Insider transactions and abnormal returns: Empirical evidence from the Dutch stock market

Insider transactions and abnormal returns: Empirical evidence from the Dutch stock market 05 07 2010 University of Amsterdam, FEB MSc Business Economics Finance Track Robin Elias 0454508 Supervisor: Dr. J.W.T. Bogers

Insider trading and abnormal returns in the Dutch stock market Abstract Regulation on the abuse of private information in trading is receiving world wide attention. To strengthen the position of the uninformed trader, rules are becoming stricter for those with easy access to so-called inside information. This thesis concentrates on the occurrence of abnormal returns when insiders trade in their company s own stock. A classic event study is conducted on the Dutch primary stock exchange from 1999 to 2008. Over 2200 insider sale and purchase transactions are examined. The dataset is divided into subsets to define specific types of trades or companies. First buy and sell orders are examined separately. Next trade size and firm size are examined. Also effects on the long and short term are taken into account. Empirical evidence is found on abnormal returns caused by insider trading. This implies that insider trading in the Netherlands cannot be neglected and should be a subject for further research. 2

Robin Elias Table of contents 1. Introduction 4 2. Motivation 5 3. Literature review 6 3.1 Relevant research 6 3.2 Related research from other countries 15 4. Hypotheses 19 5. Research plan 21 6. Data 27 6.1 Data collection 27 6.2 Data construction 27 7. Results 30 8. Interpretation and discussion 35 9. Conclusion 38 References 40 Appendices 42 3

Insider trading and abnormal returns in the Dutch stock market 1. Introduction Regulation on the abuse of private information in trading is receiving world wide attention, especially within the field of legislation, for the past two decades. To strengthen the position of the uninformed trader, rules are becoming stricter for those with easy access to so-called inside information. Especially the corporate insiders who can trade in their company s own stock receive a lot of focus on this topic. These insiders have the explicit knowledge on future company performance which is not available to the public. Each year more countries implement a government body that controls and reports the transactions made by insiders. In the Netherlands this is done by the Dutch Authority of Financial Markets (AFM). On the first of October 2005, The Market Abuse Directive was implemented in the Netherlands. The intention is to further narrow down the existing European system relating to market abuse, to expand it and partly to harmonize it in order to protect market integrity within Europe more effectively. Market integrity is very important if the markets in financial instruments are to function efficiently. Investors are to be protected. Market integrity is one of the conditions for maintaining investors confidence in the markets. A lack of confidence can result low participation of investors and other parties, in the activities on these markets. The consequences of the implementation of the Market Abuse Directive as of 1 October 2005 included an extension of the ban on market manipulation and the transfer to the AFM of supervision on the publication of price-sensitive information by companies listed on the Euronext Amsterdam stock exchange. In addition, a transparency regime was introduced for publicists who make investment recommendations, as well as an obligation for investment firms to disclose any reasonable suspicion of insider trading or market manipulation. Finally, existing provisions relating to insider trading, the disclosure regulations for 'insiders' and the insider rules were partly adapted. 1 Insider trading is regulated and reported frequently within a set time frame. But there still are strong believes that insiders trade with private knowledge and a predetermined reason. If this is actually the case, it will be very interesting to know for the uninformed investors, when making investment decisions. 1 Source: Dutch Authority for Financial Markets. http://www.afm.nl/en/professionals/afm-voor/effectenuitgevendeondernemingen/marktmisbruik/voorwetenschap.aspx 4

Robin Elias An approach to test if insiders indeed trade with private information is to examine above average stock performance after insider transactions occur. If statistically significant abnormal returns caused by these insider trades are found, one can conclude that insiders outperform the market when they trade. In this thesis a research on insider trading and stock returns is conducted on the primary stock exchange of the Netherlands, the AEX-index. A classic event study provides empirical evidence on abnormal stock returns caused by the signal of insider trading. This will finally result in an answer to the research question of this thesis: Does legal insider trading, on the Dutch primary stock market, result in abnormal returns? 2. Motivation A key characteristic of online brokerage firms and investment banks is to look for new ways to develop and implement investment strategies. When it comes to outperforming average market returns, out-of-the-box approaches are found to be interesting. One approach that has grown rapidly during the last decade is behavioral finance. It basically challenges the way people traditionally look at financial markets, stock prices and returns. The classic models are based on quite a few restrictions and assumptions. Without these assumptions, which in real life seldom are met, these theories cannot be used to predict trends with certainty. This is the part that is interesting to explore. People often do not behave rationally, although most of the classic theories assume so. As a consequence, a lot of factors should be taken into account when predicting future stock performance. These topics are eagerly waiting to be tested for abnormal price behavior as a result of imperfections caused by numerous publicly available signals. The signal this thesis focuses on is insider trading. The perfect market theory suggests that all public available information is reflected in stock prices. Still abnormal stock movements are witnessed almost everyday. One of the reasons these movements occur is unexpected company performance, which often are found around dates that companies present their yearly or quarterly figures. For every investor it would be a dream to predict these trends. Since this is virtually impossible for a random investor, it might be interesting to look at who might come 5

