Msc Business Economics: Finance
Master’s Thesis
Insider trading after the Market Abuse
Directive: Room for Improvement?
Candidate:
Giovanna Petti
10825320
July 7, 2015
Supervisor:
Patrick Tuijp
Statement of originality
This document is written by Student Giovanna Petti who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is 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.
Insider Trading after the Market Abuse
Directive: Room for Improvement?
Giovanna Petti July 2015
Abstract
The scope of this research is investigating the impact of the Market Abuse Directive (MAD I) and Market Abuse Regulation (MAD II) - both implemented on the EU territory by the European Commission respectively in April 2003 and June 2014 - on a measure for market abuse, or more precisely, insider dealing, in the period 2001-2015. Previous literature suggested a positive (decreasing) effect of the Market Abuse Directive onto the insider trading volume, and pointed to a difference in results according to cross-country variation: the variation stands in the different implementation date of the Directive per country, made possible by the nature of the regulation. Together with the country variation, a cross-industry variation is also examined by this research. Results of this analysis suggest that the MAD I significantly increases the amount of insider trading in most of the model specifications. However, there is no significant cross-country, or cross-industry, variation in the MAD I impact on the market. On the other hand, the MAD II is suggested to (non-significantly) increase the insider dealing volume, and does so in all the specifications proposed. This could be due to the fact that the introduction of the MAD II is still too recent to be properly assessed. In either case, the above findings call for further research, to be performed on a longer period after the introduction of the Regulation. Volume of abnormal returns is also examined across time, and it sustains the findings above.
Keywords: Insider trading; Market Abuse Directive; MAD; Market Abuse
Acknowledgments
I would like to thank everyone who accompanied me along the way of this project, and supported me throughout it all: my supervisor, for guiding me through the fog of my ideas towards concrete goals; fellow students and friends, for listening to my thoughts and sharing their own; and my family and loved ones, for bearing through my rantings about financial matters at unusual times during the day, and reminding me of my enthusiasm and passion for this discipline every time I lost sight of it.
Abbreviations
ARs: Abnormal Returns
CARs: Cumulative Abnormal Returns
CAARs: Cumulative Average Abnormal Returns
CSMAD: Criminal Sanction for Market Abuse Directive
EU: European Union
HFT: High Frequency Trading
MAD (I & II): Market Abuse Directive
MAR: Market Abuse Regulation
M&A: Merger & Acquisition
SEC: Security and Exchange Commission
TDP: Transparency Directive
Table of Contents
STATEMENT OF ORIGINALITY II ABSTRACT III ACKNOWLEDGMENTS IV ABBREVIATIONS V Section 1 Introduction 1Section 2 Literature Review 5
2.1. Insider trading on the economy 5
2.2. Measures of insider trading 8
2.3. On the Market Abuse Directive (MAD) and the Market Abuse Regulation
(MAR) 9
Section 3 Data And Descriptive Statistics 14
3.1 On data and sources 14
3.2 Statistics on corporate announcements 16
Section 4 Methodology 22
4.1 Research objectives 22
4.2 Event study approach 22
Section 5 Results 28
5.1. Cumulative Abnormal Returns per period 28
5.2. Incremental regression output 30
Section 6 Robustness Checks 36
6.1 Clustered and Newey West Standard Errors 36
Section 7 Conclusion 40
7.1. Discussion 40
7.2. Suggestions for further research 42
REFERENCES 45
APPENDIX A: Other Results 49
Section 1
Introduction
The Market Abuse Directive (MAD), implemented in 2003, was a measure taken by the European Parliament and the Council to establish sanctions for matters like insider trading, unlawful disclosure of information, and market manipulation in all the territories of the European Union. In fact, insider trading in Europe was an issue addressed much later than in the United States: the MAD is the first EU directive focusing on the issue, while the SEC’s rule 10b-5 from the Exchange Act dates back to 1934 (Ventoruzzo, 2014). The Market Abuse Directive condemns almost all forms of insider dealings (Report on the Proposal for a Directive of the European Parliament and of the Council on criminal sanctions for insider dealing and market manipulation), unlawful disclosure of insider information, and market manipulation (meaning transactions giving false or artificial signals to the market).
After the recent LIBOR scandal in 2012, the Commission proposed more stringent measures to fight off market abuse forms: amendments of the directive include focus on the manipulation of benchmarks, such as LIBOR and EURIBOR. Even more recent debates on the topic involve the need for a Market Abuse Regulation (MAR), rather than a Directive. The recent financial crisis and scandals (e.g. the previously mentioned LIBOR scandal) triggered an impact assessment of the Directive by a Steering Group of the
European Commission: in the text of the impact assessment (European Commission, 2011), the committee concluded that the MAD lacked some application features, such as lack of clarity in defining the target of the sanctions, and not enough powers given to regulators; hence it was not often put in practice by those. In fact, the increase in trading volume of market commodities and other financial instruments made it difficult for regulators to identify the violations of the directives (European Commission, 2011).
Thus, the proposal to create a Market Abuse Regulation was born to make up for the lacks of the MAD. In particular, the MAR would act against trading strategies which involve algorithms and automated HFT (Proposal for a Regulation of the European Parliament and of the Council on insider dealing and market manipulation (market abuse)).
Christensen et al. (2010) study the effects of the MAD and the Transparency Directive (TPD) on EU markets' cost of capital and liquidity; they make use of the cross-sectionality of their sample to understand if implementing the regulation later in time, or in a more strictly regulated country, has any effects whatsoever on the market's characteristics. They find that there are indeed some differences, across time and across entity: there is a decrease in cost of capital and liquidity after the entry into force of the two directives; in particular, liquidity decreases more in countries which had stronger prior regulations and regulators quality.
My research also makes use of the MAD, as it is conceived as a study on the impact of the Directive: its implementation gives me the opportunity to study the abnormal returns due to insider dealing in EU markets, before and after the enactment of the directive (April, 2003). Hence, my aim is to study the before and after effect of the Market Abuse Directive on measures of insider dealing for countries belonging to the EU: the cross-sectionality of
the analysis would allow for an interesting study on the effect of a diversified implementation of the law, across countries and across industry sectors.
