University of Groningen Msc - Finance
Master Thesis
An empirical analysis of legal insider trading in the Netherlands: Are insiders
able to beat the market?
Author: Kees van Bommel
Mail: k.vanbommel@hotmail.com Phone: 0641704041
Student number: s2144662
Table of Contents
1. Introduction ... 1
2. Literature review ... 4
2.1 Hypotheses ... 7
3. Insider trading legislation ... 8
4. Data and methodology ... 10
4.1 Data ... 10
4.2 Methodology ... 11
4.2.1 Calendar portfolio method ... 11
4.2.2 Event study methodology ... 14
4.2.3 Market-adjusted model ... 15
4.2.4 Mean-adjusted model ... 16
5. Empirical results ... 17
5.1 Long-term results ... 17
5.1.1 Robustness testing ... 20
5.1.2 Regression results for firm size and trade size ... 22
5.2 Event study results ... 24
Tables
Table 1. Overview of insider trading legislation in the Netherlands Table 2. Descriptive statistics for the AFM insider trading database Table 3. Correlation matrix of explanatory variables
Table 4. Equally weighted and value weighted OLS regression results for insider portfolios Table 5. Robustness tests for equally weighted insider portfolios
Table 6. Equally weighted insider portfolios based on firm size and trade size Table 7. Event study results using the mean-adjusted and market-adjusted method
Figures Figure 1. Visualization of calendar portfolio method Figure 2. Timeline for event study methodology
1
KEES VAN BOMMEL∗
Abstract
In this paper I investigate whether Dutch insiders are able to obtain abnormal returns. I use regulatory data provided by the AFM, which includes legal insider transactions between 2006 and December 2012. I find that insider purchases yield significant abnormal returns of around 6.5% to 11.5% on an annual basis. Insider sales lead to underperformance, although this is generally insignificant. Furthermore, I find that smaller transactions lead to higher abnormal returns. Transactions in small cap firms yield higher abnormal returns, but these are insignificant. In addition, I do not find outsiders to notice the insider transactions, as no impact is found on the stock price around the insider transaction date.
Keywords: Insider trading, Market regulation, Market efficiency JEL codes: G14, G18, K22
1. Introduction
Fairness has always been an important factor in the financial markets. Shefrin and Statman (1993) refer to market fairness as a “claim on entitlements”. They state that market participants have the right to equal information. Equal information means that market participants should have access to the same information sources. This means private or inside information should not be used in financial markets. However, history provides some examples of disputable trading activities. As early as the 1800’s the mighty Rothschild family possessed superior information due to a network of pigeon lofts throughout Europe. The birds were used to send information to their financial houses quicker than any others could do. The speed of this system helped them in making financial decisions. As the story goes, the Rothschild’s were the first to be aware of the defeat of Napoleon at Waterloo. They used this knowledge in financial markets to accumulate a fortune. Nowadays, the use of inside information in trading activities is regulated and illegal. To improve market transparency, disclosure of insider trades has been required already since 1934 in the United States. In the United Kingdom this is required since 1976. For the Netherlands however, similar regulation was introduced more recently. Although the use
2
of inside information has been illegal since 1989, complete disclosure of trades by insiders has been required only since 1999 in the Netherlands (Biesta et al., 2003). The regulations do not imply insiders cannot trade at all. Dutch law states, in line with directives of the European Union, that anyone who possesses private or price-sensitive information is not allowed to trade. Under these rules, it follows that trading in one’s own company is not illegal if one does not possess such information. The definition is similar to the definition given by the Securities Exchange Act of 1934 for the US market. Here, trading is banned for anyone who is in possession of material information. According to Meulbroek (1992) information is material if there is a substantial likelihood that an investor would find this particular information important in making his or her investment decisions.
From this it is clear that Dutch insiders, just like all other market participants, should only trade on publicly available information, resulting in an equal playing field for all. In effect, this implies that Dutch insiders should not be able to outperform the markets and realize abnormal returns. However, it is unknown if this is indeed true. It is possible that insiders have some gut feel, and more in-depth knowledge about the prospects and climate in their company relative to other market participants. Since this knowledge or gut feel cannot be labelled as “private”, “price-sensitive” or “material” information, trades based on this knowledge are not illegal. The aim of this study is to empirically measure the returns to long-term legal insider trading. It investigates if insiders are able to realize abnormal returns relative to a market benchmark, without performing illegal insider trades. The following research question arises from this:
Are Dutch insiders able to obtain abnormal returns?
Second, if insiders are able to do so, does the fact that they execute trades in their own companies shares affect stock prices? It is plausible that outsiders notice the insider trading activity, and incorporate this new information in the stock price. Contemporary literature for the Dutch market is scarce and provides mixed results as to whether insider trades are able to affect stock prices or not. Degryse et al. (2009) and Aktas et al. (2007) perform an event study to measure the short-term effects on stock prices resulting from legal insider trades. Although the first study provided evidence that insider trades are informative, the second did not.
3
and French (1992) can be made. Aktas et al. (2007) did not adjust for these additional factors of risk. Furthermore, the Dutch market remains relatively under researched compared to the U.S. or U.K. market, as data collection on insider transactions only recently started.
In the Netherlands, the Autoriteit Financiële Markten (AFM), which is the Dutch supervisor of financial markets, started collecting data on legal insider trades in 1999. For this study I use the AFM database on insider transactions in the Netherlands. This database provides an overview of all legal insider transactions in the Netherlands and is available from 2005 through 2012. This data is freely accessible at the AFM website. Large shareholders, insiders and family up to second degree have to disclose their transactions in their own company. For the remainder of this paper, inside trading refers to legal insider transactions covered in this register. It is assumed that no illegal trades are included in this register, as the AFM is responsible for monitoring and evaluating the fairness of the performed insider transactions which are listed in this database.
4
2. Literature review
From a theoretical perspective the subject of insider trading describes an information asymmetry situation. This means that some agents, like corporate insiders, possess more information than others, which creates the asymmetry. They could have more accurate knowledge of the company, which enables them to make a better estimation of the fundamental value of their company. Furthermore, they have better knowledge about corporate strategy, upcoming events and other happenings that could influence future stock performance. Ausubel (1990) argues that disclosure of insider trades helps to reduce the information asymmetry that exists between insiders and outsiders. Empirical literature indeed indicates the ability of insider transactions to affect stock prices. These results imply that the semi-strong form of market efficiency does not hold (Fama, 1970). Although the empirical literature on legal insider trading is extensive, the majority focusses on the U.S. market and/or is short-term orientated. Aktas et al. (2007) explain this by the lack of data for the European market. Until recently European countries did not have the legal obligation to notify insider transaction. So unfortunately, few comparable studies exist for the Dutch market.
Seyhun (1998) studies the US market using insider transaction data ranging from 1975 to 1995 for the New York Stock Exchange (NYSE), the NASDAQ and American Stock Exchange (AMEX). The dataset includes more than 300,000 insider transactions. The study analyses long-term performance, in which a purchase (sale) month for a stock is determined by netting insider buying and selling. Then the performance is calculated for a 12 month period after the purchase (sale) month. He finds an average outperformance of 4.5% for the purchase portfolio, whereas the sale portfolio results in an average underperformance of -2.7%. Moreover, he finds that two-thirds of this outperformance is realized in the first six months. Seyhun (1998) argues that for smaller firms it is easier for an insider to know information that outsiders do not, as these small firms receive less analyst and media coverage. His study indeed reveals a negative correlation between the firm’s market capitalization and the relative performance of the insider trade. Furthermore, he finds a positive correlation for the volume of the transaction and the position of the insider within the firm. This means the results are more pronounced for large trades and for trades executed by insiders holding a high position within the firm.
