on Euronext and LSE. In addition to that, she studies how liquidity is related to fragmentation and internalisation of markets. The analysis is performed through two kinds of perspectives, one from an investor who has access to all kinds of exchanges through technological advantages, and one from an investor who has only access to the primary stock exchanges. Her main results imply that an increase in competition between trading venues has a positive effect on both local and global liquidity, and that trading in dark pools4_{does not have a negative effect on stock liquidity.}

De Jong et al. (2011) studied the relationship between fragmentation and market quality. They used a sample of 52 Dutch stocks covering 2006 until 2009. Equivalent to Gresse, transparent (lit) order books and dark order books are distinguished. The results slightly differ from Gresse, as fragmentation improved global liquidity where trading venues operated with a visible order book, but local liquidity declined in those order books. The effect of improved global liquidity is transitory and ceases to exist after a certain turning point. In dark order books, fragmentation has a negative effect on liquidity.

Riordan et al. (2011) investigated whether MTF’s contribute to the price discovery process in the UK equity market. They find that Chi-X, which is one of the most profound MTF’s, contributes the most to the process of price discovery. Next to that, they conclude that MTF’s have a positive influence on market quality.

3_{https://www.esma.europa.eu/policy-rules/mifid-ii-and-mifir}

4_{Dark pools are non-transparent market venues where transactions almost do not have effect on the price on}

(7)

7

Hengelbrock and Theissen (2009) studied the entrance of Turqouise, another major MTF in the European equity market. They find a decline in quoted bid-ask spreads in 14 European regulated markets after Turquoise’s entrance in 2008. Smaller bid-ask spreads improve market liquidity which is positive for the market quality. In this sense, they have comparable results to Riordan et al. However, market depth may decline which is negatively correlated with market quality.

Boneva et al. (2016) researched the impact of fragmentation on several market quality parameters in a panel of FTSE stocks in a period ranging from 2008 to 2011. Fragmentation in both lit and dark order books lowers volatility of the total order books. However, the volatility of trading in dark pools shows an increasing variability, whereas lit trading has a decreasing variability of volatility. Local and global trading volume is higher when fragmentation in lit order books is low and when there is increased dark trading. Visible fragmentation leads to lower liquidity measured by quoted bid-ask spreads. They find no uniform solution on the optimal level of fragmentation for firms, but state that increased fragmentation is positively related to market capitalisation.

This paper’s main focus is on market quality. The mentioned authors in this paragraph generally focused on the interaction between fragmentation and market quality. However, their results form a basis in how to assess market quality and its relevant parameters. In a broad sense, it could be stated that where these authors focused on the event of fragmentation, this paper studies the enforcement of new regulation with a particular focus on the collateral damage emerging from it.

2.3 Theoretical implications on tick size changes

(8)

8

Reducing these price variations leads to a change in the level and nature of the liquidity provided. For instance, liquidity providers could decrease the number of shares they want to buy, or even move away from the quotes. In the worst case, they could leave the market when they have reached their margin. Another alteration that could be the result of decreasing tick sizes is that liquidity providers jump ahead of pending limit orders to guarantee a better place in the queue of orders.

Seppi (1997) provides a microstructure model of liquidity provision where a specialist with market power competes against a limit order book5_{. This model is}

useful to answer questions on tick sizes. Seppi compares eights and decimal pricing methods for tick sizes. This has implications for market liquidity. Larger market participants have to deal with higher tick sizes than small retail investors. However, he comes to the same conclusion as Harris that a tick size greater than zero is for both types of investors preferable.

For the purpose of this paper, the models by Seppi and Harris form a basis for what to expect from a tick size change. In the next paragraph, empirical research shows the impact of tick size changes in the past. The results of these researches are generally in line with the findings from Seppi and Harris. Researchers in the next paragraph referred a lot to Seppi and Harris.

2.4 Empirical implications on tick size changes

Regarding empirical research, Goldstein and Kavajecz (2000) investigated the impact of a reduction of minimum tick size on market liquidity concerning stocks on the NYSE. The NYSE changed the tick size from 1/8th_{to 1/16}th_{of a dollar in June}

1997. Cohesive with the mentioned theoretical models, this reduced the bid and ask spreads and market depth. This holds for the entire limit order book. The combined effect of smaller spreads and a reduction in market depth was profitable for smaller retail investors. For traders who invested larger orders in lower volume stocks, the reduction in tick size was not beneficial. The reduction became even less beneficial when stock prices were lower. These developments caused more uncertainty for liquidity demanders and are therefore not always beneficial to the welfare of investors. Furthermore, Goldstein and Kavajecz argue that investigating quoted spreads and depths is not sufficient to assess overall changes in liquidity of markets. One should know where the position of market depth is to fully understand how liquidity has changed. Next to that, market operators and regulators should consider both liquidity demanders and providers when assessing the actions on tick sizes.

(9)

9

Where Goldstein and Kavajecz only investigated the NYSE, Huang and Stoll (2001) added a sample of a dealer market: the LSE. They provide a link between market structure and market characteristics on tick sizes, bid-ask spreads, quote clustering6_{and market depth. On the LSE, there were no mandatory tick sizes}

opposite to the NYSE. They conclude that market characteristics are endogenous to the type of market structure. Their results show higher spreads, higher quote clustering and a deeper market on the LSE compared to the NYSE. For auction markets, minimum tick sizes are needed as otherwise limit orders can easily stay ahead of other limit orders or dealer quotes. Without tick sizes, traders in auction markets can easily avoid time priority. Dealer markets do not need this time priority. The higher spreads in London are positively correlated to market depth. The high depth is consistent with the large trading sizes. However, the difference in trading sizes is not as big as the difference in depth.

Oppenheimer et al. (2009) made an analysis of cross-listed stocks traded on the NYSE and in France (Euronext) after decimalisation in the US. The French stocks are large and frequently traded by institutions. Due to the decimalisation in the US, market depth has decreased. They conclude that US markets become less competitive for the institutions who trade the French stocks. The trade size of the stocks on the NYSE declines dramatically compared to the Euronext stocks after the decimalisation. Furthermore, they find that there is a positive correlation between the decline of US institutional share trading and the decline in depths on US stock markets. This decline is even greater for French institutions. Summarised, they discover an order flow migration regarding French institutions from the NYSE to Euronext after the decimalisation of tick sizes in the US.