The hereby proposed hypothesis is that the MAD does indeed reduce insider dealing across countries; however such change in insider trading is not homogeneous, but rather, expected to be different across EU countries and sectors: countries in which the regulation was implemented earlier are expected to be impacted more by the regulation, i.e. insider trading in these countries decreases more – due to the regulation - than in other ones. Results are also supposed to vary across industries: in fact, insider dealing in technology firms is expected to suffer a higher decrease due to the regulation than other industry sectors. Furthermore, the research would be interesting to EU regulators: it would shed further light on the impact of the MAD, and potentially clear their views on the implementation of the MAR.
On a bigger scale, exploring this topic could be informative to determine whether such legislative directions can be improved by policy makers: a possible tightening of the regulation may be considered by regulators, in the case in which the hereby examined measure of insider dealing would respond only slightly (or not at all) to the introduction of the laws.
Another aspect to be examined is what could be the consequent possible effect on insider gains, and subsequently, on firm value. Insider trading has in fact, shown to decrease overall economic growth, by raising the cost of capital for issuers (Bhattacharya and Daouk, 2002). Moreover, buying a stock as an insider, and then selling it at a higher price on the basis of private information, is: a form of financial arbitrage, that can result in hindering strong-form markets’ efficiency (Dothan, 2008), as there can be no
strong-form efficient market; a problem of fairness in allocations of benefits across investors (Krawiec, 2001); and a criminal violation (misappropriation), as it implies a breach of confidentiality of secret information within a corporation which gives economic advantages to the person who has knowledge of it (Uniform Trade Secret Act, 1986). Hence, the research could contribute to findings on different economic and legislative levels, by addressing: from a macro-economic point of view, policy makers; and from a micro-economic one, investors and corporations.
This paper is going to be organized as follows: Section 2 delves into previous literature on the concepts of insider trading and merger announcements laws; Section 3 exhibits summary statistics and information about data and data sources; Section 4 develops on the methodology and the model used in this analysis; Section 5 presents the results; Section 6 presents robustness checks and Section 7 maps out the main conclusions drawn from the empirical analysis, and any possible future research opening.
Section 2
Literature Review
2.1. Insider trading on the economy
Insider trading is a known economic phenomenon, which has been subject to numerous debates and scrutiny (Finnerty, 1976; Carlton and Fisher, 1983; Demsetz, 1986). The definition of insider trading by the North-American Security and Exchange Commission describes it as the act of buying or selling
a security and breaching a fiduciary duty in doing so, because of possession of
private and confidential information about said security, which may lead to a potential illicit gain. Now, while in many countries, the law still considers insiders only those who possess more than 10% of the company’s equity (Securities Exchange Act Of 1934), in the United States the definition of insider has broadened: it includes anyone who trades on private information for personal gain, or even tips said information to others, when this information has some potential effect on the security’s stock price. This phenomenon does not only have direct legal consequences towards the offenders, but it also affects the financial market by affecting share prices (Cornell and Sirri, 1992) and market liquidity (Christensen, 2010).
The rising amount of regulation on this activity has sparked a lot of dialogue and controversy during the last century: some economists (Manne, 1966; Easterbrook, 1981) go as far as saying that insider trading is a benefit, not a danger, for the economy. The reasoning behind their beliefs is that this
activity brings, although illicitly, more information to the market, thereby making the market more efficient. That is because said information is
material: meaning that it causes a change in the stock prices of the traded
securities. In particular, Manne (1966) argues that this practice brings about economic gains to both parties: those who long the stock, and those who sell it. It favors the former group, for they will make a higher profit out of the news; it also benefits the latter, for the price of the security they are selling will be higher than its market price: hence, they will also gain from this type of trading, just not as much as the holders. According to Easterbrook and Fisher (1996), the materiality requirement limits the applicability and scope of the rule 105b, which belongs to the Securities Exchange Act issued from the SEC in 1934, and dictates public disclosure of any material claim against a corporation.1
On the other hand, critics of insider trading have a different opinion on the issue: according to Levmore (1982) it is dangerous for managers’ incentives, for it allows managers (or any other employee trading on the private news) to benefit from negative corporate announcements as well; it also pushes management to delay disclosure of important corporate information, in order to better administrate these illegal gains (Schotland, 1967).
Amongst other studies on the topic, Meulbroek (1992) finds that insider trading is impounded into stock prices, resulting in a 3% value for abnormal returns: more specifically, trading volume and other trade characteristics (such as trade size, frequency and direction) signal the presence of insider trading to the market, which then responds by detecting the insider gains, and incorporating it into the stock price. Damodaran and Liu (1993)
also contribute on the research on insider gains, in the field of real estate investment trusts: they find that insiders do react on appraisal news, by buying in case of favorable appraisal, and selling in the opposite case. Such practices may, however, be of harm to the corporation, because they tend to affect corporate behavior by discouraging investments (Manove, 1989); Cheng and Lo (2006), moreover, confirm that insiders often go as far as exploiting their disclosure opportunities for their own personal profits.
The act of trading on insider information is still, as of now, shown to be significantly present in the corporate world, in spite of rising regulations on the matter: in a recent study, Augustin et al. (2014) find that about a quarter of all M&As transactions are affected by insider trading. Hence, modern regulation could not completely bring this phenomenon to a halt.
There is ample evidence on the effects of insider trading within corporations, and to the overall economy. On a macro-economic level, Bhattacharya and Daouk (2002) find that insider trading laws can reduce the cost of equity in a country by 5% and raise the cost of capital for issuers. Hence, it is essential to regulate and keep insider trading under control. Moreover, knowledge about insider trades can also bring to the table fundamental and predictive information about future stock returns to investors and corporations: Seyhun (1992) finds that knowledge of aggregate insider trades can help predict up to 60% of future aggregate stock returns, and therefore constitutes a potential driver of future great portfolio gains. This benefit may be better managed by investors, by understanding whether mergers’ announcement disclosure has any effect on insider trading, and using this information whilst building a portfolio.