5
recently underperformed. Moreover, the sale portfolio disproportionally includes growth firms and recent outperformers. After correction for these risk exposures, the authors find significant outperformance of 50 basis points a month (6% on annual basis) for the buy portfolio. Contrary to findings of Seyhun (1998), the authors do not find significant underperformance of the insider sale portfolio. They argue that insider sales arise from different motives such as liquidity and diversification needs.
A recent study for the UK market is performed by Fidrmuc et al. (2006). They focus on the short-term effects of insider trading using event study methodology. They find that insider purchases and sales are both followed with significant abnormal returns. Moreover, they conclude that larger trades result in larger abnormal returns. For sales, the impact on stock prices seems to be much smaller. Another European study on legal insider trading focuses on the German market. Betzer and Theissen (2008) use a small dataset from July 2002 to June 2004. They find remarkably large abnormal returns, both for insider purchases and sales. For an 11-day event window the authors find cumulative abnormal returns (CAR) of 2% for buy transactions. When focussing only on the larger trades, they find CARs of 3.4% for buys and 4.4% for sale transactions.1 The authors explain these large outcomes by difference in regulations. German insiders are not restricted by a so-called blackout period. This is a period before important events or news is about to go public. Unlike Dutch insiders, German insiders are allowed to trade in this period.
The Italian market is studied by Bajo and Petracci (2006). The authors conduct an event study and find that in Italy legal insider trades result in relatively high abnormal returns. They find abnormal performance for a period of 10 days after the transaction date of 3.18% for purchases and -3.67% for sales. Unlike results for other countries, it is remarkable that in Italy the sale effect is larger than the outperformance of purchases. The authors argue that this could be due to the fact that they focus on insiders owning 50% or more of the shares, which suggests a relation between size of the shareholding and abnormal return.
For the Dutch market, Biesta et al. (2003) were the first to perform a study on legal insider trading. At the time of their study, the AFM only recently started collecting data on insider transactions. The authors were the first to use this data, and covered a period from April 1999 through May 2002. They exclude options, warrants and incomplete disclosures from their sample, resulting in 2,517 insider transactions. The aim of their paper is to determine the profitability of insider transactions. Their methodology involves forming buy and hold portfolios similar to the method of Seyhun (1998). They conclude that insiders as well as outsiders mirroring insiders are able to realize abnormal returns. For a six month holding period insider buy portfolios yield abnormal returns between 8.9% and 9.3%. They
6
did not find evidence for significant underperformance of insider sell portfolios. The authors also show results for short-term effects on stock prices. Using a 21-day event window they find only significant CARs of -1.9% for sales. However, when narrowing the event window, the CARs for buy and sale transactions are no longer significant. The second paper on the Dutch market by Aktas et al. (2007) also employs an event study. Their research covers the period between January 1999 and the end of September 2005, for which they include 822 insider transactions. The authors use a very short event window ranging from two to four days. Their results for short-term effects on stock prices are similar to the finding of Biesta et al. (2003). Their analysis shows that the financial markets’ response is often insignificant for purchases and sales, and that the abnormal returns associated with the transactions do not have the expected sign. For a four day event window they find CARs of -0.92% for purchases and 0.97% for sales. However, over a longer time horizon the significance of the results improve and the average cumulated abnormal returns turn positive for stocks purchased, and negative for stocks sold by insiders. For an event window of 41 to 100 days after the event date they find significant CARs 7.45% for purchases and -4.40% for sales. According to the authors this suggests that either insiders use long-term information for their trading activities or they are able to time the market. Furthermore, they find that small insider purchases in this event window results in higher CARs, with 12.80% for purchases compared to -0.07% for large insider purchases. This is contrary to the relationship found in the U.S. market by Seyhun (1998). They argue that insiders prefer several smaller transactions, which will not be noticed and do not alert the market.
The most recent paper on insider trading in the Netherlands is by Degryse et al. (2009). They use a longer time period from 1999 to 2008. They examine the short-term price effects around insider trading days. Unlike previous literature for the Dutch market, the authors not only focus on the trading of shares, but also warrants, options and other financial instruments. They find that purchases are followed by abnormal returns, which increase as the event window widens. For a 30-day event window the CAR for purchases is around 2% and highly significant. In their sample the authors distinguish between trades performed by members of the management board, which they call “top executives” and other insiders. They find that magnitude of the CAR for these top executives is larger, although it is not statistically different from zero compared to the other insiders.
7 2.1 Hypotheses
Taking the previous literature in account, I expect to find that the size of abnormal returns obtained by corporate insiders is related to firm size, direction (buy or sell) and the value of the trade. This leads to the following hypotheses to be tested:
H1 Dutch insider purchases (sales) are able to obtain abnormal returns
Furthermore, a negative correlation between the firm’s market capitalization and the relative performance of the insider trade is expected. Therefore insiders in small cap firms could be able to earn higher abnormal returns, which results in the following hypothesis:
H2 Dutch inside trades in small cap firms result in relatively higher abnormal returns
Another expected relation is that outperformance for insider trades is more pronounced for larger trades, from which the following hypothesis arises:
H3 Large inside trades result in relatively higher abnormal returns
Finally, the literature which investigates the ability of insider transactions to affect stock prices leads to mixed results. It seems to differ among geographical and jurisdictional areas. With respect to the ability of insider trades to affect stock prices, the results for the Dutch market are not clear and ambiguous. Therefore the following hypothesis is tested for the Dutch market:
H4 Trading of shares by Dutch insiders has a significant impact on stock prices in the short term
8
3. Insider trading legislation
The AFM is responsible for the supervision and enforcement of several laws applicable to financial institutions and markets in the Netherlands. One of its duties is to monitor insider trading activity in the Netherlands. Current Dutch law on insider trading is mainly an implementation of the Insider Dealing Directive 89/592/EEC and the Market Abuse Directive 2003/6/EC issued by the European Union. The spirit of this law is comparable to regulation in the U.K. and U.S. market, where it is stated that participants who possess information that is private or is price-sensitive are not allowed to trade. Private information refers to information not available or accessible to the public. Price-sensitive information refers to the ability that sudden revelation of information can affect the price of the company’s stock.
In addition to the prohibition of trading based on private or price-sensitive information insiders are required to report trades in their own company’s shares. The AFM defines insiders as persons such as managers, supervisory board members and other staff that are exposed to potentially private or price-sensitive information. This also includes the family up to second degree (spouse, children) of the insider. Moreover, Dutch law requires listed companies to set up so-called blackout periods. Listed companies should have a set of written rules that stipulate when an insider is not allowed to trade. The notification requirement regards not only trade in shares, but also derived instruments of which the value depends on the value of the share. These rules are applicable to all companies that are registered in the Netherlands, even if they are not listed on Euronext Amsterdam. The rules apply as well to foreign companies that have financial instruments listed on Euronext Amsterdam. Table 1 provides an overview of recent insider trading law in the Netherlands and the changes that are made.
Table 1: Overview of insider trading legislation in the Netherlands
This table provides an overview of legislation and most important changes regarding insider trading in the Netherlands that are made throughout the period 1999 up until 2012.
Date Article Content
Apr. 1999 Wte 1995 Art. 46b Notification of trades. Maximum delay of 10 days after end of month.