Biais, Bisiere and Spatt (2010) researched competition between the trading platforms Island7_{and NASDAQ. Different from previous researches, Island is an}

electronic trading platform whereas the others are traditional stock exchanges. Following a tick size reduction, the competition between these trading platforms changed due to differences in market structure. The tick size reduction was only effective for NASDAQ. Island traders undercut competition with NASDAQ traders through taking advantage of lower tick sizes. This led to a reduction in the spread of Island stocks.

Jiang, Kim and Wood (2011) also performed research on tick sizes after US decimalisation. Where the other researches focus on liquidity and competition, these

6_{Quote clustering is a tendency where bid and ask orders are concentrated around a certain price and do not}

deviate much from that

7_{See http://www.fundinguniverse.com/company-histories/the-island-ecn-inc-history/ for information on}

(10)

10

researchers compare volatility and transaction costs between NASDAQ (dealer market) and the NYSE (auction market) after the change in tick sizes. They find that volatility on dealer markets is much higher than volatility on auction markets, before and after the change. A tick size change makes these differences slightly bigger. Transaction costs which are measured by quoted and effective spreads remain significantly higher on the dealer market than on the auction market. These differences do not result from the different stock characteristics traded on the exchanges. Furthermore, they find that the frequency of small trades on the auction market is much higher than on the dealer market.

Buti et al. (2013) analysed a tick size reduction in a public limit order book (PLB). They perform their research for liquid and illiquid stocks. For illiquid stocks, market quality and welfare fall after the reduction. For liquid stocks, the opposite results. Illiquid stocks are more likely to disappear after a tick size reduction resulting from the competition of alternative trading venues. These results are coherent with the decision of the SEC to increase tick sizes for illiquid stocks. In a later study, these results were confirmed8_{. The alternative trading venues are mentioned as Sub-Penny}

Venues (SPV). All traders can demand liquidity on such a venue. Therefore, total volume and welfare increases after a tick size reduction at the PLB irrespective of the level of liquidity. For stocks traded in the PLB, spreads and depths decline.

The mentioned researches generally have a focus on decimalisation and tick size changes in the US. Most of these papers research the impact on liquidity, whereas for example Oppenheimer et al. were interested in volatility. These market quality parameters will be discussed in this paper. However, distinctions are made between a dealer market and an auction market. This distinction will not be made in this paper. Nikkei 225 and Euro Stoxx 50 are mostly traded on highly automated exchanges which do not require intermediary interference by brokers or dealers. The intention of the distinction was to differentiate between market structures. In this paper, the most important factor is the introduction of new regulation. The effect of the new rules will have different consequences for different market structures. This paper uses indices traded on several markets rather than a single exchange. Therefore, market structures will be of subsequent relevance. As discussed previously, MiFID II introduces several regulatory changes to MiFID I. The mentioned papers generally focus on tick size changes where this paper will extend this view. Changes in market quality could depend on various factors. Possible other factors are external realities like disasters, bankruptcy events that influence the global market (think of Lehman Brothers) or market corrections after periods of high return.

(11)

11

3. Methodology

In the introduction, it is stated that one of the characteristics of market quality is liquidity. This means that measurements of liquidity are necessary. Liquidity relates positively to market quality, because liquid markets provide investors with the ability to trade without significantly affecting prices. Liquidity is measured by the quoted bid-ask spread. The next paragraph provides details on the quoted bid-bid-ask spread. Next to liquidity, volatility is an important measurement of market quality. High volatility has a negative influence on market quality in general. One of the main reasons for this is that investors and market operators cannot react quickly to a fast changing market. Therefore, exchanges may face operational problems and investors may face higher trading costs. Furthermore, volatility is an interesting topic to study when new regulation is implemented, because traders may face uncertainty. Predictions about the potential impact of volatility on market quality were not unambiguous according to Sellberg (2011). The same holds for trading volume. It is not clear whether trading volume will increase or decrease after implementation. Generally, increased trading volume improves market quality as higher volumes imply an increased number of orders on both the bid and ask side. This means that orders can be matched easily and prices remain significantly unaffected. The measurements of market quality parameters described in the next paragraphs are based on the methodology from Boneva et al. (2016).

In order to measure market quality, the three components are analysed separately. After the analyses, it is clear whether a component had a positive, negative or neutral impact on market quality. The overall impact on market quality is based on whether these equally weighted components combined show positive, negative or neutral signs.

3.1 Liquidity

(12)

12

effective spreads and quoted spreads. Measuring market depth also becomes difficult as some liquidity providers issue limit orders on different trading venues for the same instrument. After the trade is executed at one venue, the other orders will be cancelled9_{. The aggregated orders could overstate liquidity. However, due to the}

existence of ‘’iceberg orders’’ and dark pools, there is a substantial part of liquidity that is hidden. As order book data is not completely available, the quoted bid-ask spreads will be used. This is the formula for quoted bid-ask spread (𝑩𝑨𝒊,𝒕):

𝑩𝑨𝒊,𝒕 =

𝑷_𝒊,𝒕𝑨 − 𝑷_𝒊,𝒕𝑩 𝟏

𝟐 (𝑷𝒊,𝒕𝑨 + 𝑷𝒊,𝒕𝑩) (1)

The bid (𝑷_𝒊,𝒕𝑩) and ask (𝑷_𝒊,𝒕𝑨 ) prices indicate the intraday prices at 17:30 Central European Time, which is the time when stock markets close. The stock is indicated by i on day 𝑡.

Kyaw and Hillier (2013) discuss the relationship between liquidity, volatility and trading activity. Generally, volatility and liquidity are negatively related. This means that high volatility results in a lower bid-ask spread. It is not likely that liquidity influences volatility. Regarding the relationship between liquidity and trading activity (assuming it is almost the same as trading volume), they find that there is a positive relationship for firms with a large market value and a negative relationship for firms with a small market value.