Considering all the previously mentioned researches, insider trading appears as a double-edged sword that can affect the economy in more than
one direction: what is clear in the midst of the debate, though, is that it can bring much greater benefits to all parties, if well regulated.
2.2. Measures of insider trading
One influential paper on the matter of insider trading linked to corporate announcement is from Banerjee and Eckard (2001). The authors examine the magnitude of insider gains in the U.S. in the period going from 1897 to 1903, when insider gains were not so strongly monitored. The aim of their research is to gauge whether tighter regulations applied to insider gains had some effects in reducing the above mentioned gains, and if they promoted outsider participations.
The model used by the authors makes use of three measures for abnormal returns due to insider gains which are going to be used in this research as well:
- Runup, which stands for the average cumulative abnormal return preceding the first announcement report;
- Event Gain, which represents cumulative abnormal returns during the announcement event;
- and Postevent Drift, which measures any abnormal return following the announcement event.
Furthermore, they define a Premium, as the sum of Runup and Event Gain, and a Runup Index, as the percentage ratio of Runup over Premium. They find these cumulative abnormal returns during a three month window, by first building these measures for the full sample, and then separately, for prospective mergers, and fait accomplit mergers.
2.3. On the Market Abuse Directive (MAD) and
the Market Abuse Regulation (MAR)
In order to tackle the issues of financial markets’ lack of integrity and manipulation of financial information, the Market Abuse Directive was proposed by the European Commission, and then adopted by the European Parliament in April 2003. Its specific aim was to target and prevent any type of market abuse: meaning, the inappropriate use of financial and/or insider information, in stock trading and a number of other practices (such as spreading false information about a corporation, or the state of the economy).
The principle of sound prudential and conduct business regime (Proposal for a Directive Of The European Parliament And Of The Council on criminal sanctions for insider dealing and market manipulation, 2011) is protected by the Directive, through the use of effective sanctions applied by supervisory bodies and regimes. The proposal follows the line of action of an EU guideline which states that criminal law should be implemented into sanctioning regimes, to guarantee a sound and successful implementation of policies. This was born out of the mindset that the previously existing sanctions against market abuse were not enough: they were not functioning as they should, for they often did not leave any impression on Europe’s financial market as they were intended to do.
However, the Directive itself lacked many practical features - as it emerged from the impact assessment (Market Abuse Directive Impact
Assessment, European Commission 2011) - and was often criticized (Ferrarini,
2004; Siems and Nelemans, 2012): the text of the amendment to the proposal identifies two key areas of interests that should be reflected upon. They are
both forms of market abuse conduct: market manipulation and insider dealing. Of the two, any attempt at market manipulation can be punished; however, insider dealing can only be punished when those who own the inside information are well aware that that information is indeed confidential. Hence, both practices are considered criminal offenses, when said conditions are in place. This line of thought is much more drastic than past measure, and labels two often harmful practices with a legal status that cannot be ignored anymore.
The above mentioned considerations gave rise to a new Directive, and a completely new set of rules, named Market Abuse Regulation (MAR). The legal difference between a Directive and a Regulation is the mode of application: while a Directive cannot be applied without a conforming legislation drafted by national authorities, a Regulation does not require that type of adjustment, and can be applied directly.
The program containing the Directive and the MAR together was defined Market Abuse Directive (MAD) II. Just recently introduced (June 2nd, 2014), the MAR was in line with the ‘Better Regulation’ policy of the Commission (European Commission, 2011). According to the Regulation, the source of the Directive’s problems is the lack of powers of the regulators and the lack of sanctioning flavor of the previous rules. Most of these issues were caused by non-sufficient information and transparency on the concepts, definition and circumstances traced by the MAD. According to the European Commission’s text for the Market Abuse Regulation (2011)2 the ways in which the MAR makes up for the flaws of the previous directive are
2 European Commission (2011). Proposal for a Regulation Of The European
Parliament And Of The Council on insider dealing and market manipulation (market abuse). Brussels, 20.10.2011 COM(2011) 651 final 2011/0295 (COD)
numerous: first of all, it acts on the clarity issues by improving the set of information and definitions given to regulators and investors; secondly, it gives additional powers to authorities (for example, the power to investigate such breaches of confidence within a corporation); thirdly and lastly, it filters through the set of options previously given to authorities, for a more direct and efficient regime which will hopefully bring to an increased level of financial stability. Most EU countries have already adopted and implemented such Regulation: it is interesting to see how they absorbed the changes - if they did -, and if they all did it with the same intensity. This is one of the scopes of this research, that also wishes to confront and compare the impact of the MAD I with its amendments, and understand whether the hypothesized assessment from the Commission was correct, or whether there is still room for improvement, even for the newly introduced Market Abuse Regulation.
2.4. Effects of similar regulations on insider
trading
Several studies and impact assessments are continuously carried out in order to investigate the outcome of rules and directives on the economy. In particular, the issue of insider dealing and market abuse is closely followed by authorities and researchers, since such practices are still in place, despite the tightening constraints, and since the financial worldwide market is constantly evolving.
However, the focus on the Market Abuse Directive has been quite limited since its implementation: an important document on its effect on the economy has been published by Christensen et al. (2010), who refer to this
particular piece of regulation and examine its outcomes and spillovers in the capital markets. Their approach focuses on differences in the timing of the adoption of the MAD from various EU countries and therefore adopts a cross-sectional structure, giving me the instruments to perform a similar type of analysis on primary effects of the regulation; on the other hand, it says little about insider dealings, as its scope is restricted to indirect effects of the regulation, as the ones on cost of capital and market liquidity.
Hence, the need for a research focusing on the direct effects on insider dealing and other misappropriations related to the enactment of the Directive – and its amendments – is apparent. Prevoo & Weel (2010) were the first to study the direct effects of the directive, applying an event study methodology and calculating cumulative abnormal returns before and after the Market Abuse Directive’s enactment. Their results are informative: the Market Abuse Directive has been found to reduce insider dealing measured by cumulative abnormal returns indeed, but said reduction is diverse across industries and countries. This interesting result calls for a more recent study, to assess the real impact of the MAD and its amendments over time, especially in light of the recent changes in the regulatory framework which have brought the European Commission to draft and accept a revised Market Abuse Directive, under the name of Market Abuse Regulation.