Insider trading and abnormal returns in the Dutch stock market close. The logical approach is to look at the people who work at the top level of the subject firms. They are the ones who have first hand information on (future) company performance. Since they trade their own companies stock on a regular basis, it might be interesting if mimicking them could result in a profitable investment strategy. To examine this on an academic level, this thesis will research abnormal returns after insider transactions. This thesis will continue as follows. First a literature review will be provided. From the resulting framework hypotheses are assumed. The framework also acts as an incentive for the methodology for the research plan, which will be discussed after the hypotheses. Next the dataset construction is reviewed. In the final part, the test results will be shown and will be interpreted. At the end a conclusion on the research will be provided. 3. Literature review To give a complete overview two different aspects of previous research will be discussed. First much referred to pioneer work will be looked at. Next empirical evidence from other countries than the Netherlands will be discussed. This will lead to a framework to support the hypothesis and research methodology of this thesis. 3.1 Relevant research An often quoted article, is the one by Finnerty (1976)[10]. Though somewhat dated today, at that moment in time the research complemented the existing literature in a couple of ways. It was the first to use precise price per share and date of the insider trades. Prior to 1965 these data were not reported to the S.E.C., making thorough analysis more difficult. Furthermore, Finnerty (1976)[10] used a more extensive dataset of companies involved than had been done before, making his research comprehensive. Next to this, also the effect of insider trades on the strong-form of the efficient markets theory was researched. In previous publications there was just a focus on abnormal returns after insider transactions. A four year dataset was collected from the NYSE. In total over 30.000 individual transactions were taken into account. First they were separated into a purchase and a sales group and monthly portfolios were constructed. Subsequently all insider transactions executed in one month were pooled together and the net result 6

Robin Elias was labeled as a buy or sell portfolio. Next these monthly portfolios were used to measure abnormal returns in the following eleven months. To test whether the strong form of the efficient market theory holds, Finnerty (1976)[10] examined if insiders could outperform the market. When the theory indeed holds, this would not be the case, since no individual can have higher trading profits because of private information. According to the theory, all available public and private information is fully reflected into stock prices. However, the results of the research reject the strong-form of the efficient markets theory. Finnerty (1976)[10] finds statistically significant abnormal returns on a 10% level for the buy portfolios. He concludes that this is evidence that insiders can earn above average returns when buying their companies own stock. Also the sell portfolios are significant on a 10% level, implying that the stocks insiders sell tend to decline more than the average market declines. In general one may conclude that insiders are capable of outperforming the market. It is not uncommon for regulatory changes to occur when using a dataset that covers a couple of years. To examine whether these changes have any effect on the characteristics of insider trading, Jaffe (1974)[13] looks into three major and influential changes in insider trading laws in the USA. The effect on the volume and the profitability of insider trades is examined. Jaffe (1974)[13] first establishes that insiders indeed earn abnormal returns. The next step is to analyze insider trades just before and after each of the three chosen regulatory changes. By doing so, the author eliminates the influence of other factors. Previous research normally follows a longer time-horizon, but fails by neglecting other variables that could play a role in such a time-window. Also the combined effect of the three events is investigated. The conclusion is that the regulatory changes do not have any effect on the characteristics of insider trading. The volume does not change significantly and neither the profitably. The research ends with stating that insiders seem not to be concerned about tighter regulation when it comes to being fined for using inside information. One of the explanations the author gives is the relative low maximum amount one can be charged when found guilty. The penalties are often a small amount compared to the profits made. 7