3.2 Volatility

Volatility is the degree to which a stock price moves given a certain amount of time. High volatility is often related to a stressful market or high losses. However, high volatility could be positively interpreted under circumstances. Boneva et al. (2016) refer to Bartram et al. (2012), who state that in the US, volatility is generally higher than in other countries. It reflects innovation and competition rather than a lack of market quality. For the purpose of this paper, the volatility estimator of Rogers and Satchell (1991) is used. The equation is as follows:

𝑽𝒊,𝒕= (𝒍𝒏𝑷𝒊,𝒕𝑯 − 𝒍𝒏𝑷𝒊,𝒕𝑪 )( 𝒍𝒏𝑷𝒊,𝒕𝑯 − 𝒍𝒏𝑷𝒊,𝒕𝑶)+( 𝒍𝒏𝑷𝒊,𝒕𝑳 - 𝒍𝒏𝑷𝒊,𝒕𝑪)( 𝒍𝒏𝑷𝒊,𝒕𝑳 - 𝒍𝒏𝑷𝒊,𝒕𝑶) (2)

𝑽_𝒊,𝒕 implicates the volatility on day t for stock i. 𝑷_𝒊,𝒕𝑯, 𝑷𝑳_𝒊,𝒕, 𝑷_𝒊,𝒕𝑶and 𝑷_𝒊,𝒕𝑪 indicate the daily high, low, opening and closing prices. Furthermore, it is important to decompose volatility into temporary and permanent volatility. Permanent volatility is characterised by the underlying uncertainty of a future payoff stream of the financial instrument.

(13)

13

Temporary volatility, on the other hand, is unrelated to fundamental information about future payoff streams and is caused by reactions of traders to perceived events. Distinguishing temporary volatility from permanent volatility requires an assumption. The assumption is that permanent volatility is characterised by a smooth time trend, whereas temporary volatility is characterised by the residuals from the non-parametric regression of total volatility on a given period of time. The smooth time trend is obtained by forecasting. The method of forecasting used here, as discussed in Wooldridge (2011), is called single exponential smoothing. The following formula indicates how this method of forecasting works:

𝑬(𝒚_𝒕+𝟏) = 𝜶𝒚_𝒕+ (𝟏 − 𝜶)𝒚_𝒕−𝟏+ ⋯ + (𝟏 − 𝜶)𝒕_𝒚

𝟎 (3)

where 0<α<1 is a chosen parameter. The formula is applied for every 𝒚_𝑡. This method is called exponential smoothing, because of the exponentially declining weights on the lagged 𝒚 calculated as:

𝒚𝒕 =

∑ 𝑽𝒏 _𝒊,𝒕 𝒊

𝒏

(4) According to Wooldridge (2011), ‘’the forecast of 𝒚_𝒕+𝟏 is a weighted average of (𝒚𝑡) and the forecast of (𝒚𝒕) made at time t-1’’. After forecasting, α is approximately

0,999. This means that total volatility is close to a random walk. Temporary volatility is obtained by subtracting permanent volatility from the average total volatility per day.

Kyaw and Hillier (2013) found that volatility generally has a negative influence on liquidity. The relationship between volatility and trading activity is under all circumstances positive. Thus, an increase in trading volume causes a more volatile market. These findings are supported by previous literature regarding both theoretical and empirical frameworks, to which they refer in their paper.

3.3 Trading Volume

Trading volume is the number of orders of a certain stock. It is a measure of market participation. Global volume and volume per venue are distinguished here. Global volume indicates the trading volume of a certain asset, which is measured by turnover by volume. This is the absolute numbers of trades. The following equation indicates the normalised global volume (𝑵𝑮𝑽𝒊,𝒕):

𝑵𝑮𝑽𝒊,𝒕=

𝑮𝑽_𝒊,𝒕 𝑮𝑽_𝒊 ̅̅̅̅̅

(14)

14

where 𝑮𝑽𝒊,𝒕 is global volume, in this paper known as turnover by volume, and 𝑮𝑽̅̅̅̅̅ is 𝑖

the mean of global volume per stock i over a period from the 31st_{of December 2015}

until the 8th_{of March 2018 (t) calculated by:}

𝑮𝑽_𝒊

̅̅̅̅̅ =∑ 𝑮𝑽𝒏𝒕 𝒊,𝒕

𝒏

(6) This extended period ensures that years or periods with potential volatile trading are smoothed. The volume per venue indicates the traded volume per market venue as assets can be traded over multiple venues. In the data section of this paper, the use of multiple trading venues will be elaborated.

Kyaw and Hillier (2013) use trading activity as a variable, which is approximated by the number of trades or trading volume. As discussed before, for firms with a higher market value, higher liquidity implies less trading activity. For firms with a lower market value, the opposite is true. Trading activity and volatility are positively related.

3.4 Difference-in-difference

In order to measure the effects of the introduction of MiFID II, a difference-in-difference estimator is applied. Difference-in-difference-in-difference is especially useful when assessing short-term changes, because it assumes parallel trends over time. Based on the equation Haferkorn and Zimmermann (2014) use in their paper, the equation used in this paper is:

𝒀𝒊,𝒕− 𝒀̅ = 𝒂𝒊 𝟎+ 𝒂𝟐𝑻𝒕+ 𝒂𝟑(𝑿𝒊∗ 𝑻𝒕− 𝑿̅̅̅) + ∑ 𝒂𝒊 𝒋(𝒄𝒋,𝒊,𝒕− 𝒄̅̅̅̅)𝒋,𝒊 𝟓

𝒋=𝟏

+ 𝜺𝒊,𝒕 ₍₇₎

𝒀_𝒊,𝒕 is the relevant dependant market quality parameter as introduced before. The mean over time of the market quality parameter is expressed as:

𝒀̅ =_𝒊 ∑ 𝒀𝒊,𝒕

𝒏 𝒕

𝒏 (8)

where i represents the stock (either in the Euro Stoxx 50 or Nikkei index) and t represents time in days10_{. t runs over a period depending on one of the three time}

frames described in the data section of this paper. It is either a 10-week, 4-week or 1-week pre- and post-treatment time frame. 𝑿_𝒊 is the dummy variable that has a number equal to 1 if a stock is in the treatment group (Eurostoxx 50) and 0 if the stock is in the control group (Nikkei). By applying cross-sectional fixed effects, this dummy drops out of the regression by subtracting the mean of the dummy variable per stock:

(15)