The analysis from Prevoo and Weel (2010) is a fundamental piece of literature on the topic: its results are hereby taken into consideration as a research benchmarks to compare with these paper’s findings, while examining and assessing the effect of the regulations on insider dealings, while also adopting Christensen et al.’ (2010) cross-sectional approach. The way this research differs from Prevoo and Weel (2010) analysis is in its incremental feature: it tries to identify the impact of the first draft of regulation over time
(MAD I) while also accounting for any possible changes caused by the second version of the Directive (MAD II), all the above by keeping track of the cross-country and cross-industry possible differences of the implementation. This proves for a comprehensive analysis, which moves into three different directions: over time, over country and over industry.
Section 3
Data and descriptive statistics
3.1 On data and sources
A panel dataset with data about corporate announcements in 27 EU countries and the respective companies’ stock market returns is obtained for the period going from January 1st, 2001 through April 1st, 2015. Announcements refer to M&As (the examined companies can have the role of acquirer or target), tender offers or issue of shares, which occurred in different dates throughout the aforementioned sample period, and are related to companies listed on different European stock exchanges. Data about those corporate news and the related announcements (containing the deal announcement date, effective date, deal volume, deal status, synopsis, etc.) are retrieved via Thomson One, restricting the available observations to the sample period, and to the EU countries and stock exchanges indicated in Table 2b and 3b of Appendix B, respectively.
Announcement related firms’ closing prices are then collected via Datastream, on a daily frequency: together with the market index returns, they are used to calculate the abnormal returns around the corporate news. Those abnormal returns work as a proxy for market abuse (Runup), as they represent the deviation of a firm’s return from the market index returns at the closing time, due to the corporate announcement (Banjeree and Eckard, 2001).
The datasets are cleaned from duplicates and irrelevant observations (a variable for day of the week is created, and observations for Saturday and Sunday are dropped), and then merged together via an unique identifier variable for the companies, which is created according to the date and the company name. Furthermore, data about daily stock market index returns is also retrieved through Datastream and then added to the main dataset, joined via a variable which represents both returns’ closing date: the chosen index is EURONEXT 150, since most of the retrieved companies have been listed on the Euronext Amsterdam, Brussels, Paris and Lisbon Exchanges3.
Finally, most control variables are obtained through Datastream and added to the dataset: firm size indicators (Share Turnover and Market Value), capital structure indicators (Financial Leverage) and firm status indicators (Return Variability). Ultimately, an additional control variable for (country)
Inflation, as a proxy for the state of the economy, is extrapolated via
Worldscope and the International Monetary Fund.4 Control variables are merged to the main dataset on a daily basis, resulting in a final dataset of 18,838 observations of corporate announcements - publicized by 10,628 corporations - and 184 variables.
The year 2001 is chosen as a starting date for the dataset because it represents the latest year prior to the implementation of MAD I for which there is available information (as the data about the market index returns is only available after the introduction of the Euro); the ending date, 2015, is the latest available period, for it is important to gauge the most recent
3 For a complete list of the Market Indexes from which corporate returns have
been retrieved, refer to Table 3b in Appendix B.
4 The complete list of all data sources is displayed in the Appendix B, Table 4B.
reaction to the regulation, immediately prior to the effective application of the MAR.
3.2 Statistics on corporate announcements
The dataset is then divided in periods, by creating two dummy variables for the enactment of the MAD I, and the MAD II. MAD I takes value of 1 in case the announcement date is subsequent to April 1st, 2003 (and 0 otherwise); in a similar fashion, MAD II takes the value of 1 in the case that the announcement date follows June 2nd, 2014. A variable for the overall period which takes different values according to the type of legislation in force at the time of announcement is then created: this exercise allows to partition the dataset into three sections: a starting period prior to MAD I; an intermediate period following MAD I and prior to MAD II; and a final period following MAD II. Table 1 displays the distribution of announcements over the three periods. Looking at the yearly or monthly average distribution figures, it is possible to understand the frequency of announcements per period: these values are not too different from period to period: this allows to compare the periods’ outcomes in abnormal returns, even though the lengths of the Before
MAD I period and the After MAD II ‘s one are significantly smaller than the
TABLE 1: DISTRIBUTION OF ANNOUNCEMENTS PER PERIOD – ACCORDING TO REGULATION ONLY
Period Before MAD I In between MAD I and
MAD II After MAD II Distribution Per month (av.) Per year (av.) Total (N) Per month (av.) Per year (av.) Total (N) Per month (av.) Per year (av.) Total (N) Total Frequency (N) 117 1408 2815 124 1,488 15,625 76 758 758 18,838 Percentage (%) 14.94 81.03 4.02 100.00 Cumulative (N) 14.94 95.98 100.00 - Table 1 - continued
This table looks at the distribution of corporate announcements for the firms in the sample for the three aforementioned periods: a) before MAD I; b) in between MAD I and MAD II; c) after MAD II. Together with the total number of announcements per period, the average number of announcements per year and per month (in the specific period) are also displayed.
As mentioned in Section 4 of this paper, it is also important, in order to test the hypotheses of the research, to have an understanding of the announcements according to company’s industry and geographical region. The reason for this choice is that a previous research by Prevoo and Weel (2010) proves that there is indeed a significant difference in the reaction to the regulation according to industry – they find that technology firms’ abnormal returns before the MAD are higher than those calculated after the regulation,
for bad news announcements. Table 2 (to be found in Appendix A) shows the distributions of observation per macro industry, subdivided by the different regulation periods. Furthermore, Figure 1 shows a plot of such distribution over industry type. Finally, Table 3 instead displays the distribution of observations per country (also found in Appendix A).
3.3 Statistics on Abnormal Returns (ARs) and
Cumulative Abnormal Returns (CARs)
Abnormal returns are then calculated by regressing daily firms’ stock prices onto the EURONEXT 150 index daily values, and obtaining the residuals. Detailed summary statistics about those abnormal returns are shown below: the mean of this variable is negative all of the three examined period; however, the magnitude changes when we move forward in time (see Figure 1). In particular, Figure 1 displays the mean of the abnormal returns taken per day, over the chosen sample of firms: a low level of abnormal returns (compared to the sample average) precedes the introduction of the MAD I; the latter is followed by an irregular pattern of returns.