Insider trading and abnormal returns in the Dutch stock market Seyhun (1986)[17] follows a different approach for testing the effects of insider trading. His aim is to reinvestigate stock price behavior after insider trades occur and to check if, unlike previous research results have shown, the efficient markets hypothesis does hold. Reason to believe his approach will come up with different results lies in the alternative approach on how he conducts his event study. Most of the studies Seyhun (1986)[17] refers to, use the CAPM for measuring expected returns. Although this is an acknowledged and often applied method, the CAPM results in potential biases when measuring expected returns of securities. Therefore the results that are found using this model must be interpreted with caution. To overcome this problem, Seyhun (1986)[17] uses the Fama market model to measure the expected returns for his research. The choice for the market model is because of the expected value of the prediction errors. In the CAPM these residuals contain a systematic bias, whilst in the market model they have an expected value of zero. Hereby the bias is avoided and hence the results are more reliable. In the evaluation of abnormal returns the actual dates of reporting by the insider are used as a signal. Theory suggests that this is the moment the market reacts to the publicly available information and therefore is the correct time to start measuring for abnormal returns. Approximately 60.000 transactions are evaluated in a time period from 1975 to 1981. Seyhun (1986)[17] divides these transactions into buy and sell orders, the level of the insiders involved, equity size of the firm and the monetary volume of the trade. Overlapping sales and buys are netted into the one that exceeds the other. These sub datasets are analyzed separately to find possible effects of these different characteristics of the insider trades. The results show that insiders are capable of predicting abnormal stock returns in their companies stock. They tend to purchase stock before an abnormal rise in stock prices and sell prior to an abnormal decline. Outsiders on the other hand cannot profit from mimicking insiders. When accounting for the bid-ask spread and commission fees, their abnormal returns are negative. This is consistent with market efficiency, where no one can profit from publicly available information. Also a negative correlation between expected losses incurred by insiders and firm size is found. Meaning that the bigger the firm, the smaller the losses. This can be explained by the theory that larger firms are better priced and have smaller bid-ask spreads. Therefore the loss of trading is less. 8

Robin Elias As a starting point for their study, Givoly and Palmon (1985)[12] accept insiders that outperform the market as a given fact. Instead of assuming that this is caused by the use of inside information, the authors try to find other explanations. The first reason could lie in the effect of the trade itself. Insiders are mimicked and closely watched all the time by large groups of investors. If these groups act according to the trades of the insiders, this may create a huge flow of transactions in the same direction. These, on their turn, could cause the observed abnormal returns. Next to this, also the subsequent publication of company specific news following the insider trade is used as a possible explanation for abnormal returns. These publications often have quite an impact on stock prices. Using these as an extra signal, could create some insight on the capability of insiders to forecast profitable situations. Givoly and Palmon (1985)[12] use news publications of the Wall Street Journal as extra signals. The results support the latter assumption. Insiders indeed are capable of identifying profitable as well as unprofitable situations. According to the authors this does not mean that insiders exploit private information. They merely have a good sense on how future company performance is going to be and are in a position to identify this before other sources are capable of doing so. Also the effect of just the trade itself is proven to have a significant impact on the formation of abnormal returns. The overall conclusion by Givoly and Palmon (1985)[12] is that the trades themselves, and not the information they could contain, trigger a flow of transactions that result in abnormal returns. A more theoretical approach is followed by Ausubel (1990)[2]. By constructing a two-period model he examines the effect of government regulation on insider trading. His research extends more into the field of game theory, but comes up with some interesting conclusions for financial purposes as well. Ausubel (1990)[2] makes a difference between two possibilities of government intervention. On one side there is the consensus that if regular investors expect insiders to take advantage of them in trading, they will reduce their investment activities. This means there will be less liquidity in stock markets because of a lack of confidence in how private information can be extorted. So, on the other side one can claim that insider trading regulation promotes confidence in markets. Insiders can not profit, or at least to a lesser extent, of their explicit knowledge of future firm behavior. 9

Insider trading and abnormal returns in the Dutch stock market Using these two thinking-patters of outsiders and the way insiders react to these, a model is constructed which examines the influence of regulation. The expected outcome is that with regulation there will be a so called Pareto improvement. This means that while one person benefits, no one else has a negative effect, but only to a certain extent. Within the scope of this thesis, it is not that important to understand every detail of the model, but the basics are definitely worth looking at. There are two periods, two goods and two agents. The agents are the insiders and outsiders. The first period influences the second one. In the first period no private information is received by anyone, and the agents choose to which extent they invest in goods. In the second stage private information is given to the insiders and a double analysis is done, one where there is no regulation and one where there is. In stage two outsiders react on insiders investments. Comparing the two stages gives Ausubel (1990)[2] his answer to his central question. Given the fact that there is or isn t a regulatory regime, insiders may abstain from trade or not. The decision is made, by assumption, purely on the way the insiders think that the outsiders react on their trade: insiders always trade to maximize their own profit. After comparing the different outcomes Ausubel (1990)[2] concludes that when outsiders think that insiders will take advantage in a later stage, they will invest less in the beginning. Effective regulation, on the other hand, improves the return on outsider investment in the second stage en therefore positively influences the investments in the first stage. In other words, regulation on insider trading is beneficial for both parties involved. Although more theoretical, Ausubel (1990)[2] might still be useful for the research in this thesis. The importance of regulatory implementations on insider trading surely cannot me neglected. Therefore empirical evidence on the way regulation influences market efficiency and investor behavior might be crucial when discussing possible outcomes from this thesis. One of the most extensive researches on insider trading is from Lakonishok and Lee (2001)[16]. They examined all companies listed on the NYSE, AMEX and NASDAQ over a twenty year period from 1975 1995. In total over one million cases of insider trading were used for their computations. Motivated by the general consensus that executives know their business better than any analyst, Lakonishok and Lee (2001)[16] 10

