The Effect of the MiFID I and MiFID II / MiFIR Transparency Requirements on the Volatility: Evidence from European Equity Markets.
BSc. Economics and Business: Finance Bachelor thesis – Final Version
J.M. Meijers 10748997
Thesis Supervisor: Dr. R. C. R. van Lamoen 26 June 2018
Statement of Originality
This document is written by Julian Meijers 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.’
J.M. Meijers 10748997
Abstract
This study examines the effects of pre-and post-trade transparency requirements by analysing the Markets in Financial Instruments Directive (MiFID I). The pre- and post-trade transparency requirements are extended with a renewed directive (MiFID II) and new regulation (MiFIR). We test whether the pre- and post-trade transparency requirements of the old (MiFID I) and new (MiFID II/ MiFIR) regulatory framework structurally lower the implied and realised volatility of the European equity markets. Using the CAC 40, DAX 30 and FTSE 100 market indexes for both frameworks, evidence is provided of a structural decrease of the (implied) volatility after implementation of MiFID I. Against the expectation, a structural increase of the (implied) volatility after implementation of MiFID II / MiFIR is shown. Although not all hypotheses are supported, the expected relationship between market transparency and (implied) volatility is confirmed. However, it is hard to distinguish the ’MiFID’ effects from other effects like the financial crisis and the current economic upturn.
Key words: market transparency, mandatory disclosure, implied volatility, realised volatility, European equity markets, regulatory frameworks.
Table of Contents Table of Contents
1. Introduction ... 4
2. Literature review and Hypotheses Development ... 6
2.1 Theoretical Research on the Effects and Implications of Market Transparency ... 6
2.2 Empirical Research on the effects of Market Transparency ... 7
2.2.1 Research Concerning the Effects of Market Transparency ... 7
2.2.2 Research Concercing MiFID I and MiFID II / MiFIR ... 9
2.3 Expectations and Hypotheses ... 10
3.Overview Regulatory Frameworks ... 11
3.1 Overview MiFID I ... 11
3.2 Overview MiFID II / MiFIR ... 12
4. Data and Methodology ... 15
4.1 Sample periods ... 15
4.2 Descriptive statistics ... 16
4.2.1 Descriptive statistics MiFID I ... 16
4.2.2 Descriptive statistics MiFID II / MiFIR ... 18
4.3 Methodology ... 19
4.4 Models ... 20
4.4.1 Implied Volatility MiFID I Model ... 20
4.4.2 Implied Volatility MiFID II / MiFIR Model ... 22
4.4.3. Realised Volatility MiFID I and MiFID II / MiFIR Model. ... 23
5 Results ... 25
5.1 Results Testing for Stationarity ... 25
5.2 Results Implied Volatility MiFID I Model ... 26
5.3 Results Realised Volatility MiFID I Model ... 29
5.4 Results Implied Volatility MiFID II / MiFIR Model ... 30
5.5 Results Realised Volatility MiFID II / MiFIR Model ... 33
6. Limitations and Futher Research Suggestions ... 35
7. Conclusion ... 37
8. References ... 39
9. Appendix ... 45
Appendix A: Implied Volatility MiFID I Model ... 45
Appendix B: Realised Volatility MiFID I Model ... 47
Appendix C: Implied Volatility MiFID II / MiFIR Model ... 48
1. Introduction
Market transparency is ‘the ability of market participants to observe the information in the trading process’, according to O’Hara (1995). The information refers to public bids, current or past offer prices, quotes and volume or venue of executed transaction. Transparency issues are viewed as fundamental to the competitiveness of financial markets (Bessembinder, Maxwell and Venkataraman, 2006; Chung and Chuwonganant, 2009). Regulations regarding mandatory disclosure are one of the biggest regulatory concerns (Board, Sutcliffe and Wells, 2003). Transparency issues are therefore subject of many (policy) debates because they determine the design and regulation of the market (Chung and Chuwonganant, 2009; Dumitrescu, 2010; Madhavan, Porter and Weaver, 2005). Prior regulatory changes have been made on the belief that transparency improves the market.1 However, conclusions of existing empirical research are controversial (Bessembinder et al., 2006; Dumitrescu, 2010; Eom, Ok and Park, 2007; Kovaleva and Iori, 2015). The empirical, theoretical and experimental existing literature on the effects of information on financial markets reflect the academic interest in market transparency.
In Europe, the pre- and post-trade transparency demands are extended because of the revision of the MiFID I framework, by introducing a renewed directive (MiFID II) and a new regulation (MiFIR). MiFID I and MiFID II / MiFIR aim to assure stability and certainty in financial markets (European Commission, 2014). The little existing literature specified on the impact of these European regulatory frameworks concentrates on the liquidity and efficiency rather than the stability of the market(s). The volatility is considered to be an excellent benchmark for measuring the market’s stability (Kovaleva and Iori, 2015). Although volatility forms an important attribute of the financial markets, other attributes are emphasised in research.
Therefore, the main focus of this study is to examine whether the pre- and post-trade transparency requirements structurally lower the volatility of the European equity markets. The CAC 40, DAX 30 and FTSE 100 market indexes are used for investigation over 2005 till 2014 and 2017 till 2018. Other than prior research, this study analyses the effect of both events (i.e. MiFID I and MiFID II / MiFIR). While prior literature mostly has dealt with the transparency impact on the liquidity or efficiency, this study compares the transparency impact on the implied volatility with the impact on the realised volatility.
1 Examples are the NYSE OpenBook, Toronto Stock Exchange CATS, SEC Rule 605 and SuperMontage
(Boehmer, Saar, and Yu, 2005; Eom et al., 2007; Ferrell, 2007; Lee, Lai and Huang, 2015; Zhao and Chung, 2007).
Although both volatilities represent the variations of the market return, the realised volatility is based on the past variation and the implied volatility is referred as an estimation of future variation (Zhang, 2012 pp. 259-260). Secondly, the structural relation between MiFID I and MiFID II / MiFIR and other market attributes like the price index, bid-ask spread, high-low spread and trading volume is analysed.
The aim of this study is to advance research regarding market transparency effects on the market volatility, by relating volatility changes to implementation of European disclosure laws, which have not been brought into a relation before. Therefore, it provides useful information for the regulators and other entities in charge of the market when revising existing regulations or making recommendations in order to obtain a safe financial system (European Commission, 2014). The study contributes to a better understanding of the potential effects of markets being less or non-transparent. Subsequently, this study elaborates the trade-off for all market participants (e.g. investors and regulators) between advantages and disadvantages of increasing disclosure of information.
This paper is organised in several sections. In the next section the relevant literature is reviewed and the outcomes of similar research are discussed. Based on these findings, the expectations and hypotheses are formulated. The third section elaborates the theoretical background of MiFID and MiFID II / MiFIR. Section four discusses the data and methodology, containing descriptive statistics, the sample periods and the models. Section five discusses the results. The paper concludes with the limitations and suggestions for further research.
2. Literature review and Hypotheses Development
To provide a clear overview of earlier research regarding market transparency, the theoretical and empirical literature are discussed in the following sections. Based on these findings, the expectations and hypotheses are formulated.
2.1 Theoretical Research on the Effects and Implications of Market Transparency
In this part, we discuss the general principles and implications regarding market transparency. Transparency has multiple dimensions, generally it is defined as the amount of trading information that is made publicly available on a timely basis before or/and after a completed transaction (Dumitrescu, 2010). Pre-trade transparency concerns to the ability to observe order flow, quotes, and the identity of market participants (Boehmer et al., 2005). Post-trade transparency can be interpreted as the ability to observe execution quality, including execution prices (i.e. bid, ask and depth), price improvements, execution speed, and fill rate (Dumitrescu, 2010). Markets that disseminate little or no price data are referred to as being non-transparent, or opaque (Coslor, 2016).