TABLE 4: ABNORMAL RETURNS’ MEAN AND
SIGNIFICANCE
By Period
Before MAD I In between MAD I and MAD II
After MAD II
Absolute value (N) -0.8049*** (0.0570) -1.2065*** (0.0273) -0.9113*** (0.1081) T-value -14.1100 -44.1836 -8.6188 Table 4 - continued
This table looks at the mean and t-value for the Abnormal Returns calculated around corporate announcements for the three aforementioned periods: a) before MAD I; b) in between MAD I and MAD II; c) after MAD II. Standard errors are displayed in parentheses:
all values for the three periods are significant at the 5% statistical level. The following syntax is also used: *significant at 10% statistical level; **significant at 5% statistical level;
***significant at 1% statistical level.
FIGURE 1: ABSOLUTE ABNORMAL RETURNS’ LINE PLOT OVER TIME
Figure 1 - continued
This graph shows the plot of abnormal returns’ absolute values over the time, covering the whole sample period. The y–axis variable Abnormal Returns (Mean) is represented by the abnormal returns’ mean calculated per day, the x-axis variable Date stands for the date in which the abnormal returns were registered. Two dates are indicated among brackets: they represent the introduction of the MAD I (1/4/2003) and the MAD II (2/6/2014). This allows to partition the graph in three different sections, and compare the abnormal returns: a) prior to the introduction of any regulation; b) after the enactment of the MAD I; c) after the enactment of the MAD II.
Cumulative abnormal returns are then calculated by aligning the main date variable around the (corporate) announcement date, and using the aforementioned event window of 120 trading days: the main statistics point out a positive mean of 12.9614 (see Table 5). Furthermore, this value is not worrisome: it would suggest that on average (in this particular sample) a firm cumulates abnormal returns due to a corporate announcement equal to a value of 12.9614 – which, considering a Returns’ mean of 258.3816, translates to an average abnormal return of 5.01% - over a period of 120 trading days.
TABLE 5: DEPENDENT AND INDEPENDENT VARIABLES’ SUMMARY STATISTICS
Overall sample
Mean Standard
deviation Variance Skeweness Kurtosis Min/max CARs 12.9614 71.9933 5183.033 2.0836 8.4557 -105.5614/
261.3904 ARs -1.1269 29.7343 884.1277 -0.2091 5.8776 -82.4317/
Market Value 35895.27 138792.7 1.93e+10 11.4551 166.1689 2.04/ 3279869 Turnover (by value) 157882.6 3422684 1.17e+13 644.45 462107.3 0/ 2.59e+09 Debt (as % of common equity) 202.653 557.5091 310816.4 -1.587536 228.1083 -15967.21/ 10080 (Country) Inflation (CPI, annual %) 2.1762 1.9844 3.937679 7.1895 103.638 -4.479938/ 34.4678 Return variability 71.8230 345.9999 119715.9 10.8731 139.7271 .0144009/ 4760.204 Table 5 - continued
This table looks at the main summary statistics for the Cumulative Abnormal Returns calculated around corporate announcements for the whole sample, Abnormal Returns and all the used control variables. Variables displayed by value are: Market Value, Turnover, and
Return Variability; variables displayed as percentage of other variables are: Debt and Inflation. The main summary statistics are displayed, together with additional measures like Skeweness and Kurtosis. The following syntax is also used: *significant at 10% statistical level; **significant at 5% statistical level; ***significant at 1% statistical level.
Section 4
Methodology
4.1 Research objectives
The research’s objective consists in determining the behavior of insider dealing around the enactment of the MAD, that serves as the partitioning event around which any changes are examined: understanding if insider dealing volume went through any dramatic shifts after the Directive was effective in the territories of the EU and whether the MAD paved the way for the MAR.
These objectives are investigated via a methodological approach consisting of two phases: i) an event study of the insider dealing around the enactment of the regulation, and ii) a regression model, in which a measure for insider trading is regressed upon a dummy variable for the MAD (I and II).
4.2 Event study approach
I perform an event study5 based on a methodology very similar to Banjeree and Eckard’s (2001) and Prevoo and Weel (2010); in this case, each corporate news announcement is the event before and after which abnormal returns are investigated (t=0). The time window around the announcement is of 4
5 More information on the event study methodology from Binder’s The Event Study Methodology since 1969 (1997).
months (120 trading days), to account for insider trading around the news and any adjustments to the regulation. This choice, proposed by MacKinlay (1997), is justified by the use of daily data and of the market model, as it allows for the news to be eventually incorporated into stock prices. Moreover, the same time window was also used by Prevoo (2010) and Wong (2002), according to the same line of thought. The abnormal returns for the stocks are calculated by using the standard market model (following the same approach used by Banjeree and Eckard (2001)):
ARit=Rit – α+β Rmt , (1)
and then cumulated over the period, in order to get the cumulative abnormal returns, which represent overall insider gains:
CARi= ARi,t+k !!"
!!!!"
.
After having determined the cumulative abnormal returns, those are averaged for the whole sample, resulting in the cumulative average abnormal returns:
CAAR=N1 CARi N
i=1
.
This exercise is performed before the introduction of the MAD and repeated for the period after. Then, both samples’ summary statistics and line plot are going to be examined, to understand which direction insider (2)
trading volume took around the aforementioned event. After that, the data is going to be displayed and examined in new samples organized by industry, country, and announcement type. Tables containing information about the entry into force of the MAD in the all examined EU countries are going to be displayed in Appendix B. This is useful to prove or deny evidence in favor of the first hypothesis:
H1: Insider dealing volume is expected to decrease after the introduction of the Market Abuse Directive; however, such decrease is not expected to be homogeneous, but quite different across industries and countries, due to the difference in enactment dates and because of the apparent faults in the text of the regulation examined by previous literature. A further decrease in volume of insider dealing is expected after the introduction of the Market Abuse Regulation.