Robin Elias try to contribute to the existing literature by conducting a comprehensive research on the market s response to insider trading. They examine market reactions around the dates that insider transactions occur and are reported. Abnormal returns from daily data are summed over a five-day period from the event date. Expected is that when insider trades contain information, the market should respond substantially around the signal dates. This would be consistent with the efficient market theory. Next to this, Lakonishok and Lee (2001)[16] expect, based on previous research, the significance of insider trades to depend on company size. Larger companies should in general be priced more efficiently. Exploiting insider trades as a signal to benefit from abnormal returns should therefore be more difficult. On the other hand, in smaller companies stock there should be more potential. To test if indeed firm size plays a role in trading on signals from insiders, Lakonishok and Lee (2001)[16] divide their sample into three different groups. They construct deciles based on market capitalization. From these deciles they label companies as small, medium or large. Next to this they also clear their data on a few points. Firms with stock prices less than USD 2,- were excluded, as well as trades of less than 100 shares. This way there is a better focus on the more meaningful trades. In addition, the direction of the trades, buys vs. sells, has been examined. Previous research suggests that sales do not contain useful information whilst buys do. Their findings are mostly consistent with the existing literature. The 5-day abnormal returns around the event days are not significant, meaning that the market dismisses the information. Firm size does seem to play a role in the statistically significance of their results. As expected, insider trades in smaller firms are significant whilst the trades in medium to big firms are not. Lakonishok and Lee (2001)[16] explain this phenomenon by claiming that price pressure explains these results. Furthermore, inside information in smaller firms might somehow leak to the market before the actual trades take place. But overall the market seems to show only a small response around the signal dates. Also consistent with their expectations are the directions of the trades. Only insider buying shares are meaningful to use as a signal. Jeng et al. (2003)[14] adapt a different approach when it comes to measuring the effect of insider trading. Instead of looking at the informative effects to other 11

Insider trading and abnormal returns in the Dutch stock market investors, the return earned by insiders themselves is used as a variable in their models. Due to data limitations, the exact holding period of insider transactions cannot be derived from insider reports. Therefore all analyses, provide reliable estimates of answers, but do not perfectly reflect the real situation. To overcome this problem, Jeng et al. (2003)[14] construct value-weighted portfolios that all are held for the standardized period of six months. By following this approach, performance evaluation methods can be used. The authors claim to be the first to introduce these methods into the research on insider trading framework. The portfolios that are constructed are divided into a couple of criteria to look for insider trading characteristics. Distinctions are made for: direction of trade, volume, firm size and insider position. The first three criteria did not provide striking results. Purchases yield more return for insiders than sale transactions. The volume of the trade is positively correlated to the insiders return. Insiders earn higher abnormal returns in smaller firms than in big ones. The insiders position within the firm has the only outcome that Jeng et al. (2003)[14] did not expect. The results show that the position does not matter; higher management does not outperform lower executives. In their 2001 study, Carpenter and Remmers [6] examine whether executive stock option exercising contains significant information for outsiders. Reason to believe this, is caused by the general idea that executives have private information available about future company performance. The sample they use contains a change in the regulatory regime which, in theory, should have an impact on their test results. The major change in the regime is the elimination of the obligation to hold stock from option exercise for at least six months before selling it. Therefore they decide to split their sample into sub samples to also test if this change actually has an effect. Next to splitting the sample into two different groups by date, they also examine the effect for differences in firm size. Theory suggests that there will be different outcomes for small and big firms. They found their answers by testing for abnormal stock price performance after the option exercise of the insider. Short- as well as long-term post-event stock performance is examined. 12