Market transparency could have multiple advantages. Increasing transparency could reduce the information asymmetries, which means that more investors have the same access to relevant information (Coslor, 2016). This could result in an increasing participation of different investors, assumed that investors previously being reluctant because of uncertainties and high search and market costs. Secondly, increasing transparency could improve the ability to monitor performance and assess risks of markets. Market participants can more easily verify receipt of best execution practices, by comparing the prices of executed transactions (Bortoli, Frino, Jarnecic, and Johnstone, 2006, p.1165). On top of that, the ability to detect fraud, manipulation, unfair pricing, and other market abuses could be enhanced (Greenstone, Oyer and Vissing-Jorgensen, 2006). Fourthly, market transparency could improve efficiency and robust price formation, caused by information being more rapidly disseminated and incorporated in the market (Boehmer et al., 2005). Furthermore, transparency strengthens competition between markets by encouraging to post better prices in order to attract order flow (Bloomfield and O’Hara, 1999).
On the other side, market transparency could have disadvantages. First, there could be costs associated with collecting, disclosing and the classification of information (Hendershott and Jones, 2005). On top of that, it requires effort, time and financial resources. According to Admati and Pfleiderer (2000), direct costs are associated with producing and disseminating information when information has to be disclosed or certified
by a third party. However, information asymmetry could also lead to additional costs for all market participants, including losses caused by underinvestment and costs that arise from illiquidity in the market (Admati and Pfleiderer, 2000). Nevertheless, mostly firms do not disclose more than regulation requires. Other market participants therefore could free-ride, resulting in a potential underproduction of information (Beyer, Cohen, Lys and Walther, 2010, p.297). Regulatory frameworks to prevent such market failures are implemented to increase the disclosure of information.
Summarised, increasing market transparency could lead to increasing participation, improved ability to monitor performance and assess risks, and improved efficiency and robust price formation. Although there could be costs associated with increasing transparency, markets could be prevented from failures.
2.2 Empirical Research on the effects of Market Transparency
In the first section research regarding the transparency effects (of related regulatory frameworks) is discussed. In the second section research with a specific focus on MiFID I and MiFID II / MiFIR is discussed.
2.2.1 Research Concerning the Effects of Market Transparency
First the studies providing evidence of positive effects of transparency on the market are discussed. Boehmer et al. (2005) provide evidence of increasing market efficiency and liquidity by investigating NYSE's OpenBook service that provides limit-order book information to traders. Bessembinder et al., (2006) test institutional trades in corporate bonds before and after the initiation of public transaction reporting through the TRACE system and find a significant decline in the trade execution costs. Zhao and Chung (2007) analyse the impact of the SEC Rule 605 using data of NYSE, AMEX, and NASDAQ stocks and provide evidence of higher liquidity and lower return volatility after the implementation of the rule.2 Ferrell (2007) examines the impact of the U.S. 1964 imposition of mandatory disclosure requirements on over-the-counter (OTC) markets in terms of volatility.3 Evidence is provided of a reduction in volatility among OTC securities compared to NYSE stocks that are already subject to SEC disclosure requirements, caused by information being more quickly reflected in prices. This is consistent with the empirical
2 The SEC Rule 605 requires Market centers that trade national market system (“NMS”) securities to make
monthly electronic disclosures of information regarding execution quality on a stock-by-stock basis (Zhao and Chung, 2007, p. 658).
3 OTC trades are defined as contract where the execution of which does not take place on a regulated market
evidence of Aggarwal and Wu (2006) who show that less transparent markets, such as OTC markets in the United States, face a higher volatility caused by stock market manipulation. Hendershott and Jones (2005) investigate the Island electronic communication network (ECN) which stopped displaying its limit order book. They conclude that market liquidity and efficiency worsen when the transparency decreases. Chung and Chuwonganant (2009) examine the effect of pre-trade transparency on market quality using data before and after the introduction of SuperMontage and find a decline in quoted and effective spreads after the implementation.4 Lee, Lai and Huang (2015) investigate the relation between information transparency and idiosyncratic risk using data of listed companies violating the information disclosure rules. They conclude that companies with worse information transparency have a higher idiosyncratic risk than other companies. Eom et al. (2007) show, by investigating two discrete changes in pre-trade transparency on the Korea Exchange (KRX), that both events improve the market efficiency and liquidity. Lang, Lins and Maffett (2012) examine the relation between firm-level transparency, stock market liquidity and valuation across countries. They find lower transaction costs and greater liquidity for firms with greater transparency.
Other studies however, show negative effects of market transparency. Madhavan et al. (2005) investigate the Toronto Stock Exchange when it publicly disseminated the limit order book on both the traditional floor and on its automated trading system. They find that increasing transparency reduces liquidity and increases execution costs and volatility. Goldstein, Hotchkiss and Sirri (2007) assess the impact of last-sale trade reporting on the liquidity of BBB corporate bonds and show that the liquidity increases for very large trades when prices become more transparent. Kovaleva and Iori (2015) show that an opaque market structure preserves market stability better than any other structure, based on volatility of returns being the lowest in this structure.
On top of that, some studies find no significant effects. Angel and Roberto (2014) cannot show that the disclosure of relevant information does contribute to volatility changes in the short term. Anand and Weaver (2004) examine the effects of the abolition of the hidden limit orders in the Toronto Stock Exchange and find no significant changes in quoted spreads and quoted depths.
4 SuperMontage is a fully integrated order display and execution system for Nasdaq-listed securities (Chung
2.2.2 Research Concercing MiFID I and MiFID II / MiFIR
As previously described, research specified on the effects of MiFID I and MiFID II / MiFID is sparse. Paulo (2018) examines whether the proliferation of alternative trading venues in Western Europe after the MiFID I implementation affected the market quality. By investigating the changes in market quality of stocks that initiated trading in Multilateral Trading Facilities (MTFs), Paulo finds a significant overall liquidity increase in the short term.5 Transparency issues regarding dark pool trading are an important concern of MiFID II / MiFIR.6 Empirical research addresses these concerns. Buti, Rindi, and Werner (2017), show that dark pool trading results in illiquid markets. This is consistent with the empirical evidence of Degryse, Jong and Kervel (2015), who show that dark trading lowers liquidity in the financial markets. From the high-frequency trading point of view, Brogaard, Hendershott and Riordan (2014) find, against the argumentation of MiFID II / MiFIR, that high-frequency trading facilitates price efficiency by trading in the direction of permanent price changes and in the opposite direction of transitory pricing errors.
Summarised, the findings of existing empirical research are contradictory. Although it is shown that market transparency could have negative effects, most studies that analyse disclosure laws similar to MiFID I and MiFID II / MiFIR, find evidence of positive effects on the market (Boehmer et al., 2005; Eom et al., 2007; Ferrell, 2007; Zhao and Chung, 2007). Existing research with a specific focus on the effects of MiFID I and MiFID II / MiFIR on the market volatility is missing. Concluding, existing literature related to transparency implications provides useful insights, but should be extended.
5 Multilateral trading facility (MTF) is a multilateral system, operated by an investment firm or a market
operator, which brings together multiple third-party buying and selling interests in financial instruments (European Parliament and of the Council, 2004, Article 4).
6 Dark pools are Alternative Trading Systems that do not provide their best-priced orders for inclusion in the
2.3 Expectations and Hypotheses
Similar to the aim of MiFID I, it is expected that the renewed pre- and post-trade transparency requirements increase the overall stability of the European equity markets. It is expected that these requirements impact the implied volatility and realised volatility similarly. Despite the contrasting theories7, it is expected that the increased transparency of MiFID I structurally lowers the implied and realised volatility of the financial equity markets, which corresponds to the findings of Boehmer et al. (2005), Eom et al. (2007), Ferrell (2007) and Zhao and Chung, (2007).
Hypothesis 1: MIFID I structurally lowers the volatility of the financial equity markets. On top of that, it is investigated whether MiFID I impacts the effect of other market attributes on the volatility. It is expected that there is structural relation between MiFID I and the price index, bid-ask spread, high-low spread and trading volume.