15

𝑿𝒊

̅̅̅ = ∑ 𝑿𝒏𝒕 𝒊,𝒕

𝒏

(9) 𝑻𝒕 is the dummy variable which indicates the pre-treatment (0) and the

post-treatment period (1) with t in days. 𝑿𝒊∗ 𝑻𝒕 indicates the interaction between the two

dummy variables, the unbiased event estimator. The control variables are included in the estimator ∑𝟓𝒋=𝟏𝒂𝒋(𝒄𝒋,𝒊,𝒕− 𝒄̅̅̅̅)𝒋,𝒊 where n=5. 𝒄𝒋,𝒊,𝒕 indicates the control variable j for

stock i on day t. The mean over time is calculated by the following equation: 𝒄𝒋,𝒊 ̅̅̅̅ =∑ 𝑪𝒋,𝒊 𝒏 𝒕 𝒏 (10) The following control variables are used in the regression:

Natural logarithm of the market value: the daily market value per firm expressed in natural logarithms controls for firm size. Differences in firm size should be eliminated for regression purposes, because firm size may influence the number of orders. A bigger firm generally implies a higher number of orders. Gresse (2011) used market capitalisation as a control variable, which is the number of shares outstanding multiplied by the stock price. Market value is based on more complex calculations, but highly correlates with market capitalisation. However, Datastream provides data on market value per stock. The formula for the natural logarithm of market value is:

𝒄_{𝟏,𝒊,𝒕}= 𝒍𝒏(𝑴𝑽_𝒊,𝒕) (11)

where 𝑴𝑽𝒊,𝒕 stands for market value for stock i on day t11.

Natural logarithm of the stock price: the daily stock price per firm expressed in natural logarithms controls for stock price level. Differences in stock price level cause differences in liquidity, volatility and trading volume, for which should be accounted. Haferkorn and Zimmermann (2014) use control variables for stock levels for a similar purpose. The formula for the natural logarithm of the stock price is:

𝒄_{𝟐,𝒊,𝒕} = 𝒍𝒏(𝑷𝒄

𝒊,𝒕) (12)

where 𝑷𝒄

𝒊,𝒕 is the closing price for stock i on day t.

Price movements: the daily difference in absolute measures of stock prices controls for daily price movements. The magnitude of these movements differs per stock because of different price levels and specific firm-related events, in which we are not interested. Haferkorn and Zimmermann (2014) use these control variables for similar purposes. The formula for price movements is:

(16)

16

𝒄_{𝟑,𝒊,𝒕} = 𝑷𝒄

𝒊,𝒕− 𝑷𝒄𝒊,𝒕−𝟏 (13)

where 𝑷𝒄

Price variability: the daily difference in relative measures of stock prices is an additional control for daily price movements. Absolute measurements may not control for all movements, because price level differences are quite big. Therefore, additional relative measurements are applied. This variable is based on Haferkorn and Zimmermann (2014) where they control for both absolute and relative price movements. The formula for the relative price movements is:

𝒄𝟒,𝒊,𝒕 = 𝑷𝒄𝒊,𝒕− 𝑷𝒄𝒊,𝒕−𝟏 𝑷𝒄 𝒊,𝒕−𝟏 ∗ 𝟏𝟎𝟎 (14) where 𝑷𝒄

Volume variability: the daily difference in relative measures of turnover by volume controls for changes in volume. Trading volume of individual stocks may increase, because of specific firm-related events. These specific events are not of interest here, because MiFID II applies to all stocks in the Euro Stoxx 50 index. Based on Haferkorn and Zimmermann (2014) who control for volume variability for similar purposes, the formula is:

𝒄_{𝟓,𝒊,𝒕} =𝑮𝑽𝒊,𝒕− 𝑮𝑽𝒊,𝒕−𝟏

𝑮𝑽_{𝒊,𝒕−𝟏} ∗ 𝟏𝟎𝟎

(15) where 𝑮𝑽_𝒊,𝒕 is the global volume for stock i on day t.

The last term, 𝜺_𝒊,𝒕, is the idiosyncratic error term. Results are obtained through ordinary least squares regression, which causes the error term to be always zero.

3.5 Fixed effects model

Because difference-in-difference approaches are only useful when time trends between the treatment group and the control group are parallel, it is wise to use an additional approach. Therefore, a fixed effects model is applied. This model will follow more or less the same pattern as the difference-in-difference approach. However, there will be no control group. The following equation is used:

𝒀_𝒊,𝒕− 𝒀̅ = 𝒂_𝒊 _𝟎+ 𝒂_𝟐𝑻_𝒕+ ∑ 𝒂_𝒋(𝒄_{𝒋,𝒊,𝒕}− 𝒄̅̅̅̅)_𝒋,𝒊

𝟓

𝒋=𝟏

+ 𝜺_𝒊,𝒕 ₍₁₆₎

(17)

17

estimated by using ordinary least squares. The results section provides more details on the purpose of these regressions.

3.6 Hypotheses

The purpose of this paper is to answer the main question:

‘’What are the short-term effects on the quality of European Stock markets after the introduction of MiFID II’’

The answer to this question is based on a set of sub-questions. All sub-questions have a corresponding hypothesis. The following sub-question concerns liquidity:

“What are the effects of MiFID II on the liquidity of European stocks?”

The expectation is that overall market liquidity will improve due to tighter bid-ask spreads. However, previous research shows that market depth could decrease which mitigates the improvement in market liquidity. Therefore, it is expected that market liquidity will only slightly improve.

Another important market quality parameter, volatility, is divided into two separate sub-questions. The first question is:

“What are the effects of MiFID II on temporary volatility of European stocks?”

Here, the expectation is that temporary volatility will increase because of new regulation. New regulation could lead to uncertainty in markets, which increases volatility on a temporary basis.

The second question regarding volatility is:

‘’What are the effects of MiFID II on total volatility of European stocks?”

As total volatility depends on both the future payoff streams of the underlying asset and on temporary volatility, the expectation is that total volatility will not change significantly. The introduction of MiFID II will not directly have an impact on these future payoff streams.

The third parameter of market quality is trading volume. Trading volume is separated into global volume, which is characterised by turnover by volume, and volume per venue. The following question concerns global volume:

(18)

18

The expectation is that global volume will increase. This could be the consequence of the improved investor protection that MiFID II proposes. This effect could be temporary, because the shock effect of the new regulation decreases over time. However, this paper generally focuses on the short-term effects of MiFID II, because the regulation is introduced quite recently.

The following question concerns volume per venue:

“What are the effects of MIFID II on the volume of trading on separate European trading venues?”