Sept. 2002 Wmz 1996 Art. 2a Trades by executives require immediate notification.
Oct.2005 Wte 1995 Art.47a Notification of trades is reduced to a maximum of 5 working days delay. Exception for trades below €5,000.
Oct.2005 Market abuse Decree 2005 Art. 2
Exceptions of notification if:
i) Shares are given as part of employee compensation scheme. ii) Exercising options obtained by employee compensation
scheme on expiration date
9
10
4. Data and methodology
The first paragraph of this section covers the AFM insider transaction database and the filters that are applied to determine the sample of this study. The second paragraph discusses the methodology; which consists of a calendar portfolio method and an event study method.
4.1 Data
The data for this study is provided by the AFM. The database includes all insider transactions from October 2005 through December 2012. The data covers the period right after the implementation of the European Market Abuse Directive as discussed in the previous section. The total number of disclosures between these dates is 11,038. The disclosures not only include stocks but also warrants, options and foreign denominated transactions. I focus on insider trading in shares only and following Biesta et al. (2003) exclude transactions in options, warrants and investment funds. Investment funds cannot be regarded as true insiders, as they are not involved in the daily operations of the companies they invest in. Furthermore, the sample is reduced to Euronext Amsterdam transactions only, since the focus of this paper is on the Dutch market. Next, a quick scan of the data shows that in some years insider trading is more common than in others. For example, 2006 and 2007 experienced active insider trading with around 50% of the transactions, while the year 2005 only counts for about 1% of the total transactions. For this reason the year 2005 is excluded from the sample.
11
Table 2: Descriptive statistics for the AFM insider trading database
This table shows the descriptive statistics for the AFM database after data adjustments. Insider transactions in investment firms, foreign currencies, options, warrants and other derived products are discarded from the database. As data on the year 2005 is limited, this year is not included in the sample. A total of (N=1889) transactions are included.
Variable 2006 2007 2008 2009 2010 2011 2012
No. of trades 377 638 235 129 200 178 132
Total trades as % of total data 20.0% 33.70% 12.40% 6.80% 10.60% 9.40% 7.00%
Number of buys 212 397 190 95 128 99 91
Number of sells 165 241 45 34 72 79 41
Avg. trade size buys (€) 81,234 68,053 93,098 65,753 63,578 66,857 95,342 Med. trade size buy (€) 67,115 21,450 20,244 17,558 17,530 14,239 30,960 Avg. trade size sell (€) 195,344 144,424 139,601 158,940 133,921 74,260 245,989 Med. trade size sell (€) 130,000 124,844 35,662 88,144 34,905 25,872 121,300
The descriptive statistics show that for all years insiders execute more buy than sell transactions. When observing the value of these trades, the table shows that the average and median value of the trades is in all years higher for sale transactions. The trade value for sale transactions is in most years double or triple the value compared to buy transactions. The years 2006 and 2007 experienced the heaviest insider trading activity. Together these two years make up almost 50% of the sample. It is noticeable that the number of sale transactions prior to the large market drop in 2008 is relatively high. The years 2006 and 2007 account for 60% of all insider sale transactions in the sample.
4.2 Methodology
In this paragraph I first cover the calendar portfolio method which is used to measure the long-term returns to insider trading. Thereafter I discuss the event study methodology which measures the short-term price impact resulting from insider transactions.
4.2.1 Calendar portfolio method
12
Figure 1: Visualization of calendar portfolio method
This figure shows the calendar portfolio construction method and the calculation of daily average insider returns. At each unique insider trading day, individual buy and sell portfolios are constructed. As 587 unique insider trading days exist between 2006 and the end of 2012, this amount of portfolios is constructed. Portfolios overlap in time and the daily insider return is calculated as the equally weighted (average) return of all portfolios running at this particular day.
Date 01/01/2006 02/01/2006 03/01/2006 04/01/2006 . . . . 31/12/2006 01/01/2006 Portfolio 1 02/01/2006 Portfolio 2 03/01/2006 Portfolio 3 04/01/2006 Portfolio 4 . . . . 31/12/2012
Daily return average average average average
The weight of the stocks in the calendar portfolios is calculated according to equation (1) below.
1/ ) (1)
The equation shows that the weight for an individual stock i is determined by 1 over , where N is the number of unique firms in which insider trading activity occurred at time t.
The returns on the stocks and portfolios are calculated as in equation (2) below:
, (2)
where is the return in stock i at time t. The RI is the total return index for the stock obtained from Datastream. The total return index accounts for dividends by assuming that all cash distributions are reinvested. The portfolio return is then easily calculated as the weights on the individual securities times their respective returns. The overall daily insider return at a particular time t is calculated as the equally weighted return (average) of all portfolios that are running at this time t.
Multifactor regression analysis
To test the presence of abnormal returns, an Ordinary Least Squares (OLS) regression is applied in the context of the Capital Asset Pricing Model (CAPM). The MSCI Netherlands total return index (MSNETHL) is used to serve as market proxy. The following equation is applied:
, (3)
13
term or residual from the regression. Jensen’s (1969) alpha ( is the intercept of the OLS regression and can be interpreted as the abnormal return of the portfolio. The beta (β) for the portfolios is estimated by regressing the portfolio returns on the market returns. It shows the sensitivity in changes of the portfolio returns in relation to changes in the market return. Moreover, is the market price for risk or market premium. It is the market return at time t minus the risk free rate at time t.
It is important to check if performance is attributable to other factors as Fama and French (1992) describe in their article. They argue that returns are related to other factors of risks. These factors are market cap (known as the size effect) and book to market value (known as the growth effect). Investors require additional return for investing in low cap and high market to book stocks. To adjust for these factors of risk, additional explanatory variables are added to the equation as shown below.
(4) The first part of equation (4) is equal to the previous CAPM model. The additional factors SMB and
HML are respectively the “small minus big” and “high minus low” premiums. The SMB factor
14
Table 3: Correlation matrix of explanatory variables This table presents the correlation matrix of the explanatory variables used in the OLS regression.
-0.19 0.61
-0.19 0.17
0.61 0.17
In sample differences
As final step I test for any in sample differences. As discussed in the literature section, results can differ for small cap firms and larger trades. To divide the sample in large cap and small market cap stocks, I use a method similar to Jeng et al. (2003). The division is done by taking the average market capitalization for all 48 firms between 2006 and 2012. The results are then sorted from high to low. The highest third of this distribution will serve as the large market cap sample. The lowest third serves as the small market cap sample. Then for both samples portfolios are formed and regressed to check if insiders in small firms indeed earn higher abnormal returns. Classification based on trade size is performed comparable to Aktas et al. (2007). Both the buy and sale transaction sample are sorted from high to low based on the total value of the transaction. The first quartiles of both samples then serve as the large trade samples, whereas the last quartiles serve as the small trade samples.
4.2.2 Event study methodology
15
windows can be set to capture the actual event. The construction of the event window and estimation window is shown in Figure 2 below.
Figure 2: Timeline for event study methodology
This figure show the timeline for the event study. The estimation window is fixed and used to calculate the mean return on a particular stock for the mean-adjusted method. The selection event window shows the time period in which different event windows can be selected. The event window has to lie within the range specified in the selection window. The actual event takes place at time 0 and this is assumed to be 1 day after the date noted in the AFM database.