Hypothesis 2: MIFID I structurally impacts the effects of the price index, bid-ask spread, high-low spread and trading volume on the volatility of the financial equity markets.
The expectations for MiFID II / MiFIR are equal. Subsequently, it is expected that the increased transparency of MiFID II / MiFIR structurally lowers the implied and realised volatility of the financial equity markets.
Hypothesis 3: MIFID II / MIFIR structurally lowers the volatility of the financial equity markets.
Secondly, it is expected that there is structural relation between MiFID II / MiFIR and the price index, bid-ask spread, high-low spread and trading volume.
Hypothesis 4: MIFID II / MIFIR structural impacts the effects of the price index, bid-ask spread, high-low spread and trading volume on the volatility of the financial equity markets.
7 First, with market transparency, market participants are better informed. This could increase the (implied)
volatility when more market participants receive (bad) information (Ferrell, 2007). Furthermore, market participants could have the tendency to imitate each other, which is known as herding behaviour (Bagella, Becchetti and Hasan, 2006, p. 131). If more imitation orders the market, market transparency could result in a volatile market where additional information could trigger big reactions.
3. Overview Regulatory Frameworks
The Overview of the regulatory frameworks is divided into an overview of MiFID I and MiFID II / MiFIR.
3.1 Overview MiFID I
Markets in Financial Instruments Directive (MiFID I) is a European comprehensive regulatory regime, introduced in April 2004 and implemented in November 2007. MiFID I improved the cross-border investment guidelines under the Investment Services Directive (1993). MiFID I encompasses regulated markets, MTFs and investment firms
The construction of MiFID I is based on a four-level legislative approach for the design, implementation and enforcement of the framework. In Level 1, the legislative framework principles were approved by the European Parliament and Council in 2004. Level 2 refers to the technical implementation of the framework, which were adopted after submission to the European Securities Committee (ESC). Level 3 facilitates a coherent implementation and uniform application of the regulatory framework among the member states. Level 4 refers to addressing inconsistent or failing implementation of the legislative framework (European Parliament and of the Council, 2004). The objective of MiFID I is to create a transparent, efficient, competitive and integrated European financial market that ‘makes the most of new ways of trading’ by ‘trade securities at maximum efficiency and at minimum cost’ (European Commission, 2002). This had to be realised by providing a regulatory environment which protect investors and allows for the creation of new markets and trading platforms under a harmonised legal framework (European Commission, 2002).
Regulated markets (Article 44), MFTs (Article 29) and systematic internalisers8 that are traded on a regulated market for which a liquid market exists (Article 27)9 are required to provide pre-trade information for securities traded on regulated markets. This information contains current bid and offer prices and the depth of trading interest at those prices for securities traded during normal trading hours. Second, these trading venues are required to make public post-trade information such as the price, volume and the time at which they were concluded (Article 45, Article 30 and Article 28). This mandatory disclosure was supposed to contribute to higher market liquidity, stability and greater transaction volume. MiFID I reshaped the European markets by fostering the competition
8 Systematic internaliser is an investment firm which executes client orders outside a regulated market or an
MTF (European Parliament and of the Council, 2014, Article 4).
9 Shares with a free float of more than €500 million, and an average daily turnover higher than €2 million
between trading venues. Simultaneously, developments in new technologies have enabled the use of automated and fast trading technologies. Together, this resulted to a highly fragmented equity market in Europe (ESMA, 2016, pp. 5-12).
In July 2010, the Committee of European Securities Regulators (CESR) provided consultation paper containing a review of MiFID I and technical advice to extent the transparency requirements. CESR provided evidence of market failures post-crisis and concluded that additional post-trade transparency was needed (CESR, 2010, p. 33). In October 2011, the European Commission adopted the proposal for the revision of MiFID I by extending the transparency requirements to non-equity instruments (i.e. bonds and derivatives) traded on regulated markets, MTFs, organised trading facilities (OTFs)10 and OTC markets (European commission, 2011, pp. 2-6). In June 2014, the European Parliament and Council adopted new rules revising the MiFID I framework, by introducing a renewed directive (MiFID II) and a new regulation (MiFIR) which are implemented in January 3, 2018 (European Parliament and of the Council, 2018).
3.2 Overview MiFID II / MiFIR
The new rules of MiFID II / MiFIR fundamentally change the European trading and market infrastructure. According to the European Comission (2016), the main objectives of MiFID II / MiFIR are to realise a more transparent, responsible and safer financial system, to meet the G20 commitments,11 to address less regulated and opaque of the financial system, to improve the organisation and transparency of markets, including markets where instruments traded OTC, to enhance the oversight and transparency of commodity derivative markets and to meet new developments in market structures and technology (i.e. dark trading, algorithmic trading, and high frequency trading). This study focusses on the transparency implications, therefore, the key concepts of MiFID II / MiFIR regarding transparency requirements are discussed in detail.
First, a market structure framework that ensures trades to take place on regulated platforms is introduced. Investment firms operating an internal matching system executing client orders in shares, depositary receipts, exchange-traded funds, certificates and other similar financial instruments on a multilateral basis have to be authorised as a MTF. In addition, a new category of trading venues for non-equity instruments to be
10 Organised trading facility (OTF) is a multilateral system which is not a regulated market or an MTF and
in which multiple third-party buying and selling interests financial instruments (MiFID II, Article 4(23)).
11 G20 Commitments, made by the G20 group of countries, committed to improve the over-the-counter
derivatives market through reporting over-the-counter derivatives to Trade Repositories, increasing the use of central counterparties, executing of over-the-counter derivatives on electronic trading platforms and increasing the use of collateral and risk mitigation techniques (MiFID II, subsection 25-26)
traded on a multilateral platform is constructed, called the OTF (MiFIR, Subsection 9). New types of organised execution for bonds, structured finance products, emission allowances, derivatives and other types of organised execution that might develop in the
future, but fall outside RMs and MTFs, are captured by the OTF. Furthermore, the pre- and post-trade transparency requirements are extended in
multiple ways.12 First, MiFIR extends the requirements to a wider range of financial instruments. Under MiFID II / MiFIR, ‘Equity-like’ instruments such as depositary receipts, exchange-traded funds, certificates (MiFIR, Article 3 and Article 6) and non-equity instruments, like bonds, structured finance products, derivatives and emission allowances (MiFIR, Article 8 and Article 10) are subjected to the requirements. Secondly, MiFIR extends transparency to a wider range of trading venues. Requirements therefore not only apply to instruments traded on regulated markets, MTFs or systematic internalisers, but also to those traded on organised trading facilities (OTFs). Thirdly, the introduction of a double volume cap mechanism limits the use of reference price waivers and negotiated price waivers in an equity instrument (4% per venue cap and 8% global cap).13 Regarding dark pool trading, the double volume cap mechanism requires that a specific dark pool has to be suspended from trading if the proportion of total value traded on that dark pool exceeds 4% of the total trading volume in the respective instrument (MiFIR, Article 5).14 In case of all dark pools, the same is required if the proportion of the total value traded on all total dark pools under these waivers exceeds 8% of the total trading volume in the respective instrument. Due to limiting the amount of orders executed in dark pools, the price discovery process on public markets is protected. Lastly, consolidated tape providers (CTPs), approved publication arrangements (APAs) and approved reporting mechanisms (ARMs) are introduced to address transparency issues related to market fragmentation and OTC trading (MiFIR, Article 22). CTPs collect trade reports for financial instruments and consolidate them into a continuous electronic live data stream providing price and volume data (MiFID II, Article 4(53) & Article 65). APAs publish post-trade reports on behalf of investment firms that conduct transactions in
12 According to MiFIR, the content of the pre-trade disclosure includes public bids, offer prices and depth of
trading at those prices and post-trade disclosure includes the price, volume, time and venue of executed transactions (MiFIR, Article 6 and Article 8). Post-trade information has to be disseminated 15 minutes after the execution of the transaction, with the possibility of deferred publication or volume masking as appropriate (MiFIR, Article 7).