Probably, traditional stock exchanges will face a relatively slighter increase or decrease of trading volume compared to SI’s and MTF’s. MiFID introduced SI’s and MTF’s in 2007. Since then, these types of trading venues have gained significantly in market share. Due to improved market regulation, especially for these venues, their market share could even become bigger.

4. Data

The Euro Stoxx 50 index is the main source of data of this paper. This index consists of the 50 biggest blue chip stocks in the Eurozone. It will be fully used as the treatment group in the difference-in-difference regression. Choosing blue chip stocks instead of growth stocks is justified by the fact that growth stocks are generally riskier and more volatile than blue chip stocks. The difference-in-difference approach requires, next to a treatment group, a control group. The control group should not fall under the MiFID II regime. The control group consists of a selection of stocks from the Nikkei 225 index. This is the biggest Japanese index consisting of 225 stocks. The 50 stocks with the highest average market value measured over a period from October 2017 until March 2018 are used in the sample. Nikkei 225 is highly liquid compared to several other indices in the Pacific region. These 50 stocks are chosen because the number is balanced with Euro Stoxx 50. The decision for highest market values is made based on the fact that blue chip stocks, like in the Euro Stoxx 50 index, generally have a high market value as well. Previous research showed that American indices sometimes consist of European firms which may be influenced by MiFID II12_{. Therefore, it is wise to take an index that stands further away from}

Europe, but still has comparable characteristics.

Measuring liquidity requires the daily bid and ask prices of stocks. The measurement of volatility includes a wide range of prices, specifically the daily open,

(19)

19

close, high and low prices of all stocks. Volume is measured by using the daily turnover by volume of all stocks in the two samples. Datastream provides the above-mentioned data. The data on the separated trading venues is gathered from the website of CBOE13_{. This only concerns data on the Euro Stoxx 50 index. Instead of}

data on individual stocks in the index, the total trading volume of the index is available on the website of CBOE. The data consists of the daily notional value of the index traded on several trading venues and the daily market share. The venues included are: Turquoise, LSE Group, Euronext, Xetra and CBOE Europe. These are the five venues with the biggest market share in Euro Stoxx 50.

The obtained datasets range from the 12th_{of October 2015 until the 8}th_of

March 2018. However, the paper uses various samples to measure market quality. As mentioned in the methodology section, there are three time frames used to measure market quality. The 10-week pre- and post-event time frame ranges from the 23nd_{of October 2017 until the 8}th_{of March 2018. There are around 9023}

observations per regression. The second time frame is a 4-week pre- and post-event time frame. This sample ranges from the 4th_{of December 2017 until the 26}th_of

January 2018. Around 3448 observations are measured. The smallest time frame covers 1-week pre- and post-event, measured from the 27th_{of December until the 5}th

of January. This sample indicates the effects after the year turn. Around 450 observations are measured. Additional measurements contain the same time frames, but with two weeks, four weeks, six weeks, six months, a year prior and two years prior to the MiFID II introduction. In the results section, it will be elaborated why these extra periods are used. In appendix B, the exact dates for all measurement periods are stated.

The following table contains the descriptive statistics of the Euro Stoxx 50 dataset:

(20)

20

Table 1: Descriptive Statistics

The sample consists of all Euro Stoxx 50 stocks (50 firms) in a period ranging from the 23rd_of

October 2017 until the 8th_{of March 2018 (panel A) and a period ranging from the 4}th_of

December 2017 until the 26th_{of January 2018 (panel B). Due to the small number of}

observations, the smallest time span is ignored. The observations of prices, quoted bid-ask spread and total volatility are calculated on a daily basis for every firm. Temporary volatility is based on an exponentially weighted moving average on a weekly basis. Turnover by Volume is measured per stock, whereas total volume and the volume separated per exchange consist of the daily sums of stocks. (*) implies multiplied by 1000. (~) implies in millions. (^) implies in billions.

Panel A: 10-week pre- and post-event window

Variable Mean Std. Dev. Median Maximum Minimum Observations

Price High 70.96 63.27 51.19 260.55 2.76 4783 Price Low 69.80 62.25 50.46 256.85 2.71 4783 Price Open 70.40 62.79 50.73 259.95 2.75 4783 Price Close 70.41 62.75 55.98 259.55 2.75 4783 Price Bid 70.33 62.72 50.85 259.35 2.75 4783 Price Ask 70.37 62.76 50.87 259.55 2.75 4783

Quoted Bid-ask spread* 0.54 0.46 0.42 8.27 0.00 4783

Total Volatility* 0.13 0.18 0.09 5.62 0.00 4783

Temporary Volatility* 0.12 0.06 0.10 0.32 0.05 21 Turnover by Volume~ 42.73 54.95 10.98 488.55 0.02 4783

Total Volume^ 12.10 3.11 11.67 26.83 5.79 95

Volume on CBOE Europe^ 2.65 0.72 2.63 5.97 1.09 95

Volume on Xetra^ 2.97 0.88 2.81 6.70 1.34 95

Volume on Euronext^ 3.47 0.94 3.36 7.97 1.71 95

Volume on Turquoise^ 1.03 0.28 0.98 2.21 0.45 95

(21)

21

Panel B: 4-week pre- and post-event window

Variable Mean Std. Dev. Median Maximum Minimum Observations

Price High 71.66 63.53 62.70 254.70 2.76 1849 Price Low 70.61 62.58 60.97 252.40 2.71 1849 Price Open 71.13 63.08 61.44 253.65 2.75 1849 Price Close 71.09 63.04 55.53 254.05 2.75 1849 Price Bid 71.11 63.02 62.08 254.05 2.75 1849 Price Ask 71.15 63.06 62.12 254.20 2.75 1849

Quoted Bid-ask spread* 0.56 0.53 0.42 8.27 0.10 1849

Total Volatility* 0.10 0.11 0.07 2.21 0.00 1849 Temporary Volatility* 0.09 0.01 0.09 0.11 0.07 9 Turnover by Volume~ 39.59 50.22 9.92 368.71 0.02 1849 Total Volume^ 11.12 2.65 11.20 19.89 5.79 37 Volume on CBOE Europe^ 2.39 0.60 2.44 3.69 1.09 37 Volume on Xetra^ 2.73 0.85 2.66 6.25 1.34 37 Volume on Euronext^ 3.24 0.82 3.24 6.71 1.71 37 Volume on Turquoise^ 0.97 0.25 0.92 1.57 0.45 37 Volume on LSE^ 0.57 0.15 0.53 1.03 0.34 37

5. Results

Before the regression results are analysed, a couple of characteristics of the dataset should be considered. The Euro Stoxx 50 index consists of a wide range of stocks with varying prices and trading volumes. The highest price is above €250 and the lowest price is around €2,75. Turnover by volume shows that there are big

variations in daily transactions, because the standard deviation in the 10-week time frame is around 55 million transactions. These differences are quite big. This has influence on the minimum tick sizes, as discussed in the introduction.