Robin Elias Their results show some differences between the created sub samples. Before the regulatory change they found significantly positive abnormal returns, suggesting that insiders used private information when exercising their options. In the period after the change, only significant, negative returns are found at small firms. The other sub sets show no significant evidence of insiders using private information. Two reasons are given for the differences between the findings of the sub samples. Their sample consists almost completely of medium to big firms. Therefore smaller firms hardly influence the entire sample. Next to this, the elimination of the holding period has altered the way one can look at option exercise. Because an executive can immediately sell his shares, exercises tend to look a lot like normal sales transactions. These transactions are most of the time done for diversification and liquidity reasons and not related to the use of private information. Extensive research on the effect of the existence of insider trading laws on the cost of equity worldwide was done by Bhattacharya and Daouk (2002)[5]. First they obtained worldwide information on trading laws in countries with stock exchanges. Next they tested if prohibitions on insider trading affect the cost of equity in the different countries. The test stage was divided into two sub-stages. First there is the implementation of insider trading laws. The second stage covers the moment of real enforcement against an insider trade. By choosing not just one method for measuring the effect on the cost of equity, Bhattacharya and Daouk (2002)[5] try to find a thorough answer to their central question. Next to descriptive statistics like returns, turnover and volatility; they also use different international asset pricing models to test the effect. Finally they also use surveys of country risk forecast for credit ratings as a variable in their tests. Their findings are quite interesting to take into account when one is dealing with the topic of insider transactions. All outcomes are identical throughout the different methods. This gives the reader a reliable overview of the real effect of implementation and application of laws on insider trading. The mere existence of insider trading laws does not seem to have an empirical effect on the cost of equity. Enforcement of these laws on the other hand results in a decreasing cost of equity. 13

Insider trading and abnormal returns in the Dutch stock market The lesson to be learned is that government agencies not only have to write legislation on insider trading, but they also have to monitor and prosecute abusive behavior to really make a difference. Another approach to the way insider trading affects market pricing is follows by Aktas et al. (2007)[1]. Instead of looking strictly at abnormal returns on trading days, they investigate the information content of legal insider trading. This is done by looking at the way these trades contribute to market efficiency. The approach used is to look at the effect on the dates the trades occur instead of looking at a longer horizon, as is done more frequently by other studies. Aktas et al. (2007)[1] claim that in this way they contribute to existing literature because normally the time horizon used ranges from one month to several months. Also there still is no clear answer to the profitability to a normal investor of mimicking insiders and by conducting their research they try to fill this gap. Using a five-year period, a sample of over 50.000 trades is collected from the US market and the associated fillings from the SEC. Basing their research on previous articles, one of the more important assumptions they make is based on Choridia et al. (2005) who claim that it takes just five minutes for investors to start taking actions that create efficiency after a signal has gone public. So the main focus of the research lies on the same day the insider trades occur. Next to this Aktas et al. (2005)[1] also look at the more standard research methods for examining insider trading. These latter methods come up with already known results. Two-day and fiveday abnormal returns come up significantly for purchases but not for sales. Also evidence is found that market impact does appear to increase with trade size. Finally, looking at the new part of their research, Aktas et al. (2005)[1] find that insiders indeed significantly contribute to faster price adjustments. Therefore they conclude that legal insider trading contributes to market efficiency. 14

Robin Elias 3.2 Related research from other countries Since most of the articles reviewed are based on American stock exchanges; in the following part empirical evidence from other countries will be discussed. Looking at the results from these publications, one might come up with some contradicting or maybe new interesting results. Next to American stock exchanges, other frequently used markets are the ones in England. Friedrich et al. (2002)[11] examine the short-term returns around insider trading days on the London Stock Exchange. Triggered by results from previous research, that did not find evidence on abnormal stock price movements in the month of the insider trade, Friedrich et al. (2002)[11] try to find an explanation. Their motivation for believing that there was something missing came from the idea that if abnormal returns occur immediately after a trade, the use of monthly data will not recognize this effect. If this would be true, it should be a phenomenon not neglected in drawing conclusions on insider trading. Next to this, the authors find two other reasons for their research to be interesting. First there is the welfare and regulatory concern to take into account if profitable insider trading is supported. This would also mean that the strong-form of market efficiency would be violated. Second, if outside investors would be able to make a profit by mimicking insiders, the semi-strong form of market efficiency would also not stand. In their research, Friedrich et al. (2002)[11] concentrate on smaller firms. Undervaluation is more likely and larger in smaller firms than in bigger ones. Therefore, the effect of insider trading should be substantial. Significant abnormal returns are found around insider trading days. These are consistent with insiders short-term market timing. Their economic significance should nonetheless not raise any concern. Other findings are that buy trades are more informative than sell trades and the strongest signals come from medium-sized trades. This could be explained by informed traders trying to hide their information based trades by not making big or block trades. One of the first to take a look at the influence of the position insiders have and the difference in their profitability when trading their own stock were Baesel et al. 15