13 Negotiated Trade Waiver are systems that formalise negotiated transactions, which takes place at or within
the current volume-weighted spread. Reference Price Waiver are systems where the price is determined by reference to a price generated by another system and the reference price is widely published and regarded generally by market participants as a reliable reference price (MiFIR, Article 4).
14 Dark pools are Alternative Trading Systems that do not provide their best-priced orders for inclusion in the
bonds, structured finance products, emission allowances and derivatives traded on a trading venue (MiFIR, Article 21 & Article 22). ARMs are authorised by the European Securities and Market Authority (ESMA) to collect investment firm trade transaction reports.
Third, new trading obligations for derivatives (Article 28, MiFIR) and equities (Article 23, MiFIR) are introduced to improve efficient competition between trading venues. To meet the G20 commitments, every trading venue has non- discriminatory and transparent access to trading venues and central counterparties (CCPs) (MiFIR, Recital 28).15 The trading obligation for shares aims to limit equity trading conducted OTC, by stimulating trading to take place on regulated trading venues and on platforms of Sis to increase transparency and to improve the quality of the price discovery process (MiFIR, Recital 11).
Beside these key concepts, MiFID II / MiFIR specified trading controls to reduce systematic risk arising from algorithmic trading activities (MiFID, Article 17). Furthermore, it strengthened supervisory powers (MiFID II, article 69), harmonised position limits regime for commodity derivatives (MiFID II, article 56) and introduced a minimum common regulatory framework that ensure certainty and uniform treatment of third-country firms accessing EU markets (MiFIR, Subsection 41).
15 Central counterparties stand between the two counterparties to a (derivatives) contract to reduce systemic
4. Data and Methodology
This part elaborates the data and methodology. First the sample periods are discussed, followed by the description of the statistics. Second, the methodology and the models are provided.
4.1 Sample Periods
Theories in similar research regarding the choice of the sample periods emphasise the importance of a long post-event period. Boehmer et al. (2005) state that involved parties will not change their (trading) strategies before the implementation of an event (i.e. NYSE OpenBook), however, a long time is needed to accommodate the change of an event afterwards.
First, the sample window to analyse MiFID I is discussed. The pre-event period consists of 2 years before the implementation of MiFID I. To be able to eliminate the impact of the financial crisis of 2007-2008 on the implied volatility, the post-event window consists of data two years after the crisis. Even though, not all European markets are affected at the same time according to Dajcman, Festic and Kavkler (2012), it is assumed that the major effects of the financial crisis on the European equity markets started in 200816 and ended in 2012.17 Hence, the pre-event period from November 1, 2005 till October 31, 2007 and the post-event period from November 1, 2007 till December 31, 2014 are used to analyse MiFID I.
For MiFID II / MiFIR. preparations for the implementation had been going on since 2011 (ESMA, 2018). Due to significant challenges to meet the MiFID II rules, the implementation of MiFID II / MiFIR was delayed (ESMA, 2015). Subsequently, possible announcement effects or changes in trading protocols to meet the new rules could had been taken place. However, significant changes are assumed to occur only after the new rules have been implemented. As stated before, it is assumed that a long post-event period is needed to accommodate the effect of an event (i.e. at least one year). However, only 5 months of data after the implementation is available. Hence, the pre-event period from January 1, 2017 till December 31, 2017 and to the post-event period from January 3, 2018 till May 31, 2018 are used to analyse MiFID II / MiFIR.
16 Lehman Brothers shocked all capital markets worldwide by filing for Chapter 11 bankruptcy protection on
September 15, 2008 (SEC, 2008). However, earlier signs bank failures and financial market pressures started at the end of 2007 (SEC, 2007).
17 Signs of the recovery of EU’s economy started in 2011 and continued the following year (European
4.2 Descriptive statistics
The constituents of the CAC 40, FTSE 100 and DAX 30 market indexes are used for analysis, like earlier research did when measuring the volatility of European stock markets (Ozenbas, Schwartz and Wood, 2002; Charles, 2010). These three market indexes represent respectively the 40, 100 and 30 highest market capitalisation traded stocks on their European trading venue. Where CAC 40 is being traded on the Euronext Paris, FTSE 100 on the London Stock Exchange and DAX 30 on the Frankfurt Stock Exchange. The data is collected from Data stream (Thom Reuters), which consists of daily closing prices, bid-ask spreads, high-low spreads and trading volumes. For the implied volatility data, the volatility indexes of the CAC 40, FTSE 100 and DAX 30 market indexes are used, containing daily expectations of 30-day volatility. All data is filtered for unlikely values (outliers).
4.2.1 Descriptive statistics MiFID I
The summary statistics of the data, regarding MiFID I, are provided below.
Table 1: Summary statistics of the market indexes
Market Spread (bid-ask) Spread (high-low) Trading volume Price index Pre Post Post* Pre Post Post* Pre Post Post* Pre Post Post* CAC 40 -0.86 2.18 -0.02 48.93 66.06 49.28 114.39 142.26 122.88 4931.89 3836.39 3854.51 DAX 30 -4.53 1.57 0.63 62.95 114.03 102.54 118.43 132.92 97.95 5805.23 6909.76 8851.76 FTSE 100 -2.76 0.10 -1.36 57.23 87.19 64.00 463.13 494.32 450.12 5738.05 5705.79 6282.72 -2.72 1.28 -0.25 56.37 89.09 71.94 231.98 256.50 223.65 5491.72 5483.98 6329.66
Summary statistics for all three datasets; CAC40, DAX30, LFTE 100, during the pre-event period (01/11/2005-31/10/2007), the post-event period (01/11/2007-31/12/2014) and the post*-event period (01/01/2013-31/12/2014). Where the values are averages. Note that the trading volume is given in millions and the post* period (01/01/2013-31/12/2014) equals the post-event period without the crisis period.
It can be concluded that on average, the average bid-ask spreads, high-low spreads and trading volume are higher in the post-event period (Table 1). The average price index is slightly lower in the post-event period. Remarkable is the big decrease in the post-event period of the CAC 40 price index in contrast to the big increase of the DAX 30 price index. This difference is even bigger when analysing the price index of both indexes in the post*-event period. Both the average bid-ask spreads and high-low spreads are slightly higher
in the post*-event period, compared to the pre-event period. There are no common differences in the post*-event trading volumes across the indexes.
The summary statistics of the volatility index for the MiFID I sample period are reported below (Table 2). Figure 1 graphically illustrates these statistics.
Table 2: Summary statistics volatility indexes
Volatility index
Pre Post Post*
CAC 40 17.00 24.63 16.72 DAX 30 17.83 24.57 16.99 FTSE 100 15.53 21.70 13.76 Average 16.79 23.64 15.82
Summary statistics volatility indexes for all three datasets; CAC40, DAX30, LFTE 100. Where the pre, post and post* values are averages. Note that the post* period (01/01/2013-31/12/2014) equals the post-event period without the crisis period.
Figure 1: Volatility indexes during sample period of MiFID I
Volatility indexes of the CAC 40, DAX 30 and FTSE 100 during sample period of MiFID I. Note that the first vertical line indicates the implementation date of MiFID I and the second vertical line indicates the start date of the post* event period.
All three volatility indexes depicted in Figure 1 show an increase after the implementation of MiFID I. The post-event period contains the three highest volatility peaks of the sample. This translates into a higher average volatility index in each post-event period (Table 2). These findings could be explained by the assumption that during extreme events, like the financial crisis in the post-event period, the (implied) volatility is higher because of increasing uncertainty and risk (Hassan and Wu, 2015). Furthermore, all three volatility
indexes represent a decrease starting at the end of 2011. This reflects the lower average volatility index in each post*-event period.