This chapter is divided into two parts. First, the trends of the market quality parameters are analysed. It gives an idea how Euro Stoxx 50 and Nikkei relate to each other. Second, the regression results are analysed, which leads to a conclusion regarding changes in market quality.

5.1 Analysis of trends

(22)

22

less the same pattern. There are also no significant differences between both indices. Temporary volatility is calculated by subtracting permanent volatility from total volatility. Permanent volatility is a smooth time trend, so temporary volatility can take a negative number, although in practice this is not possible. As explained before, permanent volatility is not a major subject here. Although generally both indices follow the same patterns, there are some slight differences. In November 2017, for both temporary and total volatility, Nikkei shows a peak whereas Euro Stoxx 50 stabilises. At the end of December 2017, Nikkei volatility reaches an all-time low whereas Euro Stoxx 50 stabilises toward the end of the year. Furthermore, both indices peak in February 2018. Most likely, this peak is caused by a market correction after periods with high returns. Volatility is slightly higher at the end of the sample than at the beginning. The differences between total and temporary volatility are not very clear. However, Nikkei and Euro Stoxx 50 show more similarities in temporary volatility than in total volatility. Permanent volatility just has a weak impact on total volatility. Daily movements and shocks are the most significant factors determining total volatility.

The average bid-ask spreads of both indices are presented in panel c). It is clear that the Euro Stoxx 50 is a more liquid index than Nikkei 225. The line of Euro Stoxx 50 never exceeds the line of Nikkei. Euro Stoxx 50 has an average bid-ask spread of around 0,006 cents, Nikkei’s average is around 0,01 cents. Bid-ask spreads of the Nikkei sample are more volatile than the Euro Stoxx 50 spreads. The bid-ask spreads of the Euro Stoxx 50 peak in the last week of 2017. In the first week of 2018, these numbers move back to the old level. The numbers show that these patterns were also visible in previous years14_{. The Christmas holidays, where traders are often on}

vacation, and therefore orders decrease, are the most likely cause of these patterns. The graph shows that for a less liquid index, bid-ask spread variations are higher. This could have to do with a lack of matching orders in the index. For example, there could be too much orders at a certain bid price, without having enough orders for the same ask price. Bid orders could not be executed. There are generally enough matching orders in the Euro Stoxx 50 index. Minimum tick sizes seem to be at a sufficient level to ensure high liquidity.

14_{Weekly meaning by date shows that at the end of 2015 the average was 0,000865 against 0,000617 in the}

(23)

23

Figure 1: Market quality parameters

Panel a) represents the weekly mean of total volatility of Euro Stoxx 50 and Nikkei over the sample period. Panel b) represents the weekly mean of temporary volatility of both indices. Temporary volatility is calculated by subtracting permanent volatility from total volatility. Panel c) represents the weekly means of the quoted bid-ask spreads of both indices. Panel d) represents the weekly means of turnover by volume for both indices expressed in thousands.

a) Total volatility b) Temporary volatility

c) Bid-ask spreads d) Volume

(24)

24

Figure 2 shows the trading volume per venue of Euro Stoxx 50. Euronext is the leading venue in trading volume for Euro Stoxx 50. Xetra, CBOE Europe, LSE and Turquoise follow. The peak in February 2018 and the fall in December 2017 are also visible. However, Euronext and Xetra show a peak before the fall in December whereas the other trading venues do not show such a pattern. Euronext generally covers stocks in The Netherlands, Belgium and France, whereas Xetra is a German trading venue. Dutch, French and German stocks compose a big part of Euro Stoxx 50. Therefore, it seems reasonable that Euronext and Xetra have the biggest share in trading volume. Turquoise is the only MTF in this selection of trading venues. It was introduced in 2008 after the first MiFID directive. Trading volume stays around 1 billion trades per day over the time frame. The shocks are not as big as the shocks on trading venues with higher volume. In February 2018, there is also a peak in trading volume, which is probably caused by the market correction.

5.2 Analysis of regression results

Table 2 shows the difference-in-difference regressions for the market quality parameters that were possibly affected by the introduction of MiFID II. Quoted bid-ask spread and turnover by volume show significant effects in the 1-week and the 10-week time frames. Furthermore, turnover by volume is affected in the 4-10-week time frame. Volatility remains unaffected. Difference-in-difference assumes parallel time trends of the compared indices. For stock indices, this is generally not true.

Figure 2: Separated volume

The coloured lines indicate the trading volume of Euro Stoxx 50 per trading venue. The trading venues are: Euronext (red), Xetra (green), CBOE Europe (black), Turquoise (blue), LSE (purple). These are the five trading venues where Euro Stoxx 50 stocks are traded most frequently.

0.00E+00 1.00E+09 2.00E+09 3.00E+09 4.00E+09 5.00E+09 6.00E+09 7.00E+09 8.00E+09 9.00E+09 10/23/2017 11/23/2017 12/23/2017 1/23/2018 2/23/2018 N u mb e r o f tr ad e s Date

Volume per venue

(25)

25

Table 2: Difference-in-difference regression results

Regression results are based on three balanced time periods where the MiFID II-introduction is the event. Time effects are used to cancel out the effect of the turn of the years. Significant results are highlighted. T-statistics are presented in brackets. *** is significant at 1%, ** is significant at 5%, * is significant at 10%.