Figure 1: Timeline event study Revisions of Stock Recommendations by Investment Banks. Sample selection
The event study is executed for buy and sale transactions separately. To ensure the quality of the data, several filter rules are applied to arrive at the final sample. First, non-Euro transactions and transactions not listed on Euronext Amsterdam are excluded from the sample. Second, to prevent the inclusion of delayed trades, only transactions with a value of €5,000 and up are included. This method is different from the method applied by Aktas et al. (2007) who focus only on transactions of 100 shares and up. However, their method does not ensure that delayed transactions with a value below €5,000 are excluded. The last step is netting the insider transactions as described by Fidrmuc et al. (2006). For example, if on a given day buy and sell transactions are made on a particular stock the direction of the trade is determined by subtracting the sale volume from the buy volume. If more stocks are bought than sold on that day the transaction is regarded as a buy event. If more stocks are sold than bought on that day, the transaction is regarded as a sale event. Following these adjustments and the elimination of the net transactions reduces the sample to a total of 478 events of which 326 a buy events and 152 are sale events.
The final part of the methodology covers the calculation of the test statistics for both the market-adjusted and mean-market-adjusted model.
4.2.3 Market-adjusted model
With the market-adjusted model the abnormal returns ! for stock " on day # are calculated by formula (5):
! , (5)
where is the return on stock i at day t and is the return on the market m, at day t. For the market return again the MSCI Netherlands is used as market proxy. The average abnormal
T0 = t-140 T1 = t-30 T2 = t-10 T3 = t-1 t T4 = t+1
Estimation window
T5 = t+10
Event window Selection event window
16
returns !! $$$$$$ for day # are then calculated by summing the returns at period # and dividing by the number of events as defined by formula (6):
!!
$$$$$$$ 1 % ! & '
(6)
Summing the !! $$$$$$ during the event window results in the cumulative average abnormal returns ((!! $$$$$$$$ . This (!! $$$$$$$$ can be used to draw inference about the event window and is defined as follows:
(!! *, *
$$$$$$$$$$$$$$$$$$$ % !! $$$$$$$ +,
'+-,
(7)
The test statistic T of the market model can be calculated by dividing the (!! *$$$$$$$$$$$$$$$$$$$ by its , * standard deviation as specified in equation (8):
* (!! *$$$$$$$$$$$$$$$$$$$, *
.(!! *$$$$$$$$$$$$$$$$$$$ , * (8)
4.2.4 Mean-adjusted model
With the mean-adjusted model the abnormal returns ! for stock " on day # are calculated by subtracting the average return (/̂) of stock " during the estimation window from the return of stock " on day # as defined by equation (9):
! /̂ (9)
The average abnormal returns !!$$$$$$ for day # are calculated by summing the returns for period # and dividing by the number of events as shown in equation (10):
!!
$$$$$$ 1 % ! & '
(10)
Summing the !!$$$$$ during the event window results in the cumulative average abnormal returns ((!! $$$$$$ as presented by equation (11): (!!*, * $$$$$$$$$$$$$$$$$ % !!$$$$$$ +, '+-, (11)
The test statistic of the mean-adjusted model is calculated by dividing the (!!*$$$$$$$$$$$$$$$$$ by its standard , * deviation illustrated by equation (12):
* (!!*$$$$$$$$$$$$$$$$$, * .(!!*$$$$$$$$$$$$$$$$$ , *
17
5. Empirical results
This section provides the main empirical results. First, the results for the long-term calendar portfolio based method are discussed. Thereafter, the results of the short-term event study are treated.
5.1 Long-term results
First, daily raw returns are calculated for the insider portfolios between 2006 and May 2013 and transformed into a return graph. The aggregate returns are calculated assuming that €100 is invested in the insider portfolio at the first transaction day and kept until the first of May 2013. Figure 3 shows the aggregate equally weighted portfolio returns for both the insider buy and sale portfolio for several different assumed holding periods The first notable finding is that for all holding periods the buy portfolio yields higher raw returns than the market index, whereas the insider sell portfolio underperforms the market index. At first sight these graphs seem to support the hypothesis that corporate insiders are able to obtain abnormal returns; however note that these raw returns are not risk- adjusted.
Display A in Figure 3 shows the portfolio returns when a 60 day holding period is taken into account. It is clear that the sale portfolio does not perform well compared to the market index. This shows that insiders do well by selling these stocks i.e. they are able to prevent underperformance. However, for the first years the sale portfolio does not deviate much from the market index. The largest underperformance seems to arise after the year 2008 and 2009. The buy portfolio also closely follows the return on the market index in the first years; nevertheless, it is able to deliver higher returns starting from the year 2010. When increasing the holding period the results for the buy portfolio improve. It can be seen that the returns on the buy portfolio are the highest under a 125 day holding period. In this case, on average, shares bought by insiders would have doubled in value. Again most of the outperformance is obtained after the big market declines in 2008 and 2009. After these years when markets recovered, the value of the insider buy portfolio and the market portfolio diverge, where the insider buy portfolios rises faster than the market index.
18
Figure 3: Raw equally weighted insider trading portfolio returns
This figure shows the raw portfolio results for several holding periods. The portfolios are equally weighted at the day of construction. The results are calculated under the assumption of a €100 investment in the insider trading portfolios at day 1. Holding periods are denoted in trading days.
A.) 60 Day holding period B.) 125 Day holding period
19
Next, risk adjustments are made and the statistical significant of the returns are assessed. Table 4 on the next page shows the main regression results for several different holding periods. In general the regressions show that insiders buy (sell) stocks in their own firm, which are typically small cap and value firms, as the SMB and HML factors are significant in most regressions. These results are in line with the findings of Jeng et al. (2003) who find that insiders purchase much more shares in small firms, value firms and firms that recently underperformed. In addition, as can be found in Appendix 1, most insider transactions in the AFM database indeed arise from relatively small companies.
Panel A in Table 4 shows that for a ten day holding period the buy portfolio exhibits a negative alpha, although not statistically significant. The sale portfolio exhibits an annualized2 negative alpha of -16.0% which is significant at the 5% level. This negative Fama-French alpha means that the portfolio is earning a lower return than its exposures would otherwise yield. Or, put differently, higher returns can be achieved with an alternative portfolio that mimics the characteristics of this insider sale portfolio. This means investment in the insider sale portfolio is suboptimal and insiders made a favorable decision by selling these shares. However, the results should be interpreted with caution, as the majority of the sale transactions occurred prior to the market declines in 2008 and 2009. When the holding periods are increased to 60 and 125 trading days the sale portfolio still exhibits negative daily alpha, but this is no longer statistically significant.
For the buy portfolio, the highest alpha (13.0%) is obtained under a 125 day holding period, which is in line with the figure on the previous page. However, the abnormal performance for this holding period is not significant. For 250 and 500 holding days, the alpha decreases to respectively 9.25% and 6.50% but both are significant. These results are comparable, but lower, than findings of Biesta et al. (2003) who finds abnormal returns for Dutch insider purchases of around 9.00% for a six month period. However, Seyhun (1998) argues that two-thirds of insider outperformance is realized in the first six months. Moreover, the study of Biesta et al. (2003) includes a different time period than this study. Jeng et al. (2003) found annual risk adjusted outperformance for insider purchases of 6.0% in the U.S. market.
For robustness the calendar portfolios are also constructed based on transaction value (value weighting). With this method small trades no longer receive the same weight in the portfolio as large trades. The weight on a particular stock i is then determined by equation (13).
∑ 11
, (13)
2
20
where is the weight in the portfolio on stock i determined by the value of the trade 1 divided by the sum of all transactions, ∑ 1 ), made at time t.