4.2.2 Descriptive statistics MiFID II / MiFIR
The summary statistics of the data, regarding MiFID II / MiFIR, are provided below. Table 3: Summary statistics of the market indexes
Market Spread (bid-ask) Spread (high-low) Trading Volume Price index
Pre Post Pre Post Pre Post Pre Post
CAC 40 -1.11 -1.33 47.60 51.03 91.63 85.42 4912.96 5371.06 DAX 30 -2.81 7.57 101.46 151.50 87.25 104,51 12373.26 12652.36 LFTE 100 -1.65 -0.52 56.12 68.42 783.03 814.93 7442.46 7443.45 Average 3.98 5.72 68.39 90.32 320.64 334.95 8242.89 8488.96
Summary statistics for all three datasets; CAC40, DAX30, LFTE 100, during the pre-event period (01/01/2017- 31/12/2017) and the post-event period (01/01/2018-31/05/2018). Where the pre- and post- values are averages. Note that the trading volume is given in millions.
It can be concluded that the average high-low spread and price index is slightly higher in each post-event period. On average, the average volume and bid-ask spread of the three indexes are higher in the post-event period (Table 3), however there are no common differences across the indexes. The average bid-ask spread is slightly lower in the
post-event period.
The summary statistics of the volatility index for the MiFID I sample period are reported below (Table 4). Figure 2 graphically illustrates these statistics.
Table 4: Summary Statistics Volatility Indexes
Volatility index Pre Post CAC 40 14.14 14.76 DAX 30 14.26 17.32 FTSE 100 10.88 12.90 Average 13.09 14.99
Summary statistics volatility indexes for all three datasets; CAC40, DAX30, LFTE 100. Where the pre and post values are averages.
Figure 2: Volatility indexes during sample period MiFID I
Figure 2: Volatility indexes of the CAC 40, DAX 30 and FTSE 100 during sample period MiFID II / MiFIR. Note that the first vertical line indicates the implementation date of MiFID II / MiFIR.
All three volatility indexes depicted in Figure 2 show an increase after the implementation of MiFID II / MiFIR. Furthermore, the post-event period contains the highest volatility peak of the sample. These findings reflect the higher average volatility index in each post- event period compared to the pre- event period (Table 4).
4.3 Methodology
This research examines whether the implementation of MiFID I and MiFID II / MiFIR structurally change the volatility of European financial markets.
Most earlier research methodologies are based on event studies. This methodology examines the effects of an event by comparing the market before the announcement or implementation of the event in the pre-event period, with the market afterwards in the post-event period (Eom et al., 2007; Ferrell, 2007; Madhavan et al., 2005). The theoretical basis for interpreting volatility is the variance-bound model, which indicates that lower volatility is consistent with increased stock price accuracy (Ferrell, 2007, p. 216). Early empirical studies have indicated that implied volatility could be informationally superior to historical volatility (Blair, Poon and Taylor, 2001; Giot, 2003). This research uses the implied and realised volatility for analysis. The implied volatility is referred as the market's volatility forecast and is said to be forward looking (Dufour, Garcia and Taamouti, 2012). To ensure structural validity of the outcome of this research, a robustness test is applied by redoing the estimation process using the realised volatility instead of the volatility index for analysis.
4.4 Models
Fluctuations of the volatility may not only result from the impact of MiFID I or MiFID II / MiFIR, therefore, control variables are added to the models to determine the effect on the implied volatility more accurately. The models can be estimated by applying an ordinary least square regression (OLS). OLS is a widely used method for measuring a linear relationship between assets or return distributions assuming normal distributed errors. Complications could arise when using OLS if variable(s) is (are) correlated with the error term, which is known as endogeneity (Klomp and De Haan, 2009). In addition, the variables are first tested for multicollinearity. The volatility could be affected by rare or extreme events (Trapin, 2018). This results in volatility skew patterns, making OLS regression inappropriate to use (Liu et al., 2005). On top of that, the volatility could by determined by the past level of an explanatory variable. Therefore, the use of lagged values could be more appropriate. The optimal number of lags for each variable is determined using the Schwarz information criterion (Schwarz, 1978).18
4.4.1 Implied Volatility MiFID I Model
For the dependent variable, the three volatility indexes (i.e. CAC 40, FTSE 100 and DAX 30) are used, resulting in three regression models. Regarding the independent variables, the control variables are discussed first.
It is assumed that the price index has a significant negative relationship with the (implied) volatility (Dennis et al. 2006; Giot, 2005; Dufour et al., 2012). Earlier research has shown that the relationship between the (implied) volatility and stock index returns is asymmetric (Hibbert et al., 2008).19 To eliminate this, the variable is transformed by taking the log. Evidence of correlation between the (implied) volatility and the lagged price index is provided by Bollerslev, Litvinova & Tauchen (2006). Subsequently, different number of lags are tested for significant effects. Secondly, the trading volume is assumed to have a positive relationship with the (implied) volatility, which corresponds to the findings of earlier research (Wang and Wu, 2015). Chiang, Qiao and Wong (2010) show that lagged values of trading volume could impact the (implied) volatility. Subsequently, different number of lags are tested for significant effects.
18 The Schwarz information criterion is, in contrast to other criterions like the Akaike-criterion, the
Hannan-Quinn-criterion and the final prediction error, strongly consistent in higher dimensions of the lags when the sample gets larger (Lütkepohl, 2005, p.150)
19 An asymmetric relationship implies that the negative change in the stock market has a higher impact on
Thirdly, earlier research has shown the bid-ask spread to be significantly related to the (implied) volatility (Frank and Garcia, 2011). Furthermore, Corrado and Truong (2007) demonstrate that the intra-day high-low price range is an important attribute of the (implied) volatility. Based on the findings of earlier research, only the present values of both spreads are assumed to have a significant positive relation with the (implied) volatility (Corrado and Truong, 2007; Frank & Garcia, 2011). Subsequently, the price index, trading volume and the bid-ask spread together with the high-low spread are added to the model. A dummy variable is created to eliminate the potential effect of the financial crisis of 2007-2008 on the implied volatility. As described above, the crisis period is assumed to start in 2008 and end in 2012. The Crisis Dummy variable is equal to 1 for the volatility index observed between the crisis time window January 1, 2008 till December 31, 2012 and equal to zero otherwise.
The potential effect of MiFID I is tested by implementing a dummy variable. The MiFID I Dummy variable equals 1 for the volatility index observed after November 1, 2007 and equals zero otherwise. Interaction terms are created by multiplying the MiFID I dummy variable with the control variables, to test whether MiFID I impacts the effect of other market attributes on the volatility. Still, the implied volatility of financial markets could depend on multiple other uncontrollable factors, like the quality of the information provided, monitoring and controlling mechanisms, which could make it hard to advocate the effect of the implementation of MIFID II / MiFIR. However, these factors are assumed to have a small effect on the implied volatility. The model is provided below. Where Spread(
is equal to the bid-ask spread, Spread) is equal to the high-low spread.
𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦𝑖𝑛𝑑𝑒𝑥5= 𝛽8+ : 𝛽( ; <=( log 𝑃𝑟𝑖𝑐𝑒𝑖𝑛𝑑𝑒𝑥5D<+ : 𝛽) ; <=( 𝑇𝑟𝑎𝑑𝑒𝑣𝑜𝑙𝑢𝑚𝑒5D<+ 𝛽I𝑆𝑝𝑟𝑒𝑎𝑑(5+ 𝛽L𝑆𝑝𝑟𝑒𝑎𝑑)5 + 𝛽N𝑀𝑖𝐹𝐼𝐷𝐼5+ 𝛽S𝐶𝑟𝑖𝑠𝑖𝑠 + : 𝛽V ; <=( 𝑀𝑖𝐹𝐼𝐷𝐼5∙ log 𝑃𝑟𝑖𝑐𝑒𝑖𝑛𝑑𝑒𝑥5D<+ : 𝛽X ; <=( 𝑀𝑖𝐹𝐼𝐷𝐼5∙ log 𝑇𝑟𝑎𝑑𝑒𝑣𝑜𝑙𝑢𝑚𝑒5D< + 𝛽Y 𝑀𝑖𝐹𝐼𝐷𝐼5∙ 𝑆𝑝𝑟𝑒𝑎𝑑(5+ 𝛽(8𝑀𝑖𝐹𝐼𝐷𝐼5∙ 𝑆𝑝𝑟𝑒𝑎𝑑)5+ 𝜀5 𝑀𝑖𝐹𝐼𝐷 𝐼= 1, 𝑓𝑜𝑟 𝑡 = 𝐴𝑓𝑡𝑒𝑟 𝑁𝑜𝑣𝑒𝑚𝑏𝑒𝑟 1, 2007 𝑎𝑛𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒. 𝐶𝑟𝑖𝑠𝑖𝑠= 1, 𝑓𝑜𝑟 𝑡 = 𝐽𝑢𝑛𝑒 1, 2007 − 𝐷𝑒𝑐𝑒𝑚𝑏𝑒𝑟 31, 2012 𝑎𝑛𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒.