4.3 Incremental regression approach
The next step consists in ascertaining the impact of the regulation’s introduction into the legal and corporate system, and their effect on market abuse - in general, and insider trading - in particular. This would allow to double check the results of Section 4.2, and moreover, assess the significance of the regulations’ impact on the abuse.
Firms’ abnormal returns, representing a measurement for insider trading, are hence used as the dependent variable of an incremental regression model: they are regressed onto several dummy variables for the regulations enactments at different points in time (which indicated whether the
corresponding regulation was in place – with value of 1 – or not – with value of 0 - for that specific time), some additional control variables, and some fixed effects for country and industry. The regression model would then take the following form (extended variable labels and descriptions to be found in Appendix B):
ARit=β1MADIit+β2MADIIit+γControlsit+δFixedEffectsit+εit
The dependent variable of the model above is represented by the insider dealing, measured by the abnormal returns that were used to test the previous hypothesis. The main independent variables, displayed on the right hand side of the equation, are two dummy variables (indicated by the coefficients going β1 and β2) representing the enforcement of the regulations,
i.e. they take the value of 1 if the Directive was already enforced at the time of the deal announcement, and 0 otherwise. In particular, MADI stands for MAD I, the first draft of the Market Abuse Directive implemented in April, 2003; MADII instead, stands for MAD II, the second draft of the directive, which comprehends the Market Abuse Regulation (MAR) and the Criminal Sanction for Market Abuse Directive (CSMAD). Finally, country and industry fixed effects and some control variables are also added to the model. The control variables chosen for the model are Market Value, Share Turnover,
Return Variability, Inflation and Financial Leverage. The chosen variables
represent factors affecting the amount of trading done inside the company and
with the company (i.e. share repurchases and purchase or sale of other
companies’ shares linked to the examined corporate announcements – M&As (4)
and tender offers), as well as the whole trading volume at the time of the announcement.6 The hypotheses to be tested by this regression model are the following:
H2: Abnormal returns representing insider trading are expected to be significantly related to the independent variables MADI and MADII: moreover, the former are expected to be negatively related to the latter, with a stronger expected significance of the MADII, as it is expected to amend to some of the faults of the first version of the MAD.
Moreover, differences across countries and industries are going to be examined through the summary statistics on the entry into force of the regulation, and the CARs relative to different EU countries: in case of a huge variance in absolute value and statistical significance of results arising from the cross sectional nature of the model, a corresponding legislative explanation is probably behind such phenomenon. Christensen et al. (2010) already examined differences across industries and countries due to later implementation dates, with regards to market liquidity and cost of capital, rather than insider trading. They find evidence of a substantial variance in outcomes, due to differences in implementation of the regulation, and of hysteresis, meaning that countries having weaker regulatory systems find it difficult to quickly catch up with strongly regulated countries (capital market effects from new rules are broader than those effects due to past regulations). This moves me to formulate the following hypothesis:
H3: It is expected that the magnitude of CARs in countries in which the regulation was implemented at an earlier date is lower than the magnitude of
6 More information about the control variables to be found in table 1B of
CARs corresponding to a country in which the MAD I and II were implemented at a lower date.
Section 5
Results
5.1. Cumulative Abnormal Returns per period
The next step of the analysis after calculating the abnormal returns per stock, and then cumulating them over a period of 120 trading days, in order to obtain the CAR per stock, implies averaging them across the sample, to obtain the CAAR for the whole timeframe; this exercise is going to be repeated three times: once for the period before the introduction of the MAD I, once for the period after the MAD I and before the MAD II, and finally, once for the period after the introduction of the MAD II.
In order to obtain the abnormal returns per stock, the closing prices of the company that is being analyzed are regressed against the market index EURONEXT 150 returns, overall the sample period: information about the slope and intercept of the regression is obtained in this fashion, for all the companies whose news announcements are part of this analysis.
ARit = Rit - α+β Rmt
The standard market model above displayed allows to calculate the abnormal returns that are then going to be cumulated over the specific period of 120 trading days, around the announcement event. CARs are then differentiated among the three aforementioned period, and then compared in
Table 5. Table 6 displays the values of CARs and CAARs across firms: for each period, the total value of each variable is given, together with a yearly average across firms.
Observing the behavior of CAARs, it is shown that those seem to drop very close to 0 after the Directive is introduced, and then waver again towards slightly higher values after the Regulation is introduced.7 This result hints at the following finding: the event study performed in this analysis suggests that the Market Abuse Directive did have some positive effects on the Cumulative Average Abnormal Returns (CAARs) present in this sample, by giving lead to a slight decrease of those abnormal returns after its introduction. The Market Abuse Regulation, on the other hand, seemed to increase the CAARs, hence being less successful in its aim than its antecedent measure. However, it proves interesting to confirm this finding by applying the incremental regression approach, displayed in Section 5.2.
7 The reason why CAARs are more informative than CARs in this case is that
the latter display the total value of abnormal returns cumulated over time, while on the other hand, the former present the average value of abnormal returns cumulated over time (i.e. the total number of abnormal returns divided by the number n of observations), thus being a better instrument for comparison across periods.
TABLE 6: EVENT STUDY OUTPUT: CUMULATIVE (AVERAGE) ABNORMAL RETURNS PER PERIOD
Period Before MAD I In between MAD I
and MAD II After MAD II Whole sample CARs & CAARs Per year (av.) Total (N) Per year (av.) Total (N) Per year (av.) Total (N) Per year (av.) Total CARs (N) 6.4809 12.9618 1.1785 12.9640 8.6379 12.9568 0.9258 12.9614 Count (N) 120,834 241,668 108,765 1,196,421 59,458 89,187 109,091 1,527,276
CAARs(N) 0.00005 0.00005 0.00001 0.00001 0.00014 0.00014 8.487e-06 8.487e-06
Table 6 - continued
This table looks at the output of the event study for the three aforementioned periods: a) before MAD I; b) in between MAD I and MAD II; c) after MAD II. Cumulative abnormal returns are displayed in absolute value and relative value for each period. The absolute value corresponds to the volume of the cumulative returns per period; the relative value is calculated by dividing the former by the count of CARs for that particular period. Hence, relative values (CAARs) are better measures for comparison between the periods.