Insider trading and abnormal returns in the Dutch stock market (1979)[3]. Their research differs from previous work because two subgroups are created, one consisting of regular insiders and one group of directors. As well as being the first to use Canadian data to provide independent evidence to compare. Also the use of a control variable in their models was an addition. For building the model the CAPM is used for estimating expected returns. Certain significant criticism is recognized by the authors on the use of this model, like the argument that the model is very difficult to test. Nonetheless they claim that the use of the random control sample offsets this argument. They base their research on the Fama theorem of efficient markets and the general consensus that directors have access to better or more valuable information than regular insiders. Especially when making their hypotheses we can see this. To focus on the innovative part of this research, the hypothesis of the difference between the two sub-groups has been taken into consideration. Based on the general consensus, Baesel et al. (1979)[3] expect that directors perform better when trading on inside information in comparison to regular insiders. The results support this view; also more regular answers were found. Overall insiders perform significantly better than other investors. In addition, insiders buy trades seem to be more informative than insider sell trades. The Taiwan Stock Market is used by Chiang et al. (2004)[7] in their research for abnormal returns earned by insider trading. After being triggered by results shown in America on the NYSE and AMEX, they decided to apply the methods used on the TWSE. As a starting point they made some assumptions based on previous research. Insider trading should result in abnormal returns. Firm size does matter; insiders in smaller firms tend to gain abnormal returns whilst in bigger companies this will not be the case. Finally, on the overall existence of insider trading, Chiang et al. (2004)[7] assume it aligns interests of different groups of investors, by giving private information open to the public and allowing it to be reflected in stock prices. In a four-year window, almost 300 companies and 7500 trades are being used for their research. Next to a traditional single-factor model with the CAPM, also the conditional alpha approach from Jensen is used. 16

Robin Elias Their findings do not confirm previous research and differ from what they expected. Insiders do not outperform the index. So other assumptions, like differences in firm size, are not supported either. The profitability of insider trading in the Spanish stock market is investigated in a paper by Del Brio et al. (2002)[8]. Next to profitability they also take a look at the information content of these trades and if an outsider can also profit from them by mimicking. At the moment of publishing this was a unique research on the Spanish market. The authors had little references but could nevertheless contribute significantly to the existing literature. In their research they try to conclude whether insider trades are profitable. If the effects of these trades on security pricing are harmful. Also the effectiveness of the trading laws implemented in Spain is examined. The latter was added to the research because, next to the existence of the laws, no case was ever filed for the prosecution on insider trading. As a benchmark for their research, the efficient markets theory was used to check the information content of insider trades. In order to check if the laws were effective they decided that this would be so if no insider could profit from personal information. Del Brio et al. (2002)[8] conclude that insiders in fact do earn abnormal returns in the Spanish stock market. Because of this possibility, the strong form of the efficient markets theory is rejected and the situation looks more like the semi-strong form. Another conclusion is that the current legislation is not effective in prohibiting insiders to use private information to gain profits. Next to these, another, quite rare, result came up. Del Brio et al. (2002)[8] found that only insider sale transactions statistically contain information and purchases do not. In other research, this is almost always the other way around. Eckbo and Smith (1998)[9] take on the Oslo Stock Exchange (OSE). Next to using a new market, they also try out new empirical methodology. Instead of more classic event study methods, they develop new ways to measure performance of insider trades. It is new because it mimics the true performance of these trades. In comparison to more conventional models this one is more accurate. A substantial part 17

Insider trading and abnormal returns in the Dutch stock market of the paper is, not surprisingly, also dedicated to explaining how their models are built. Since this paper will not use them, I will focus more on the results that are found by Eckbo and Smith (1998)[9]. Because their findings were contradicting in a number of ways, they also decided to test their sample with more traditional methods. The original results did not show any statistical evidence of insiders earning abnormal returns with their trades. Even for different trade characteristics like size and direction, nothing significant came up. Because this contradicted other findings, classic techniques were implemented and indeed there was proof of abnormal returns. The reason for this difference is not clearly explained by the authors, but it obviously has something to do with the use of their own developed methodology. Nonetheless Eckbo and Smith (1998)[9] claim that their results are solid and provide explanation that in a market like the OSE there probably is not much insider information that can often be obtained. Driven by contradicting evidence from previous research, as well as taking on a new country, Bajo and Petracci (2006)[4] take a look into the Italian stock market. Another challenge the authors take on is the weak and ineffective law enforcement on insider trading in Italy. On the other hand, this flaw makes it all the more important to come up with empirical results on the profitability of insider trading. Using standard event study methodology, Bajo and Petracci (2006)[4] do not come up with real surprising results. Around the trading days there is weak evidence of market reaction but on the longer horizon, one to three months, insiders indeed earn abnormal returns. Previous research on the Dutch stock market was done by Kabir et al. (1996)[15]. Especially a new addition to the already existing regulation on insider trading is used in their research. Since 1987 it is forbidden by Dutch law for insiders to trade two months prior to an earnings announcement. The impact this restriction had on stock prices and other related factors is the central topic. The results show that trading volume before earnings announcements declined after the implementation of the new restriction. According to Kabir et al. (1996)[15] this shows that the desired effect, being insiders not trading, was achieved. Excess returns on the other hand did not change. This could mean that the market itself seems to forecast good and bad news without the information provided by reported insider 18