The structural change of the volatility is measured by testing the variables regarding MiFID I (i.e. Dummy and interaction terms) to have a jointly significant impact on the implied volatility.
4.4.2 Implied Volatility MiFID II / MiFIR Model
The variable choice of the MiFID II/MiFIRmodel is based on variables used in the MiFID I model. Hence, for the dependent variable, the volatility indexes (i.e. CAC 40, FTSE 100 and DAX 30) are used, resulting in three regression models. The price index, trading volume, bid-ask spread and high-low spread are added to the model as control variables. Again, the log of the price index is used to eliminate potential biases in estimation. Different number of lags of the price index and trading volume are tested for significant effects. The potential effect of MiFID II /MiFIR is tested by implementing a dummy variable. The MiFID II Dummy variable is equal to 1 for the volatility index observed after January 3, 2018 and equal to zero otherwise. Additionally, interaction terms are created by multiplying the MiFID I dummy variable with the control variables, to test whether MiFID I impacts the effect of other market attributes on the volatility. Still, the implied volatility of financial markets could depend on multiple other uncontrollable factors, like the quality of the information provided, monitoring and controlling mechanisms, which could make it hard to advocate the effect of the implementation of MIFID II / MiFIR. However, these factors are assumed to have a small effect on the implied volatility. The model is provided below. Where Spread( is equal to the bid-ask
spread, Spread) is equal to the high-low spread.
𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦𝑖𝑛𝑑𝑒𝑥5= 𝛽8+ : 𝛽( ; <=( log 𝑃𝑟𝑖𝑐𝑒𝑖𝑛𝑑𝑒𝑥5D<+ : 𝛽) ; <=( 𝑇𝑟𝑎𝑑𝑒𝑣𝑜𝑙𝑢𝑚𝑒5D<+ 𝛽I𝑆𝑝𝑟𝑒𝑎𝑑(5+ 𝛽L𝑆𝑝𝑟𝑒𝑎𝑑)5 +𝛽N𝑀𝑖𝐹𝐼𝐷𝐼𝐼 + : 𝛽S ; <=( 𝑀𝑖𝐹𝐼𝐷𝐼𝐼 ∙ log 𝑃𝑟𝑖𝑐𝑒𝑖𝑛𝑑𝑒𝑥5+ : 𝛽V ; <=( 𝑀𝑖𝐹𝐼𝐷𝐼𝐼 ∙ 𝑇𝑟𝑎𝑑𝑒𝑣𝑜𝑙𝑢𝑚𝑒5+ 𝛽X𝑀𝑖𝐹𝐼𝐷𝐼𝐼 ∙ 𝑆𝑝𝑟𝑒𝑎𝑑(5+ 𝛽Y𝑀𝑖𝐹𝐼𝐷𝐼𝐼 ∙ 𝑆𝑝𝑟𝑒𝑎𝑑)5+ 𝜀5 𝑀𝑖𝐹𝐼𝐷𝐼𝐼= 1, 𝑓𝑜𝑟 𝑡 = 𝑎𝑓𝑡𝑒𝑟 𝑗𝑎𝑛𝑢𝑎𝑟𝑦 3, 2018 𝑎𝑛𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 .
The structural change of the volatility is measured by testing the variables regarding MiFID II/ MiFIR (i.e. Dummy and interaction terms) to have a jointly significant impact on the implied volatility.
4.4.3. Realised Volatility MiFID I and MiFID II / MiFIR Model.
The realised volatility is often measured by using the standard deviation, however this method could have some limitations (Klomp and De Haan, 2009). The standard deviation does not take the average growth of the variable into account, while the average growth and variability are likely to be jointly determined (Mobarak, 2005). Therefore, the relative standard deviation is used is dependent variable. The relative standard deviation is defined as follows.
𝜎5= 1 |𝑃𝐼ooop|
q∑(𝑃𝐼5− 𝑃𝐼ooop))
𝑛 − 1
Where 𝜎5 is the realised volatility, 𝑃𝐼5 is the price index at time t, 𝑃𝐼ooop is the average price index over the whole period 𝑇 and 𝑛 is the number of observations over the period 𝑇. The relative eliminates the impact of potential growth in the price index by dividing the standard deviation by the mean of the price index (Klomp and De Haan, 2009).
Again, the price index, bid-ask spread, high-low spread and trading volume are added to the models as control variables. To analyse MiFID I, the methodology provided in Section 5.4.1. is used. The potential effect of the regulatory framework is tested by implementing a dummy variable. Additionally, interaction terms are created by multiplying this dummy variable with the control variables. To eliminate the potential effect of the financial crisis of 2007-2008 on the implied volatility, an additional dummy variable is implemented. 𝜎5= 𝛽8+ : 𝛽( ; <=( 𝑃𝑟𝑖𝑐𝑒𝑛𝑑𝑒𝑥5D<+ : 𝛽) ; <=( 𝑇𝑟𝑎𝑑𝑒𝑣𝑜𝑙𝑢𝑚𝑒5D<+ 𝛽I𝑆𝑝𝑟𝑒𝑎𝑑(5+ 𝛽L𝑆𝑝𝑟𝑒𝑎𝑑)5 + 𝛽N𝑀𝑖𝐹𝐼𝐷𝐼 + 𝛽S𝐶𝑟𝑖𝑠𝑖𝑠 + : 𝛽V ; <=( 𝑀𝑖𝐹𝐼𝐷𝐼 ∙ log 𝑃𝑟𝑖𝑐𝑒𝑖𝑛𝑑𝑒𝑥5D<+ : 𝛽X ; <=( 𝑀𝑖𝐹𝐼𝐷𝐼 ∙ 𝑇𝑟𝑎𝑑𝑒𝑣𝑜𝑙𝑢𝑚𝑒5D< Y + 𝛽Y𝑀𝑖𝐹𝐼𝐷𝐼 ∙ 𝑆𝑝𝑟𝑒𝑎𝑑(5+ 𝛽(8𝑀𝑖𝐹𝐼𝐷𝐼 ∙ 𝑆𝑝𝑟𝑒𝑎𝑑)5+ 𝜀5 𝑀𝑖𝐹𝐼𝐷 𝐼= 1, 𝑓𝑜𝑟 𝑡 = 𝐴𝑓𝑡𝑒𝑟 𝑁𝑜𝑣𝑒𝑚𝑏𝑒𝑟 1, 2007 𝑎𝑛𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒. 𝐶𝑟𝑖𝑠𝑖𝑠= 1, 𝑓𝑜𝑟 𝑡 = 𝐽𝑢𝑛𝑒 1, 2007 − 𝐷𝑒𝑐𝑒𝑚𝑏𝑒𝑟 31, 2012 𝑎𝑛𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒.
Regarding MiFID II / MiFIR, the methodology provided in Section 5.4.2. is used. The potential effect of the regulatory framework is tested by implementing a dummy variable. Additionally, interaction terms are created by multiplying this dummy variable with the control variables.