The results in Table 4, panel B show that weighing based on value does not have a major impact on the regression results. The results do not deviate much from the results of the equally weighted regressions (see Appendix D for value weighted return graphs). However, with the value weighted approach, the buy portfolio does yield a significant alpha of 11.50% for a 125 day holding period. Furthermore, the sale portfolio still underperforms, but the alpha is insignificant in all cases. As the results for these weighing methods do not show much difference, I stick for convenience reasons to the equally weighted method for further testing.
Table 4: Equally weighted and value weighted OLS regression results for insider portfolios
This table presents the results for the OLS Fama-French regressions for both insider buy and sale portfolios. Portfolios are constructed for each unique insider trading day and overlap in time. The holding period for these portfolios is displayed in trading days and varies from 10 to 500. The equally weighted and value weighted returns of these portfolios are calculated at each trading day between the start of 2006 and the end of 2012 and used as dependent variable. The explanatory variables are the Dutch market index (beta), represented by the MSCI Netherlands and the small minus big (SMB) and high minus low (HML) risk premiums. The alpha is denoted as annualised, assuming 250 trading days a year. Statistical significance is denoted by * for 10%, ** for 5% and *** for 1%.
Buy portfolio Sale portfolio
Holding period Alpha Beta SMB HML Alpha Beta SMB HML Panel A. Equally weighted portfolio construction
10 -0.75% 0.59*** 0.54*** 0.01 -16.00%** 0.49*** 0.23*** -0.11*** 60 7.00% 0.80*** 0.55*** -0.03 -10.50% -0.85*** 0.47*** -0.18*** 125 13.00% 0.63*** 0.44*** -0.02 -9.75% 0.75*** 0.46*** -0.01*** 250 9.25%** 0.73*** 0.55*** -0.07*** -2.25% 0.88*** 0.52*** -0.16*** 500 6.50%* 0.77*** 0.52*** -0.11*** 1.25% 0.92*** 0.58*** -0.12*** Panel B. Value weighted portfolio construction
10 -1.00% 0.59*** 0.54*** 0.01 -15.50% 0.49*** 0.23*** -0.11*** 60 7.25% 0.79*** 0.54*** -0.03 -10.00% 0.85*** 0.47*** -0.18*** 125 11.50%* 0.76*** 0.55*** -0.00 -8.00% 0.94*** 0.49*** -0.20*** 250 9.50%** 0.72*** 0.55*** -0.07*** -2.05% 0.88*** 0.53*** -0.16*** 500 6.43%* 0.76*** -0.11*** 0.52*** 1.00% 0.91*** 0.58*** -0.12*** 5.1.1 Robustness testing
Exclusion of market breakdown
21
the holding periods a significant alpha is found on the sale portfolio. This leads to the conclusion that shares sold by insiders do not significantly underperform the market in this time period. The results for the buy portfolio however improve. For the 125 and 250 trading days the alpha is highly significant at the 1% level. For the 500 trading day holding period the alpha is still significant but at the 5% level. The 125 day holding period leads to an annual alpha of 28.25%, which is much higher compared to the alpha of 13.00% found previously for the whole time period. No clear explanation can be given for the large differences. It is possible that insiders make a more accurate estimation of the true value of their company and bought shares when they felt their company was undervalued during the market panic.
Exclusion of dominant firm in database
It is important to mention that the results so far can be driven by one specific firm. In Appendix B a list of firms and total number of transactions is included. One particular firm, Boussard & Gavaudan Holding, seems to dominate the database with a total of 341 insider transactions. This firm accounts for about 18% of the whole sample. Panel B in Table 5 presents the regression results when this firm is removed from the sample. The results prove to be robust in this setting, although the significance level of the alphas reduces to 10% for holding periods of 250 and 500 trading days. Also the 125 day holding period is no longer found to be significant. The magnitude of the alpha coefficient does not seem to change a lot compared to the previous results.
Table 5: Robustness tests for equally weighted insider portfolios
This table shows the results for the OLS Fama-French regressions for both insider buy and sale portfolios. Portfolios are constructed for each unique insider trading day between 2006 and the end of 2012 and overlap in time. The holding period for these portfolios is displayed in trading days and varies from 10 to 500. The equally weighted returns of these portfolios is calculated at each trading day between the start of 2006 and the end of 2012 and used as dependent variable. The explanatory variables are the Dutch market index (beta), represented by the MSCI Netherlands and the small minus big (SMB) and high minus low (HML) risk premium. In panel A. the years 2006, 2007 and 2008 are excluded to correct for the large market drop in these years. Panel B excludes the firm Boussard & Gavaudan which dominates the sample. The alpha is denoted as annualised, assuming 250 trading days a year. Statistical significance is denoted by * for 10%, ** for 5% and *** for 1%.
Buy portfolio Sale portfolio
Holding period Alpha Beta SMB HML Alpha Beta SMB HML Panel A. Sample corrected for bear market. The years 2006, 2007 and 2008 are excluded
10 -2.50% 0.52*** 0.50*** 0.08 -15.00% 0.36*** 0.19*** -0.07** 60 20.00% 0.71*** 0.41*** 0.06 -5.00% 0.64*** 0.35*** -0.16*** 125 28.25%*** 0.67*** 0.41*** 0.08** 0.00% 0.64*** 0.32*** -0.15*** 250 25.00%*** 0.63*** 0.39*** 0.08** 5.00% 0.63*** 0.32*** -0.14*** 500 21.25%** 0.64*** 0.39*** 0.08** 2.50% 0.64*** 0.35*** -0.13*** Panel B. Sample corrected for dominant firm
22 5.1.2 Regression results for firm size and trade size
On the next page, Panel A and B in Table 6 provide the regression results when the insider trading sample is split into large and small cap firms. As previously described, information asymmetry between insiders and outsiders is hypothesised to be larger for small cap firms, leading to higher risk adjusted returns. The results show that indeed the alpha coefficient on the buy portfolio for small cap firms is higher and almost double the size compared to large cap firms, although both are insignificant. With respect to the insider sale portfolio, the small cap sample yields negative alpha for all holding periods, whereas the large cap sample only does for 10, 60 and 125 holding days. It is remarkable that for the small cap sample the alpha is significant at the 5% level for holding periods of 125 and 250 trading days. The annual underperformance is around -25.00% which is much larger than the underperformance found for the total sample and the large cap sample. The results seem to support the hypothesis that insiders in small cap firms are indeed able to obtain higher abnormal returns (or prevent higher underperformance) compared to large cap firms. However, this only seems to hold for insider sale transactions.
23
Table 6: Equally weighted insider portfolios based on firm size and trade size
This table presents the OLS Fama-French regression results for both insider buy and sale portfolios. Portfolios are constructed for each unique insider trading day and overlap in time. The holding period for these portfolios is displayed in trading days and varies from 10 to 500. The equally weighted returns of these portfolios are calculated at each trading day between the start of 2006 and the end of 2012 and used as dependent variable. The explanatory variables are the Dutch market index (beta), represented by the MSCI Netherlands and the small minus big (SMB) and high minus low (HML) risk premiums. Firms are sorted in descending order based on market capitalisation. Panel A include only trades from the highest third percentile and serves as high cap sample. Panel B includes only the lowest third percentile. When trades are sorted based on transaction value, panel C includes the highest quarter of the distribution and serves as high value sample and panel D includes the lowest quarter and serves as low value sample. The alpha is denoted as annualised, assuming 250 trading days a year. Statistical significance is denoted by * for 10%, ** for 5% and *** for 1%.