Dependent variable

Independent variable 10-week pre- and post-event frame 4-week pre-and post-event frame 1-week pre-and post-event frame Quoted bid-ask spread Volatility Turnover by Volume Quoted bid-ask spread Volatility Turnover by Volume Quoted bid-ask spread Volatility Turnover by Volume Coefficient -0.02* -0.001 -0.07 5E-03 4E-04 3.19 -0.05 9E-03 -66.8

(1.95) (-0.19) (-0.01) (0.3) (0.11) (0.28) (-0.13) (0.33) (-0.84) Time effects -1E-05 4E-05*** 0.05*** -3E-05 8E-06 0.07*** 5E-04*** 5E-05*** 0.48***

(-0.82) (8.44) (5.08) (-1.19) (1.60) (4.25) (3.26) (5.12) (15.9) Treatment effects -5E-05** -7E-06 0.04*** -4E-05 -8E06 -0.05** -1E-03*** 7E-06 -0.11***

(-2.22) (-0.98) (3.14) (-1.42) (-1.25) (-2.33) (7.23) (0.65) (-3.07) Log Price -3E-03** -6E-04 -1.75** -3E-04 1E-04 0.88 -0.02 1E-03 -0.29

(-2.14) (-1.39) (-2.01) (-0.12) (0.30) (0.54) (-0.46) (0.33) (-0.03) Log Market Value 0.003** 3E-04 0.87 -2E-04 -1E-04 -0.58 0.01 -1E-03 5.29

(2.09) (0.75) (1.00) (-0.09) (-0.20) (-0.36) (0.28) (-0.33) (0.47) Price Movement -1E-08 -6E-08*** -5E-05* 3E-07*** 6E-08*** -5E-05 2E-07 1E-07*** 1E-04 (-0.2) (-3.54) (-1.70) (3.57) (3.97) (1.07) (0.53) (3.44) (1.04) Volume Variability 6E-07*** 1E-06*** 6E-03*** 2E-07 7E-07*** -5E-03*** -7E-08 5E-07*** 4E-03***

(4.70) (30.0) (69.4) (1.22) (17.8) (45.6) (-0.06) (5.80) (16.4) Price Variability -1E-05** 4E-08 -5E-03* -6E-07 -6E-06*** 0.03*** 1E-04 -4E-06 -0.04***

(-2.43) (0.03) (-1.69) (-0.09) (-4.06) (5.43) (2.44) (-0.98) (-3.46) R-squared 0.38 0.18 0.44 0.46 0.22 0.49 0.44 0.50 0.85 Observations 9023 9023 9023 3448 3448 3448 500 500 500

(26)

26

Figure 3: Difference-in-difference regression comparison

Panel a) represents difference-in-difference regression results for quoted bid-ask spread with the middle point exactly 2 years, 1 year, 6 months, 6 weeks, 4 weeks 2 weeks before and the middle point exactly during treatment. Panel b) represents the same for volatility. Panel c) represents the same for turnover by volume after dividing the numbers through its mean as explained in the methodology section. The regressions are based on the same independent variables as in table 2. Coefficients of the points are captured in panel a) of appendix C.

-0.0015 -0.001 -0.0005 0 0.0005 0.001

2 years 1 year 6 months 6 weeks 4 weeks 2 weeks Treatment

C oe ff ic ie n t

Time before treatment

a) Quoted bid-ask spread

1 week span 4 week span 10 week span

-0.00015 -0.0001 -0.00005 0 0.00005 0.0001 0.00015 0.0002

C oe ff ic ie n t

b) Volatility

-0.3 -0.2 -0.1 0 0.1 0.2 0.3

C oe ff ic ie n t

c) Turnover by volume

(27)

27

Price movement, price variability and volume variability are major determinants of the market quality parameters of interest. For example, high price variability generally causes high volatility. Furthermore, because volume and the level of prices determine tick sizes according to the tick size table, control variables are also important to control for effects in the quoted bid-ask spread. Turnover by volume is in almost all circumstances affected by these control variables as shown in the regression results. This has probably to do with the fact that turnover by volume includes orders on the buy and ask side, which means that on a highly volatile day, where prices drop, turnover by volume will increase. The log variables for market value and price level generally have an insignificant effect on the dependent variables.

Table 3 shows the fixed effects regression results, so without the Nikkei sample. This regression assesses trends in Euro Stoxx 50 for the same time frames and dependent variables. During the 1-week and 10-week time frames, there are treatment effects for all market quality parameters. The 4-week time frame shows these effects solely for the quoted bid-ask spread.

Table 3: Fixed effects regression results

Regression results are based on three balanced time periods where the MiFID II-introduction is the event. Significant results are highlighted. T-statistics are presented in brackets. *** is significant at 1%, ** is significant at 5%, * is significant at 10%.

Dependent variable Independent

variable

10-week pre- and post-event window 4-week pre-and post-event window 1-week pre-and post-event window

Quoted

bid-ask spread Volatility

Turnover by Volume

Quoted

bid-ask spread Volatility

Turnover by Volume Quoted bid-ask spread Volatility Turnover by Volume Coefficient -0.03* 9E-05 12.7 0.01 -5E-03 -0.05 -0.21 0.03 -20.8

(-1.87) (0.02) (1.30) (0.29) (-0.58) (-0.00) (-0.58) (0.77) (-0.28) Treatment effects -5E-05*** 3E-05*** 0.09*** -8E-05*** 2E-06 0.01 -6E-04*** 5E-05*** 0.30***

(-4.38) (6.31) (9.78) (-3.26) (0.42) (0.65) (-5.93) (4.68) (14.6) Log Price -3E-03* -6E-04 0.30 7E-04 -6E-04 1.91 -0.05 5E-03 4.71

(-1.80) (-0.81) (0.22) (0.12) (-0.56) (0.61) (-0.93) (0.92) (0.46) Log Market Value 0.004* 2E-04 -1.19 -1E-03 7E-04 -0.57 0.04 -4E-03 0.36

(1.88) (0.28) (-0.87) (-0.22) (0.58) (-0.18) (0.70) (-0.82) (0.04) Price Movement -3E-07 -1E-05*** 0.01** -2E-05 1E-05*** -0.03*** 1E-05 3E-06 0.02*

(-0.04) (-3.60) (2.17) (-1.08) (3.48) (4.06) (0.15) (0.47) (1.74) Volume Variability 3E-07* 1E-06*** 0.01*** 6E-07 1E-08*** 0.01*** 2E-06 1E-06*** 3E-03***

(1.94) (22.4) (51.7) (1.90) (15.9) (33.8) (1.25) (10.3) (11.3) Price Variability -8E-06 1E-05 -0.01** 2E-06 -1E-05*** 0.01 1E-04 -1E-05 -0.03*

(28)

28

Figure 4: Fixed effects regression comparison

Panel a) represents fixed effects regression results for quoted bid-ask spread with the middle point exactly 2 years, 1 year, 6 months, 6 weeks, 4 weeks 2 weeks before and the middle point exactly during treatment. Panel b) represents the same for volatility. Panel c) represents the same for turnover by volume after dividing the numbers through its mean as explained in the methodology section. The regressions are based on the same independent variables as in table 2. Coefficients are captured in panel b) of appendix C.