Robin Elias trades. The next thing to question is if the regulation for insiders indeed was necessary, because without them trading abnormal returns did not change. It looks like beforehand insiders did not abuse private information or something alike. The next Section will use the presented research as a framework to build hypotheses. Especially the researches from Seyhun (1986)[17], Jaffe (1974)[13], Jeng et al. (2003)[14] and Lakonishok and Lee (2001)[16] are used. Given the research from this thesis in comparison to theirs, it seemed appropriate to hold their findings into account when formulating hypotheses. 4. Hypotheses Based on the literature review dealt with in Section 3 hypotheses are formed. In total we investigate five hypotheses. With the use of these we can give a complete and encompassing answer to the central question. Also the sub sample characteristics, which are discussed in Section 6, play a role in the formation of the hypotheses. Hypothesis 1 Insider transactions will lead to abnormal returns. Theory suggests that insiders have access to private information which they can use when trading stocks of their own company. This information provides the insider with better know-how on future performance of the company. Since they are allowed to trade these stocks they will exploit their position to earn abnormal returns. Using the insider trade as a signal, abnormal returns are expected in the period after the trade. Hypothesis 2 Buy transactions by insiders will result in statistically significant abnormal returns while insider sell transactions will not. Based on previous research, insiders tend to have other motivations for selling stock than earning abnormal returns. Sales are often executed for liquidity and diversification reasons. These events do not stand for insiders having information on declining stock prices. On the other hand, buy orders do tend to lead to significant abnormal returns. This is often interpreted ad the insider having good faith in the company s future performance. To test hypothesis 2, buy and sell orders will be 19

Insider trading and abnormal returns in the Dutch stock market investigated separately and their statistically significant influence on abnormal returns is tested. This thesis also tries to identify trade and company characteristics that possibly result in statistically significant abnormal returns. This leads us to the following hypotheses. Hypothesis 3 Insider trading in small firms will lead to statistically significant abnormal returns while this will not be the case in big firms. Big firms are known to be priced more efficiently than small firms. As a result, publicly available information tends to have hardly any effect on stock prices of big firms. When an insider trade gets public attention, the price will adjust accordingly. Small firms on the other hand should witness more of an effect because of less efficient pricing. Therefore abnormal returns will be realized after trades in small firms and not in big firms. Hypothesis 4 Big trades will lead to statistically significant abnormal returns while small trades will not. Higher volume trades shows stronger believe of the insider on future company performance. Therefore the public announcement of a big trade will cause a grander effect on the stock price. Small trades are found to be less informative by the public. These trades do not show real commitment of an insider to the company. Hypothesis 5 Abnormal returns will be statistically significant in the long run and not in the short run. Efficient markets theory suggests that all public available information is reflected in stock prices. The moment an insider trade is made public; this information will be reflected in the price. There has to be a reason for the insider to execute the trade nonetheless. This motive could be found in the long term performance of the company. Therefore only on this timeframe statistically significant evidence is expected to be found. 20

Robin Elias 5. Research plan To test for abnormal returns a classic event study will be conducted. For calculating expected returns the market model as described by Fama will be used. By choosing this model, like Seyhun (1986)[17] in his research on insider trading, the prediction errors have an expected value of zero, unlike the CAPM. This way abnormal returns realized from insider trading will not be overstated, Seyhun (1986)[17]. In the event study of this paper there are three indices, the index for all the stocks of the AEX index, denoted as j, the index for the days of the set time frame, denoted as t and the index for all the individual events, denoted as y. The event study contains the following steps: Step 1 The market model as described by Fama is used for estimating expected returns: r jt = α j + β j r mt + ε jt (1) With r jt is the total return for stock j over day t; α j and β j are the OLS estimates from the regression coefficients; r mt is the return on the market portfolio and ε jt is a random variable. As we have defined our return generating process, the next step is to calculate the expected returns. Step 2 The expected returns can be calculated from relation (1): E rj = α j + β j E rm (2) with (3) E(ε jt ) = 0 (4) With E rj is the expected return for stock j; α j and β j are the OLS estimates from the regression coefficients; E rm is expected return on the market portfolio and E(ε jt ) is the expected value of the random variable which by construction of the model is zero as shown by Seyhun (1986)[17]. 21