Buy portfolio Sale portfolio
Holding period Alpha Beta SMB HML Alpha Beta SMB HML Panel A. Sample split for large cap firms only
10 1.60% 0.31*** 0.26*** -0.08*** -7.13% 0.30*** 0.17*** -0.11*** 60 2.75% 0.76*** 0.52*** -0.09*** -9.33% 1.01*** 0.37*** -0.34*** 125 1.88% 1.13*** 0.64*** -0.12*** -9.55% 1.22*** 0.51*** -0.33*** 250 5.80% 1.12*** 0.68*** -0.15*** 1.21% 1.25*** 0.55*** -0.25*** 500 6.48% 1.20*** 0.67*** -0.16*** 3.35% 1.27*** 0.64*** -0.24*** Panel B. Sample split for small cap firms only
10 4.48% 0.15*** 0.29*** -0.11 -5.00% 0.34*** 0.22*** -0.11 60 6.65% 0.72*** 0.52*** -0.29*** -7.45% 0.39*** 0.27*** -0.15*** 125 13.25% 0.61*** 0.36*** -0.11* -25.43%** 0.58*** 0.35*** -0.23*** 250 16.13% 0.57*** 0.39*** -0.01 -24.40%** 0.69*** 0.47*** -0.16*** 500 13.95% 0.55*** 0.38*** -0.02 -10.20% 0.73*** 0.52*** -0.05 Panel C. Sample split for large trades only
10 11.88% 070*** 0.55*** 0.08 -14.18% 0.91*** 0.55*** -0.15** 60 7.70% 0.70*** 0.47*** 0.12 -3.33% 1.01*** 0.66*** -0.22*** 125 5.80% 0.75*** 0.49*** -0.04*** -3.03% 1.21*** 0.64*** -0.25*** 250 6.43% 0.72*** 0.45*** -0.14*** 0.30% 1.10*** 0.51*** -0.21*** 500 4.15% 0.77*** 0.44*** -0.15*** 2.25% 1.03*** 0.61*** -0.16*** Panel D. Sample split for small trades only
24 5.2 Event study results
First the cumulative abnormal returns are plotted for the insider buy and sale transactions separately and included in Appendix F . The graph does not look promising, as no clear change is visible for both trade types around the event day. Panel A in Table 7 confirms this, as for all event windows the abnormal returns are insignificant. Both the T and Z-values are very low. However the CARs show an intuitive sign change around the event day. For sale transactions the CARs become negative, whereas the CARs for buy transactions turn positive. Unfortunately, the results are not convincing, and lead to the conclusion that insider trades, in general, are not able to affect stock prices.
Cross sectional analysis
Existing literature indicated abnormal returns after inside trades to be related to firm size or size and the trade (Betzer and Theissen, 2008; Bajo and Petracci, 2006; Degryse, et al., 2009). The hypothesis is that larger trades and trades in small cap firms will lead to higher CARs. Large trades show more confidence by the insider, and for small cap firms the information asymmetry between insiders and outsiders is larger. Panel B in Table 7 shows the CARs for both large cap and small cap firms. To classify firms by size I use a method comparable to the method used by Jeng et al. (2003) which is described in the methodology section of this paper. Although the T and Z-values are higher compared to the total sample, the first thing to notice is that the CARs are not significant for all event windows. Furthermore the results are not quite as one would expect. Sale transactions in large firms for example lead to higher price reaction than for small caps, with mean-adjusted CARs of -2.03% and 0.59% respectively using an event window from day 0 to 10. In addition, sale transactions do not have the expected negative sign for small cap firms. With respect to buy transactions the results do not show much difference between small and large cap firms. The CARs are positive, but again too small to be statistically significant. Therefore, it is concluded that insider transactions in small cap firms do not lead to a (higher) impact on stock prices.
25
sample (transactions of €5,000 and up) where on each event day buy and sale transaction in one specific firm are netted. The buy sample consists of N=326 events and the sale sample consists of N=152 events. Panel B includes the results for the top and bottom third of the firms when sorted based on market capitalisation. The large cap sample has N=69 for buys and N=49 for sales. The small cap sample has N=95 for buys and N=44 for sales. Panel C provides the results when, based on value of the trade the highest 25% is taken into account with N= 69 for buys and N=52. Statistical significance is denoted by * for 10%, ** for 5% and *** for 1%.
Event windows
CAR-10, +1 T (z) CAR-5, +1 T (z) CAR-5, +5 T (z) CAR-3, +3 T (z) CAR-1, +1 T (z) CAR-1, +5 T (z) CAR0, +5 T (z) CAR0, +10 T (z) CAR+6, +10 T (z)
Panel A. Total sample
Net purchases Mean adj. -0.57% -0.08 -0.34% -0.07 -0.19% -0.03 -0.09% -0.01 0.13% 0.03 0.08% 0.01 0.15% 0.03 0.27% 0.04 0.12% 0.03
(-0.02) (-0.02) (-0.02) (0.03) (0.02) (0.02) (0.02) (0.03) (0.02)
Market adj. -0.54% -0.08 -0.32% -0.07 -0.09% -0.01 -0.08% -0.01 0.08% 0.02 0.13% 0.02 0.23% 0.04 0.41% 0.07 0.18% 0.04
(-0.02) (-0.01) (-0.01) (0.01) (0.01) (0.01) (0.02) (0.05) (0.06)
Net sales Mean adj. 1.06% 0.19 0.69% 0.17 0.55% 0.09 0.50% 0.10 0.17% 0.05 -0.02% 0.00 -0.14% -0.03 -0.65% -0.10 -0.51% -0.12
(0.24) (0.21) (0.11) (0.13) (0.06) (-0.01) (-0.04) (-0.13) (-0.16)
Market adj. 0.88% 0.14 0.44% 0.09 0.11% 0.01 0.19% 0.03 0.03% 0.01 -0.27% -0.05 -0.33% -0.06 -0.74% -0.11 -0.41% -0.10
(0.18) (0.12) (0.02) (0.04) (0.01) (-0.06) (-0.08) (-0.14) (-0.12)
Panel B. Sample split for large and small cap firms Large caps
Net purchases Mean adj. 1.49% 0.06 1.63% 0.42 1.78% 0.31 2.39% 0.44 0.48% 0.13 0.87% 0.17 0.15% 0.04 0.51% 0.11 0.37% 0.12
(0.08) (0.53) (0.39) (0.55) (0.16) (0.21) (0.04) (0.14) (0.15)
Market adj. 1.55% 0.25 1.63% 0.43 1.81% 0.03 2.00% 0.37 0.32% 0.09 0.66% 0.13 0.14% 0.03 0.56% 0.12 0.46% 0.15
(0.29) (0.50) (0.04) (0.50) (0.12) (0.18) (0.04) (0.16) (0.20)
Net sales Mean adj. 0.86% 0.14 0.80% 0.19 0.38% 0.06 0.87% 0.17 0.23% 0.07 -0.42% -0.07 -0.41% -0.08 -2.03% -0.27 -1.61% -0.33
(0.20) (0.27) (0.08) (0.24) (0.12) (-0.10) (-0.11) (-0.38) (-0.47)
Market adj. 0.18% 0.03 -0.01% 0.00 -1.08% -0.13 -0.34% -0.05 -0.26% -0.07 -1.22% -0.17 -1.08% -0.16 -2.32% -0.29 -1.24% -0.27
Small caps (0.04) (0.00) (-0.20) (-0.10) (-0.11) (-0.27) (-0.26) (-0.48) (-0.42)
Net purchases Mean adj. -1.55% -0.17 -0.99% -0.17 -0.69% -0.08 -1.63% -0.20 0.35% 0.06 0.25% 0.03 0.30% 0.04 0.62% 0.08 0.32% 0.05
(-0.21) (-0.22) (-0.10) (-0.25) (0.08) (0.04) (0.06) (0.10) (0.07)
Market adj. -0.76% -0.10 -0.88% -0.17 -0.02% 0.00 -1.18% -0.14 0.49% 0.08 0.78% 0.11 0.86% 0.12 1.16% 0.16 0.30% 0.05
(-0.18) (-0.31) (-0.01) (-0.20) (0.15) (0.20) (0.22) (0.31) (0.09)
Net sales Mean adj. 1.16% 0.20 0.37% 0.08 0.49% 0.07 0.00% 0.00 0.08% 0.02 0.32% 0.06 0.12% 0.02 0.59% 0.09 0.47% 0.12
(0.42) (0.17) (0.16) (0.00) (0.04) (0.13) (0.05) (0.20) (0.38)
Market adj. -5.16% -0.89 -3.79% -0.89 -3.79% -0.42 -1.41% -0.29 0.23% 0.06 0.49% 0.10 1.27% 0.27 0.50% 0.09 -0.83% -0.23