-0.0008 -0.0006 -0.0004 -0.0002 0 0.0002

C oe ff ic ie n t

a) Quoted bid-ask spread

-0.0001 -0.00005 0 0.00005 0.0001 0.00015

C oe ff ic ie n t

b) Volatility

-0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

C oe ff ic ie n t

c) Turnover by volume

(29)

29

Figure 4 shows the comparison of these fixed effects with fictive prior treatment periods. Panel a) shows for both the treatment period and 2 years prior the MiFID II introduction a highly negative coefficient for the 1-week time frame of the quoted bid-ask spread. The coefficient for 2 years prior is around -0.0004 against -0.0006 for the MiFID II introduction. It is clear that these reductions in quoted bid-ask spread are caused externally. The most likely cause for the treatment period is the new tick size regime of MiFID II. Previous literature shows that reduced tick sizes generate lower bid-ask spreads. However, this effect is only visible for a short time period. The 4-week and 10-4-week time frames show a coefficient that is almost nil. This means that the effect of a new tick size regime fades away after a very short period of time.

Volatility is affected in the 1-week and the 10-week time frames. However, panel b) of figure 4 shows many random patterns for volatility. It is unlikely that the introduction of MiFID II had an impact on volatility. Comparing the two regression approaches yields the same result. These regressions are performed on total volatility. Because temporary volatility follows more or less the same trend as total volatility, it is assumed that temporary volatility was not significantly influenced. The lack of effects on volatility could be seen as a positive thing. Volatility is a market quality parameter with often a negative load. Low volatility is generally a good thing whereas high volatility is bad. As stated in the methodology section, volatility could be seen positively. It could reflect innovation and competition. However, this is generally true for permanent volatility. Temporary volatility does not reflect innovation and competition, but usually uncertainty. Because total volatility is affected by temporary volatility rather than permanent volatility, the regression results indicate that MiFID II was introduced without causing high volatility which is a positive development.

(30)

30

However, this will also be the case for the Nikkei sample. For the 1-week time frame, difference-in-difference is the most reliable measurement approach.

Figure 2 shows trading volume separated per exchange. The methodology section stated that MTF’s would probably gain the most from MiFID II. The red line in Figure 2 shows the development of Turqouise, the only MTF in the analysis. The Turquoise trading volume does not show very different behaviour compared to the other trading venues. MiFID II did not affect MTF’s directly after the introduction. However, as these electronic trading venues thrived rapidly after the introduction of the first directive of MiFID, it is possible that the second directive could be a new incentive for growth. It will probably take some time before the effects of MiFID II on MTF’s and SI’s are visible. It is worth keeping an eye on the development of these types of trading venues in the future.

6. Conclusion

This paper addresses the question of whether the introduction of MiFID II had an effect on the market quality of European Stock markets. Three market quality parameters of the Euro Stoxx 50 index are considered. These are: liquidity, volatility and volume. Liquidity is measured by the quoted bid-ask spread. A smaller quoted bid-ask spread generally implies that a stock is more liquid. The results show a big decrease of bid-ask spreads for Euro Stoxx during the first week after the introduction of MiFID II. The decrease is only temporary, because the spreads move back to the old level during the month of January. The introduction of MiFID II consisted of a new tick size regime, which implied a tick size reduction for Euro Stoxx 50 on average. Liquidity only improved for a very small period of time.

Another market quality parameter, turnover by volume, increased after the year turn. However, this effect is most probably caused by the end of the Christmas holidays. After these holidays, trading resumes. The same effect was visible two years prior to the introduction of MiFID II. Next to that, difference-in-difference regression results do not imply very different behaviour compared to other years. Concluding, the introduction of MiFID II did not have a significant effect on turnover by volume.

(31)

31

introduction of the first MiFID directive. This paper is too short-term focused to study the impact of MiFID II on these venues thoroughly.

The introduction of MiFID II did not trigger volatility. Total volatility is relatively stable after the introduction. This could be considered as a positive development. Generally, it means that the introduction of MiFID II went smoothly. Temporary volatility is the biggest determinant of total volatility. Temporary volatility followed the same trend as total volatility. Therefore, one can conclude that temporary volatility also remains quite stable.

Overall, there was a short-term improvement of market quality after the introduction of MiFID II, because liquidity improved. Long-term effects of MiFID II on market quality could not be measured at this point in time. These effects should be considered in the future. However, it is clear that MiFID II added regulation which is complex. Investors and national competent authorities will have to manage the implementation and enforcement of the new set of rules, which could disrupt trading activity in the future. Not all capital market participants were prepared for MiFID II and may require some time to fully adjust to the new regulation. This paper shows that the introduction of new regulation does not necessarily cause shocks in stock markets, even if not everybody is prepared. Although investors and authorities should be prepared for disruptions, it is not definite that markets will face trouble. Next to that, tick size changes have only a short-lasting effect on liquidity. Trading venues should be aware of this.

7. References

Bartram S.M., Brown G., Stulz R.M. (2012). Why are U.S. stocks more volatile? Journal of Finance 67: 1329–1370. doi: 10.1111/j.1540-6261.2012.01749.x

Biais, B., Bisiere, C., & Spatt, C. (2010). Imperfect Competition in Financial Markets: An Empirical Study of Island and NASDAQ. Management Science, 56(12), 2237-2250. doi:10.1287/mnsc.1100.1243

Boneva, L., Linton, O., & Vogt, M. (2016). The effect of fragmentation in trading on market quality in the UK equity market. Journal of Applied Econometrics, 31, 192-213. doi:10.1002/jae.2438

Buti, S., Rindi, B., Wen, Y., & Werner, I. M. (2013). Tick Size Regulation and Sub-Penny Trading. Retrieved from http://ssrn.com/abstract=2324862

Buti, S., Rindi, B., Wen, Y., Werner, I. M., & Consonni, F. (2015). Sub-Penny and Queue-Jumping. Retrieved from

The implications of MIFID II on the quality of European Stock Markets