𝜎5= 𝛽8+ : 𝛽( ; <=( 𝑃𝑟𝑖𝑐𝑒𝑛𝑑𝑒𝑥5D<+ : 𝛽) ; <=( 𝑇𝑟𝑎𝑑𝑒𝑣𝑜𝑙𝑢𝑚𝑒5D<+ 𝛽I𝑆𝑝𝑟𝑒𝑎𝑑(5+ 𝛽L𝑆𝑝𝑟𝑒𝑎𝑑)5+ 𝛽S𝑀𝑖𝐹𝐼𝐷𝐼𝐼 + : 𝛽V ; <=( 𝑀𝑖𝐹𝐼𝐷𝐼𝐼 ∙ log 𝑃𝑟𝑖𝑐𝑒𝑖𝑛𝑑𝑒𝑥5D<+ : 𝛽X ; <=( 𝑀𝑖𝐹𝐼𝐷𝐼𝐼 ∙ 𝑇𝑟𝑎𝑑𝑒𝑣𝑜𝑙𝑢𝑚𝑒5D< + 𝛽Y𝑀𝑖𝐹𝐼𝐷𝐼𝐼 ∙ 𝑆𝑝𝑟𝑒𝑎𝑑(5 + 𝛽(8𝑀𝑖𝐹𝐼𝐷𝐼𝐼 ∙ 𝑆𝑝𝑟𝑒𝑎𝑑)5+ 𝜀5 𝑀𝑖𝐹𝐼𝐷𝐼𝐼= 1, 𝑓𝑜𝑟 𝑡 = 𝑎𝑓𝑡𝑒𝑟 𝑗𝑎𝑛𝑢𝑎𝑟𝑦 3, 2018 𝑎𝑛𝑑 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 .
5. Results
The results of the stationarity test are provided first. Then, the results of the MiFID I and MiFID II / MiFIR models are discussed.
5.1 Testing for Stationarity of Volatility Measures and Other Variables
Before estimating the models, all variables are tested for stationarity during the MiFID I and MiFID II / MiFID sample periods. It can be concluded that the variables during the MiFID I sample window are stationary (Table 5).
Table 5: Augmented Dickey Fuller test results for the MiFID I sample window
CAC 40 DAX 30 FTSE 100
DF statistic
(p-value) DF statistic (p-value) DF statistic (p-value)
Volatility index -4.541*** (<0.001) -4.520** (0.024) -3.179 ** (0.021) Realised volatility -6.217*** (<0.001) -5.929** (0.016) -6.378** (0.014) Log (Price index) -5.236*** (<0.001) -6.931** (0.015) -6.308 *** (0.003)
Spread (bid-ask) -45.930*** (<0.001) -44.532*** (<0.001) -48.712*** (<0.001) Spread (high-low) -12.585*** (<0.001) -22.908*** (<0.001) -22.917*** (<0.001)
Volume -13.324*** (<0.001) -12.207 *** (<0.001) -12.535*** (<0.001)
The augmented Dickey Fuller test, tests the null hypothesis that a unit root is present in the sample window or, equivalently, that the sample window follows a random walk (Dickey and Fuller, 1979). The value between brackets is equal to the p-value. ***, **, * denote statistical significance at a 1%, 5% and 10% level respectively. Based on the results we can reject the null hypotheses of a unit root at a 5% significance level for all variables, meaning that all variables are stationary.
Furthermore, it can be concluded that the same variables during the MiFID II / MiFIR sample window are stationary too (Table 6).
Table 6: Augmented Dickey Fuller test results for the MiFID II / MiFIR sample window
CAC 40 DAX 30 FTSE 100
DF statistic
(p-value) DF statistic (p-value) DF statistic (p-value)
Volatility index -3.198** (0.020) -4.571*** (0.004) -3.541*** (0.007) Realised volatility -5.129*** (0.004) -5.663*** (0.002) -5.312** (0.021) Log (Price index) -4.238** (0.012) -3.345** (0.013) -3.308 ** (0.016)
Spread (bid-ask) -16.543*** (<0.001) -16.732*** (<0.001) -48.712*** (<0.001) Spread (high-low) -12.932*** (<0.001) -12.569*** (<0.001) -11.959*** (<0.001)
Volume -6.363*** (<0.001) -7.875 *** (<0.001) -9.279*** (<0.001)
5.2 Results Implied Volatility MiFID I Model
The estimated models are tested for multicollinearity. The highly correlated price index and volume interaction terms are omitted for estimation for all three models (CAC 40; Table 11, DAX 30; Table 12 and FTSE 100; Table 13). The results of the Schwarz information criterion are provided in Table 14. In the CAC 40 model, the present value of the price index and trading volume are used for estimation. For both the DAX 30 and FTSE 100 model, the lagged value of the price index and trading volume are used for estimation. The final estimated models are provided in Table 7 on the next page.
Table 7: Regression estimation results for the implied volatility MiFID I model
CAC 40 DAX 30 FTSE 100
(t-stat) Beta VIF (t-stat) Beta VIF (t-stat) Beta VIF
Cons 27.197*** 26.745*** 37.875*** (32.77) (36.54) (37.66) Price index -23.358*** 1.59 -17.783*** 2.18 -29.148*** 1.60 (-41.96) (-27.21) (-34.28) Spread(bid-ask) 0.035 1.03 0.002 1.02 0.003* 1.02 (1.52) (1.11) (1.75) Spread(high-low) 0.101*** 1.83 0.074*** 1.79 0.090*** 1.25 (27.08) (30.08) (41.30)
Volume 2.31e-08*** 1.86 2.03e-08*** 1.72 3.24e-09*** 1.11
(8.53) (8.77) (7.42) MIFIDI -2.510*** 1.31 -4.344*** 2.72 -2.676*** 1.58 (-8.61) (-9.29) (-8.94) Crisis 4.577*** 1.83 6.298*** 1.40 5.510*** 1.55 (15.65) (23.30) (22.98) MiFID*Spread1 0.001** 1.18 0.005** 1.29 0.003*** 1.76 (3.41) (2.25) (4.21) MiFID*Spread2 0.063*** 4.45 0.056*** 6.21 0.090*** 1.26 (5.78) (9.54) (41.40) R-squared 0.6747 0.6297 0.7209 Number of obs 2328 2332 2324
The value between brackets is equal to the t-statistic (test for significance). ***, **, * denote statistical significance at the 1%, 5% and 10% level respectively. A variable with a significance level of 5% (or lower) is considered to be significant. Each estimated coefficient, apart from the price index’s Beta, can be interpreted the following way: a change with 1 unit in the explanatory variable, leads to a change of 𝛽 units in the volatility index. The Beta of the price index can be interpreted the following way: an 1% change in the price-index, leads to a unit change in the volatility index of 100/𝛽 units. Regarding the Multicollinearity, a VIF value lower than 10 is said to be acceptable (Hair et al. 1995). Note that the VIF test only applies to the explanatory variables, hence there is no VIF value given for the constants.
Regarding the CAC 40 model, the coefficient of determination20 is equal to 0.6747, meaning that 67.47% in the variance of the volatility index is explained. For the DAX 30 model, the coefficient of determination is equal to: 0.6297, meaning that 62.97% in the variance of the volatility index is explained. The FTSE 100 model has a coefficient of determination equal to 0.7209, meaning that 72.09% in the variance of the volatility index
20 The Coefficient of determination (R-squared) shows how much variance of the dependent variable is
explained by the explanatory variables. The adjusted R-squared is used to check whether explained variance of the model improves after adding an additional explanatory variable to the model (Jin et al., 2006; Hutton et al., 2009).
is explained. Because the results of the estimated coefficients of each market index show similar effects, the results are jointly discussed.
Firstly, the estimated price indexes have a significant negative impact on the volatility index, which means that a higher price index results in a lower implied volatility. This could be explained by the assumption when the price index increases, the confidence in the market increases. Resulting in a lower implied volatility. Secondly, all the high-low spread coefficients have a significant positive impact on the volatility index. Implying that when the high-low spread increases, the implied volatility increases. This could be explained by the assumption that, when the intraday volatility increases (i.e. high-low spread), the uncertainty about the stock price increases, which translates to a higher implied volatility. The bid-ask spread coefficients however, have no significant positive impact on the volatility index. The estimated volume coefficients have a significant positive impact on the volatility index. Meaning, that a higher trading volume on daily basis, increases the implied volatility. This could be explained by the assumption that, when the trading activity per day increases, the probability for price fluctuations increases, translating into in a higher implied volatility.