(-1.08) (-1.08) (-0.43) (-0.36) (0.07) (0.12) (0.32) (0.10) (-0.28)
Panel C. Sample split for large trades
Net purchases Mean adj. -0.18% -0.03 0.15% 0.04 -0.64% -0.13 0.07% 0.02 0.04% 0.01 -0.61% -0.13 -0.79% -0.20 -0.41% -0.08 0.37% 0.10
(-0.03) (0.05) (-0.14) (0.02) (0.01) (-0.17) (-0.29) (-0.09) (0.14)
Market adj. -0.15% -0.02 0.16% 0.04 -0.65% -0.13 0.02% 0.01 0.03% 0.01 -0.67% -0.15 -0.84% -0.22 -0.44% -0.10 0.40% 0.12
(-0.04) (0.11) (-0.36) (0.01) (0.01) (-0.24) (-0.35) (-0.20) (0.19)
Net sales Mean adj. 1.90% 0.36 1.54% 0.40 0.40% 0.06 0.42% 0.08 -0.77% -0.24 -1.29% -0.22 -1.14% -0.24 -1.41% -0.28 -0.27% -0.06
(0.84) (0.91) (0.12) (0.19) (-0.55) (-0.52) (-0.57) (-0.56) (-0.09)
Market adj. 2.32% 0.46 1.84% 0.51 0.66% 0.10 0.53% 0.11 -0.68% -0.21 -1.25% -0.22 -1.17% -0.25 -1.28% -0.23 -0.11% -0.02
26
Dutch insiders are allowed to trade in their company’s stock if certain conditions are met. The AFM is responsible for monitoring these insider trades and records the trades in a publicly available register. Using this data, I investigated whether insiders in the Dutch market are able to obtain abnormal returns. Second, I examined if outsiders notice the transactions by insiders and if this results in a significant effect on stock prices.
First, I find that Dutch insider stock transactions mainly arise in small cap firms and high book to market value firms. This supports findings by Jeng et al. (2003), who conclude that insiders in the U.S. seem to invest mainly in small firms and value firms. Second, after making the appropriate risk adjustments, I find that generally the returns from the insider buy portfolios exhibit a significant alpha of around 6.5% to 11.5% annually. This is only found when the calendar portfolios are kept for a relatively long holding period (125 to 500 trading days). This is similar to results of Biesta et al. (2003) who find abnormal performance for insider purchases of around 9.00% in six months and Aktas et al. (2007) who find cumulative abnormal returns of 7.45% during a 60 day event window. However, Aktas et al. (2007) did not make adjustments for additional factors of risk. The results indicate that corporate insiders buy their shares with long-term information and/or are able to make a better estimate of the true fundamental value of the company.
Third, I find the insider sale portfolio to exhibit insignificant underperformance, although this seems to decline when the holding period for the portfolios increases. This contradicts the findings of Seyhun (1998) for the U.S. market and Aktas et al. (2007) who do find significant underperformance in the long run for insider sales. However, Biesta et al. (2003) employ a comparable methodology to the one used in this paper and did not find Dutch insider sales to significantly underperform the market. They argue that unlike insider purchases, insider sales arise from different motives such as liquidity or diversification needs, which explains the differences.
Furthermore, return differences between insider portfolios and trade characteristics are analysed. Contemporary literature suggest insider returns to be negatively related to firm size (Seyhun 1998). I find indeed that purchases in small cap firms, assuming a 250 trading day holding period, generate annual alpha of 16.13%, while the large cap purchases yield alpha of only 5.80%. However, the alphas are no longer significant for both samples. For insider sales however, I find the sales in small firms to significantly underperform around -25.00% annually, whereas insider sales in large firms do not.
27
contradictory to findings of Seyhun (1998), Fidrmuc et al. (2006) and Betzer and Theissen (2008) who suggest an opposite relation. However, the study of Aktas et al. (2007) also focuses on the Dutch market and came to similar findings, where smaller trades result in larger abnormal returns. They argue that an explanation for this difference is found in the stealth trading hypothesis of Kraakman (1991). Insiders prefer several smaller transactions above one large transaction so that the market and market supervisors do not notice these trades.
With respect to the short-term effect of insider trading on stock prices, I find the cumulative abnormal returns to be positive for purchases and negative for insider sales. However, the effects are too small to be significant. Biesta et al. (2003) and Aktas et al. (2007) came to similar conclusions and also conclude that inside trades do not affect stock prices in the short run. Moreover, differentiating between firm size and trade size does not lead to different results. Even for larger trades or trades in small cap firms the results are insignificant. The results are unambiguous and lead to the conclusion that insider transactions do not impact stock prices in the Netherlands.
Limitations
A few limitations of this study have to mentioned. First, the AFM data is insufficient to determine the actual holding period for transactions by insiders. As this period is unknown, the returns calculated in this study are based on an assumed holding period. Second, it is assumed that the data does not contain illegal trading activity. If the data does contain some illegal trades, this would likely result in higher returns, which leads to an upward bias of the results in this study. A third limitation is the focus on small firms. Although this limitation arises from the data, it could be that insiders in large firm trade financial products other than shares which underrepresents them in this study. Furthermore, the sample splits performed in this study result in relatively small sub-samples, which could lead to lower power of the statistical tests. Finally, the AFM data does not provide the exact date on which the insider transaction is made public. This date can be anywhere between the five days after the actual transaction. As a result noise can enter the event windows which will lead to higher T-values for the statistical tests.
Suggestions for future research
28
Appendix A: Filters applied to AFM database
This table shows the filters applied to the AFM insider transaction database to arrive at the final sample of 1889 insider transactions.
Initial sample 11,038
Deletion of
Non-share transaction i.e. derivatives 7,100
Non-Euronext transactions 361
Investment funds 1,217
Incomplete disclosures 371
Outliers 'data winsorising' cutting of 5% 99
Final sample 1,889
Appendix B: Composition of sample after filter applications Firms included in sample (N=1889) and total number of trades per firm (buys plus sells).