Insider trading and abnormal returns in the Dutch stock market Now that we have defined the model to estimate α j and β j, the following steps will rely on the outputs that will be found using the OLS method. Per event α j and β j are estimated according to the time frame that can be found in Figure 1. This will result in α OLS j and β OLS j. The combination of the latter two variables and the known stock and market returns will result in possible abnormal returns. This leads to step 3. Step 3 From relation (2) abnormal returns are computed using: AR jt = r jt (α OLS j + β OLS j*r mt ) (3) With AR jt is the abnormal return for stock j for day t; r jt is the actual return of stock j over day t; α OLS j and β OLS j are the OLS estimates from the regression coefficients for stock j and r mt is the actual return of the market portfolio. The next step is to add abnormal returns per event within the given timeline of this event study. This is done individually per event per stock. So we cumulate over time T. The window for adding the abnormal returns is denoted from t(1) as the day after the event, till t(5) as the 5 th day after the event for the short-term analysis and t(180) as the 180 th day after the event for the long-term analysis. The result will be a cumulative abnormal return per event for both time frames, CAR j. So for every event y, of stock j a CAR jy is calculated. Step 4 The cumulative abnormal returns per stock, first for the short-term and second for the long-term, are calculated as: CAR jy = AR j,t(1) +.. + AR j,t(5) CAR jy = AR j,t(1) +.. + AR j,t(180) (4)(a) (4)(b) With CAR jy is the cumulative abnormal return per event y, for stock j and AR jt is the abnormal return for stock j for day t from relation (3). 22

Robin Elias After calculating al the CAR jy s, it now is time to sum all of these and then average them. So in the next step all the events are cumulated and averaged for the entire (sub) sample of insider trades that is under evaluation. Step 5 The average cumulative abnormal return is calculated for all the events of all the companies using: (5) With CAAR is the cumulative average abnormal return for the whole (sub) sample; CAR jy is the cumulative abnormal return per event y, per stock j from relation (4); M is the total number of firms in the (sub) sample and N is the total number of events in the (sub) sample. The sign from relation (5) will give us information about the abnormal returns after the insider trade. If CAAR > 0, then the average cumulative stock return is higher than can be expected. The stock shows abnormal positive returns after the insider trade. If CAAR < 0, then the average cumulative stock return is lower than can be expected. The stock has an abnormal negative return after the insider trade. After defining the CAAR, we need to test it for statistical significance. This is necessary to draw valid conclusions on the hypotheses. The following two steps test if the CAAR from relation (5) statistically differs from zero. This will lead to the answer if insider trading indeed causes significant abnormal returns. Statistical significance in this paper is found by t-statistics. As discussed by Brooks(2008)[19], t-statistics give a measurement figure to draw conclusions when using samples as a representation for a population. To come to a t-value first the standard deviation of the sample is needed. 23

Insider trading and abnormal returns in the Dutch stock market Step 6 The standard deviation for the sample is calculated as: (6) With s is the standard deviation of the sample; N is the number of events; M is the total number of firms in the (sub) sample; CAR jy is from relation (4) and CAAR is from relation (5). The next step is to calculate the t-value. First the mathematical representation is given; next arguments are provided why exactly this t-test is used. Step 7 A traditional t-test is used: (7) With t is the t-value for the sample; N is the number of events; CAAR is from relation (5) and s is from relation (6). To test whether the found t-values support or reject the hypotheses, a couple of assumptions are made. To start we will look at why this specific t-test is suited for this research. The first assumption is that the abnormal returns are independently and identically distributed. By assumption they approximately follow a standard normal distribution if N is large enough. According to Jaffe (1974)[13] this will be the case for N bigger than 49. Since all (sub) samples satisfy this restriction, the assumption is legitimate. Since standard normal distribution is assumed, a traditional t-test is the most common test for a null hypothesis of no abnormal returns. It provides the user with t- values which can be interpreted when it comes to supporting or rejecting hypotheses. Cut-off points for the critical regions are calculated to draw a conclusion on the 24

Robin Elias significance of the t-values. For this the tables in Brooks(2008)[19] are used. A 5% significance level is used, which corresponds to a 95% confidence interval as used by the tables. Next the degrees of freedom are chosen per sub sample as N-2. This is standard for a traditional t-test according to Brooks(2008)[19]. In total there will be twenty event studies conducted on the different sub sets of the data. The first two are the purchase and sale transactions. Next there is the subset according to firm size, consisting of small en big firms. Per category, small and big, the sale and purchase transactions are split. So the firm size subset contains four different groups. There is the purchases group for small firms, sales group for small firms, purchases group for big firms and sales group for big firms. This is also done for the subset based on trade size. This results in a small purchases group, small sales group, large purchases group and large sales group. This comes to ten groups. These groups are analyzed in the short- and long-run, resulting in the total of twenty event studies. As discussed earlier, different timelines will be used in this study, one for the short-term analysis and one for the long-term. Both timelines will follow the layout of figure 1, but with different periods after the announcement date. Figure 1: Timeline of the event study For both the short-term and the long-term analysis the estimation window for the parameters of the market model will be from: - T 0 = 12 months prior to the event - T 1 = 1 month prior to the event 25