All Crisis dummy variables have a significant positive impact on the volatility index. Regarding the CAC 40 model, the implied volatility during the crisis period is 4.58% higher; for the DAX 30 model: 6.30% higher; for the FTSE 100 model: 5.51% higher. All bid-ask spread and high-low spread interaction terms have a significant positive effect on the implied volatility. This implies that after the implementation of MiFID I, the effects of both the bid-ask spread and high-low spread on the volatility index significantly increased for all three indexes. The structural relation between MiFID I and the both spreads is therefore confirmed, which is in line with the second hypothesis.
Similar to the expectation, the estimated MiFID I dummy variables have a significant negative impact on the implied volatility. Regarding the CAC 40 model, this means that the implied volatility after the implementation of MiFID I has decreased with 2.51%; for the DAX 30 model a 4.34% decrease is observed; for the FTSE 100 model a 2.68% decrease is observed. Furthermore, it can be concluded that the variables jointly have a significant impact on the implied volatility for all three models (Table 15). In accordance with the first hypothesis, the implied volatility of all three market indexes is structurally lower after the implementation of MIFID I. These results correspond to the findings of earlier research on similar disclosure laws (Boehmer et al., 2005; Eom et al., 2007; Ferrell, 2007; Lee et al., 2015; Zhao and Chung, 2007).
5.3 Results Realised Volatility MiFID I Model
The highly correlated price index and volume interaction terms are omitted for estimation for all three models (CAC 40; Table 16, DAX 30; Table 17 and FTSE 100; Table 18). The present values of the price index and trading volume are used for estimation (Table 19). The final estimated models are given below.
Table 8: Regression estimation results for the realised volatility MiFID I model
CAC 40 DAX 30 FTSE 100
(t-stat) Beta VIF (t-stat) Beta VIF (t-stat) Beta VIF
Cons 16.379*** 18.124*** 37.875*** (68.86) (20.50) (37.66) Priceindex 0.801*** 2.57 0.717*** 2.33 -0.703*** 2.74 (31.28) (17.29) (27.12) Spread(bid-ask) -0.050* 1.16 0.003* 1.23 -0.002 1.14 (-3.61) (1.78) (-0.85) Spread(high-low) 0.056*** 2.44 0.0394*** 3.00 0.045*** 2.21 (9.64) (5.46) (10.88)
Volume -7.29e-09** 1.94 -6.66e-08*** 1.81 -6.28e-09*** 1.67
(2.93) (6.05) (-3.58) MIFIDI -4.1477*** 5.03 -4.344*** 5.02 -2.472*** 4.17 (-25.81) (-10.48) (-6.24) Crisis 3.213*** 1.71 6.446*** 1.59 8.879*** 1.88 (11.17) (-9.94) (22.19) MiFID*Spread1 0.001** 1.17 -0.011** 1.27 -0.014** 1.14 (2.01) (-1.17) (-1.83) MiFID*Spread2 0.011*** 5.01 0.037*** 5.78 0.004*** 5.15 (27.08) (2.77) (3.55) R-squared 0.5341 0.5460 0.4974 Number of obs 371 368 364
Regarding the CAC 40 model, the coefficient of determination is equal to: 0.5341, meaning that 53.41% in the variance of the realised volatility is explained. For the DAX 30 model, the coefficient of determination is equal to: 0.5460, meaning that 54.60% in the variance of the realised volatility is explained. The FTSE 100 model has a coefficient of determination equal to 0.4974, meaning that 49.74% in the variance of the realised
volatility is explained. Compared to the implied volatility model, the coefficient of determination for all three indexes are slightly lower.
In contrast with the implied volatility model, the estimated price index coefficients show different results compared to the implied volatility model. On top of that, the estimated volume coefficients have a significant negative impact on the realised volatility. In accordance to the implied volatility model, the bid-ask spread coefficients have no significant impact on the realised volatility. The high-low spread coefficients of all three models have a significant negative impact on the realised volatility as well.
Secondly, the Crisis dummy variables for all three indexes have a significant positive impact on the realised volatility. Regarding the CAC 40 model, the realised volatility during the crisis period is 3.21% higher; for the DAX 30 model: 6.45% higher; for the FTSE 100 model: 8.88% higher. All high-low spread interaction terms have a significant positive effect on the realised volatility too. This implies that after the implementation of MiFID I, the effects of the high-low spread on the realised volatility significantly increased for all three indexes. The results of the bid-ask spread interaction terms, however, differ across the indexes. For the CAC 40 model, the bid-ask spread interaction term has a negative effect on the volatility index, while for the DAX 30 and FTSE 100 models the effect is positive. Nevertheless, the structural relation between MiFID I and the both spreads is confirmed, which is in line with the second hypothesis.
Similar to the expectation, the estimated MiFID I dummy variables have a significant negative impact on the realised volatility for all market indexes. Regarding the CAC 40 model, the realised volatility after the implementation of MiFID I has decreased with 4.15%; for the DAX 30 model: 4.34% decreased; for the FTSE 100 model: 2.47% decreased. It can be concluded that after the implementation of MIFID I, the realised volatility is structurally lower for all three market indexes. Furthermore, the variables jointly have a significant impact on the realised volatility for all three models (Table 20). These findings correspond to the implied volatility models and are in line with the first hypothesis. Subsequently, the consistent outcome of the implied volatility models is confirmed.
5.4 Results Implied Volatility MiFID II / MiFIR Model
First, the highly correlated price index, high-low spread and volume interaction terms are omitted for estimation in the CAC 40 (Table 21) model and DAX 30 (Table 22) model. In the FTSE 100 model, all interaction terms are omitted because of high correlation (Table 23). The results of the Schwarz information criterion are provided in Table 24. For the CAC 40 model, the lagged value of the price index and trading volume are used. Regarding the DAX 30 and FTSE 100 model, the lagged value of the price index and the current value of the trading volume are used. The final estimated models are given below.
Table 9: Regression estimation results for the implied volatility MiFID II / MiFIR model
CAC 40 DAX 30 FTSE 100
(t-stat) Beta VIF (t-stat) Beta VIF (t-stat) Beta VIF
Cons 54.553*** 48.926*** 65.565*** (6.73) (24.14) (14.10) Log(Priceindex) -40.033*** 1.43 -33.024*** 1.06 -53.143*** 1.36 (-34.12) (-24.08) (-23.62) Spread(bid-ask) 0.013*** 1.00 0.004*** 1.05 -0.004*** 1.36 (2.62) (2.67) (-9.53) Spread(high-low) 0.026*** 1.42 0.014*** 1.40 0.015*** 1.25 (3.22) (5.92) (3.92)
Volume 2.17e-08*** 1.38 3.55e-09 1.24 -1.10e-10 1.10
(3.32) (0.65) (-0.33) MiFID II 2.262*** 1.32 2.982*** 1.26 1.545*** 1.12 (7.04) (12.08) (8.73) MiFID*Spread1 0.021*** (6.21) 1.08 0.002*** (5.44) 1.90 Omitted R-squared 0.6054 0.5673 0.6543 Number of obs 2328 2332 2324
Regarding the CAC 40 model, the coefficient of determination is equal to: 0.6054, meaning that 60.54% in the variance of the volatility index is explained. For the DAX 30 model, the coefficient of determination is equal to: 0.5673, meaning that 56.73% in the variance of the volatility index is explained. The FTSE 100 model has a coefficient of determination equal to 0.6543, meaning that 65.43% in the variance of the volatility index is explained. Firstly, the estimated price index coefficients have a significant negative impact on the volatility index. The estimated bid-ask spreads have a significant positive effect on the