# Firm Total number of trades As % of total sample
1 Aalberts industries N.V. 19 1.01%
2 Aegon N.V. 3 0.16%
3 Akzo Nobel N.V. 7 0.37%
4 AMG N.V. 1 0.05%
5 AND Publishers 8 0.42%
6 Atrium Real Estate 1 0.05%
7 Batenburg Beheer N.V. 18 0.95%
8 BE Semiconductor 14 0.74%
9 Beter Bed Holding N.V. 15 0.79%
10 Boussard & Gavaudan Holding 341 18.05%
11 CTAC N.V. 8 0.42% 12 Corio N.V. 1 0.05% 13 CSM N.V. 8 0.42% 14 D.E. Masterblenders 4 0.21% 15 Delta Lloyd N.V. 12 0.64% 16 Dockwise 29 1.54% 17 DPA Group N.V. 6 0.32% 18 Exact Holding N.V. 3 0.16% 19 Fugro N.V. 7 0.37% 20 Gemalto N.V. 8 0.42% 21 HAL Trust 2 0.11% 22 Hunter Douglas N.V. 135 7.15% 23 Hydratec Industries 6 0.32% 24 Kendrion N.V. 10 0.53% 25 BAM Groep N.V. 12 0.64% 26 Philips N.V. 12 0.64% 27 Ten Cate N.V. 113 5.98% 28 Vopak N.V. 28 1.48% 29 Wessanen N.V. 12 0.64%
30 Macintosh Retail Group 96 5.08%
31 Neways Electronics N.V. 52 2.75%
32 Nieuwe Steen Investments N.V. 3 0.16%
29
Appendix C: Calculation of Z-statistic according to the Cowan rank test
This appendix shows the calculation of the Z-statistic for the non-parametric Cowan (1992) rank test.With the Cowan rank test abnormal returns calculated on both the estimation and event window and ranked in ascending order. The Z-statistic is calculated according to the equation below:
3 4/5 ! 67$$$$ 8 9∑+'+:;-,<;! 6$$$ 8 5 ∑ ! 6$$$ 8 5 +, '+-, != (14)
In the formula above ! 6$$$ is the average rank of the abnormal returns across the number of stocks on day #. Where ! 67$$$$ represents the average rank of the abnormal returns across the number of stocks and 4 days of the event window. The MAX represents the maximum possible rank. The U represents the mean rank.
34 OctoPlus N.V. 5 0.26% 35 Ordina N.V. 9 0.48% 36 Qurius N.V. 8 0.42% 37 Randstad Holding N.V. 162 8.58% 38 SBM Offshore N.V. 135 7.15% 39 SimacTechniek N.V. 6 0.32%
40 Sligro Food Group N.V. 65 3.44%
41 Stern Group N.V. 62 3.28% 42 TIE Holding N.V. 40 2.12% 43 Unilever N.V. 55 2.91% 44 Unit4 N.V. 50 2.65% 45 USG People N.V. 4 0.21% 46 Vastned Retail N.V. 1 0.05%
47 Volta Finance Limited 108 5.72%
30
Appendix D: Raw value weighted insider trading portfolio returns
The figure below shows the raw portfolio results for several holding periods. The portfolios weights are based on transaction value. The results are calculated under the assumption of a €100 investment in the insider trading portfolios at day 1.Holding periods are denoted in trading days
A.) 60 Day holding period B.) 125 Day holding period
31
Appendix E: Bera-Jarque test for normal distribution of mean-adjusted abnormal returns
The T-test to assess the statistical significance of the abnormal returns is based on the assumption that returns approximately follow a normal distribution i.e. they have skewness and excess kurtosis of around zero. The graphs below show that this assumption is not satisfied for both the buy and sale sample.
A.) Normality test results event study buy sample
B.) Normality test results event study sale sample
32
Appendix F: Cumulative abnormal returns for insider buy and sale transactions.
The CARs are plotted for the total buy sample (N=326) and sale sample (N=152) using the mean-adjusted method. Insider transactions with a value of €5,000 and up are taken in account.
0.00% 0.20% 0.40% 0.60% 0.80% 1.00% 1.20% 1.40% -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
CAR Sale transactions
-0.70% -0.60% -0.50% -0.40% -0.30% -0.20% -0.10% 0.00% 0.10% 0.20% -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
33 References
Aktas, N., De Bodt, E., Riachi I., De Smedt, J., 2007. Legal insider trading and stock market reaction: Evidence from the Netherlands. Unpublished working paper. Louvain School of Management, Louvain
Ausubel, L., 1990. Insider trading in a rational expectations economy. American Economic Review 80, 1022-1041.
Bauer, R., Koedijk, K., Otten, R., 2005. International evidence on ethical mutual fund performance and investment style. Journal of Banking and Finance 29, 1751-1767
Bera, K., Jarque, M., 1980. Efficient test for normality, heteroskedasticity, and serial independence of regression residuals. Economic Letters 6, 255-259.
Betzer, A., Theissen, E., 2008. Insider trading and corporate governance: The case study of Germany. European Financial Management 15, 402-429.
Biesta, A., Doeswijk, R.Q., Donker, H.A., 2003. The profitability of insider trades in the Dutch stock market. Unpublished working paper. Erasmus University, Rotterdam
Brooks, C., 2008. Introductory econometrics for finance. Cambridge University Press, Cambridge.
Brown, S., Warner, J., 1985. Using daily stock returns: The case of event studies. Journal of Financial Economics 14, 3-31.
Carlton, D., Fischel D., 1983. The regulation of insider trading. Stanford Law Review 35, 857-895.
Corrado, C., 1989. A nonparametric test for abnormal security-price performance in event studies. Journal of Financial Economics 23, 385-395.
Cowan, R., 1992. Nonparametric event study tests. Review of Quantitative Finance and Accounting 2, 343-358.
34
Degryse, H., De Jong, F., Lefebvre, J., 2009. An empirical analysis of legal insider trading in the Netherlands. Unpublished working paper. Tilburg University, Tilburg
Fama, E., 1970. Efficient capital markets: A review of theory and empirical work. Journal of Finance 25, 383-417.
Fama, E., French, K., 1992. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 3-56.
Fidrmuc, S., Goergen, M., Renneboog, L., 2006. Insider trading, news releases and ownership concentration. Journal of Finance 61, 2931-2974.
Huddart, S., Shi, C., 2007. Jeopardy, non-public information, and insider trading around SEC 10-K and 10-Q filings. Journal of Accounting and Economics 43, 3-36.
Jaffe, J., 1974. Special information and insider trading. Journal of Business 47, 411-428.
Jeng, L., Metrick, A., Zeckhauser, R., 2003. The profits to insider trading: A performance-evaluation perspective. Review of Economics and Statistics 85, 453-471.
Jensen, M., 1969. Risk, the pricing of capital assets, and the evaluation of investment portfolios. Journal of Business 42, 167-247.
Kothari, S., Warner, J., 1997. Measuring long-horizon security price performance. Journal of Financial Economics 43, 301-339.
Kraakman, R., 1991. The legal theory of insider trading regulation in the United States. In: Klaus J., (Ed.), European Insider Dealing. Butterworths, London.
Lyon, J., Barber, B., Tsai, C., 1999. Improved methods of tests of long-horizon abnormal stock returns. Journal of Finance 54, 165-201.
MacKinlay, C., 1997. Event studies in economics and finance. Journal of Economic Literature 35, 13-39.
35
Meulbroek, K., 1992. An empirical analysis of illegal insider trading. Journal of Finance 47, 1661-1699.
Seyhun, H., 1998. Investor Intelligence from Insider Trading. MIT Press, Cambridge.