5.2. Incremental regression output
After shaping the dataset as a panel data analysis, the regression of the abnormal returns on the two dummies for MAD I and MAD II is ready to be run. Said regression is presented as a panel data regression, run with and
without fixed effects, using the option robust on Stata, to make up for too big standard error values. The group variable is represented by the Company Unique Identifier, and the date variable is the Date in which said company’s closing prices are announced. Table 7 displays the regression’s outcome:
TABLE 7: REGRESSION OF ABNORMAL RETURNS (DUE TO INSIDER TRADING) ONTO MAD I AND MAD II DUMMY VARIABLES
Dependent variable: ARs
(1) (2) (3) (4) (5) Incremental regression with constant Incremental regression with controls Incremental regression with country fixed effects Incremental regression with industry fixed effects Incremental regression with country and industry fixed effects MAD I 8.1543*** (0.5273) -0.7166*** (0.3383) 7.4316*** (0.8222) 7.8901 (8.8432) 7.4286 (14.8061) MAD II 4.0533*** (0.8202) 12.4590*** (1.5281) 15.1660*** (1.5326) 15.1956 (17.8765) 15.1760 (22.5404) Constant -7.0951*** (0.4811) No No No No
Table 7 - continued
This table looks at the output of different incremental regressions of abnormal returns (or ARs) onto two dummy variables for two different types of regulation (MAD I and MAD II). In column 1, the panel regression is performed with constant; in column 2, the constant is excluded and controls are included. Fixed effects are included in column 3-4-5: in particular, country fixed effects are included in column 3; industry fixed effects are included in column 4; and both types of fixed effects are included in the estimation displayed in column 5. Robust standard errors are reported in parentheses. The following syntax is also used: *significant at
10% statistical level; **significant at 5% statistical level; ***significant at 1% statistical level.
The regression’s output calls attention on the main variables: MAD I and MAD II. It is important to note that all the coefficients for MAD I and II are significant at the 5% significance level across most specifications of the model. Looking at specification (1): the MAD I dummy is significant at the 5% level, and presents a positive coefficient, pointing at a positive relation with the Abnormal Returns. This suggests that when the directive was in
Control variables No Yes Yes Yes Yes
Country fixed effects No No Yes No Yes Industry fixed effects No No No Yes Yes N 168,640 168,640 168,640 168,640 168,640 Adj R-sq (within) - 0.0005 0.0011 0.0005 0.0007 Wald-chi2 or F-value 283.30*** 177.36*** 27.94*** 25.62*** 12.87*** Prob>chi2 (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)
place at the moment of the announcement, the related abnormal returns were 8.1543% higher than abnormal returns related to announcements that happened when MAD I was not effective yet. The coefficient for MAD II is also positive, of higher magnitude, and statistically significant. This would suggest that in principle, any abnormal returns’ related to announcements that took place when MAD II was effective present a higher magnitude (4.0532%) then returns related to times where the directive was not effective. By excluding the constant in the regression model in output (2), and also adding control variables to the regression model, it is shown how the magnitude of the coefficient for MAD I changes significantly; but the one for MAD II does not at all. In this case, the positive effect of MAD I is suggested to be lower than before, providing evidence in favor of the hypothesis that the constant term seemed to absorb some of the predictive power of the MAD I variable8. The coefficient for MAD I changes in sign: being negative in (2), it would suggest that if the directive was in place at the moment of the announcement, the related abnormal returns were 0.7168% lower than abnormal returns related to announcements that happened when MAD I was not effective yet. MAD II’s coefficient is still positive and significant: this would suggest that the introduction of the regulation did not reduce the sample’s abnormal returns, but acted in the opposite direction, inflating them.
Columns (3), (4), and (5) show fixed effects regressions: all three specifications differ from each other, and from (2): the significance of the
8 The other specifications automatically correct for this problem by the use of
Fixed Effects. That is because a fixed-effects model eliminates everything that is constant within a panel unit. That is the very purpose of running a fixed effects model. The constant is - as the word implies - constant within a panel unit and does therefore not contribute to the parameter estimation.
coefficients for the main variables in (3) confirms that there is indeed cross-country variation in the way abnormal returns were affected by the regulations, and confirms Christensen et al (2010)’s theory. However, little interpretation of cross-industry variation is suggested by specification (4), that displays non-significant results. Comprehensively, both types of fixed effects – in (5) - also yield non significant results. This has important implications for hypothesis 3.
Drawing the lines on this approach, this outcome would economically imply that while the effect of MAD I is well positive only in one specification – meaning that in that case, it lowered the amount of abnormal returns during the sample period; MAD II’s results on abnormal returns as a measure of insider trading are negative across all regression exercises. Thus, MAD I had the effect expected by regulators in only one case; hence, looking at the overall outcome, both MAD I and MAD II were quite surprising in their contributions: instead of lowering the presence of insider dealing after their introduction, they seemed to have contributed to its spreading, and even significantly so.
The ambiguous result for MAD I is interesting, for it clashes with the outcome intended by the regulators. Moreover, an explanation for the behavior of abnormal returns around MAD II in this sample may be found in some specific and unaccounted macro-economic or financial occurring over the last two years, or it could rather be due to the early date of implementation of the regulation compared to the Directive: the fewer number of observations in the sample for the period After MAD II could be a consistent driver of this result, keeping in mind that for all of the countries in question, the regulation has been introduced less than a year ago. Both examinations call for an
interesting future analysis on the same topic, with a special focus on the current regulation in place (MAD II): it would prove interesting to perform a similar study in the future, by the time the Regulation will have already been rooted in the regulatory and corporate system for a time span equal or longer than the Directive’s (see Section 7.2).
Section 6
Robustness checks
Before delving into the specific of the robustness checks performed on the analysis, it is important to state that each specification of the Incremental
Regression Output section was performed by using robust standard errors: this
already involves correcting for heteroskedasticity and high values of standard errors.
6.1 Clustered and Newey West Standard Errors
Christensen et al. (2010) performed additional checks for their analysis, in the form of clustered standard errors regressions. As the way this analysis presents itself is very similar to their analysis, using the same check might prove useful to assess the robustness of the results. Therefore, the same regression exercises of Table 7 in Section 5.2 are run with an additional option for clustered standard errors according to Country.
An additional robustness check to the standard panel regression approach used in Section 5 (Results) involves the use of Newey-West standard errors: clustered and Newey West Standard Errors are jointly calculated for the same of the specifications examined in Table 7, and displayed in the brackets.
TABLE 8: CLUSTERED AND NEWEY-WEST STANDARD ERRORS OF ABNORMAL RETURNS (DUE TO INSIDER TRADING) ONTO MAD I AND MAD II DUMMY VARIABLES
Dependent variable: Abnormal Returns (ARs)
(1) (2) (3) (4) (5) Incremental regression with constant, but no controls Incremental regression with controls Incremental regression with country fixed effects Incremental regression with industry fixed effects Incremental regression with country and industry fixed effects MAD I 8.1543*** (0.4250) -0.7166*** (0.1176) 7.4316*** (0.8156) 7.8901 (9.6501) 7.4286 (14.3382) MAD II 4.0533 (17.2167) 12.4590 (19.9105) 15.1660 (20.2154) 15.1956 (27.0642) 15.1760 (28.8101) Constant -7.0951*** (0.3631) No No No No Control variables
No Yes Yes Yes Yes
Country Fixed Effects No No Yes No Yes Industry Fixed Effects No No No Yes Yes N 168,640 168,640 168,640 168,640 168,640 R-squared 0.0002 0.0005 0.0011 0.0005 0.0007 F-test 1.26 (0.3036) 60.45*** (0.0000) 70.69*** (0.0000) 65.98*** (0.0000) 13.11*** (0.0000)
Table 8 - continued
This table looks at the output of different incremental regressions of abnormal returns (or ARs) onto two dummy variables for two different types of regulation (MAD I and MAD II). In column 1, the panel regression is performed with an included constant; in column 3, controls are included. Fixed effects are included in column 3-4-5: in particular, country fixed effects are included in column 3; industry fixed effects are included in column 4; and both types of fixed effects are included in the estimation displayed in column 5. Newey-west and country-clustered standard errors are reported in parentheses, instead of OLS standard errors values. Some values could not be reported due to multicollinearity of the independent variables. The following syntax is also used: *significant at 10% statistical level; **significant
at 5% statistical level; ***significant at 1% statistical level.
The clustered and Newey West standard errors’ check displays the consistency of the previous results: the significance of the coefficients is similar to the one showed in Table 7 for most variables, except for MAD II. The coefficients for the independent variable MAD I are significant at a level of 5% in specifications (1), (2) and (3): the effect of MAD I results again significantly positive (decreasing) on the abnormal returns of the sample in specification (2), and significantly negative (increasing) in (1) and (3), confirming that the Directive caused a significant reduction of market abuse in only one of the proposed specifications. On the other hand, the coefficients for MAD II are significant at the 5% level across none of the specifications. This hints again at the possibility of low predictive power of MAD II due to the recent enactment of the Regulation (see Section 7 for further discussion).
Furthermore, coefficients for the main independent variables in specifications (3), (4) and (5) are not significant at the same statistical level: this result supports the finding that using country-clustering, and accounting for cross-country or cross-industry differences, does not really have much
meaning for this analysis, as it does not suggest any significant variation in the result due to the different implementation dates of MAD I at different points in time per country. Hence, it is again proved that the results of Section 5.2 are indeed robust in their assessment of the regulation’s impact onto the measure for insider dealing.
Section 7
Conclusion
7.1. Discussion
The effects of the enactment of the Market Abuse Directive (hereby defined as
MAD I) on a measure for insider dealing in EU during the period 2001-2015
has been studied in this analysis; as well as any possible developments due to the introduction of an amendment to the Directive, called Market Abuse
Regulation (hereby defined as MAD II).
Cumulative abnormal returns and cumulated average abnormal returns calculated via the event study technique for the sample around corporate news announcements suggest a lower volume of insider trading activity during the period following the introduction of the MAD I, and a higher volume of such activity during the period following the introduction of the MAD II. This result is only partially endorsed by the incremental regression approach performed on the sample, where a measure for insider dealing (abnormal returns) is used as dependent variable to be regressed onto a dummy variable for MAD I, a dummy for MAD II, and a set of control variables. The regression outcomes suggests that the MAD I only has a positive (decreasing) and significant effect on the amount of insider trading in the sample when fixed effects are not included in the model; while MAD II has a consistently negative (increasing) and significant effect on insider dealing. Newey-West and clustered standard errors calculated for the same regression exercises confirm the significance of the variable MAD I; however, MAD II is found to be
non-significant in this case.
Furthermore, a cross-country and cross-industry variation in the model is investigated: the fixed effects regressions performed for Country and (macro) Industry separately, and then jointly, all report non-significant coefficients for the main independent variables, suggesting no interesting cross-sectional variation in these two terms. This conclusion is confirmed by the use of clustered standard errors as a robustness checks.
Hence, the results provide mixed evidence in favor of the hypothesis - formulated at the beginning of this research - that the MAD I indeed reduced the amount of market abuse after its enactment, adding onto previous literature on the matter (Prevoo and Weel, 2010). Moreover, the expected strong cross-sectionality of the impact on insider trading is not empirically ascertained: this suggests that the MAD I did not have any particular stronger or weaker effect for countries which embodied the legislation earlier (than others), or for certain industry types.
Additionally, the negative (increasing) effect of the MAD II is also surprising to see. It is, however, to be noted that the MAD II has just been recently introduced in EU (as of June 2014 for all the EU countries used for this analysis), and it may be pre-mature to assess its impact on the market and market abuse measures, as its full effect may not be present yet. Moreover, the period following MAD I and prior to MAD II is consistently longer (about 11 years), proving for a more substantiated estimation of the MAD I’s impact onto market abuse and insider dealing. Therefore, this difference in length of time since the enactment of the laws has to be taken into consideration (which also implies a difference in the number of amendments, impact assessments, and tests usually performed after a law has
