Insider trading patterns and signaling power of insider purchases before and during the 2007 – 2009 financial crisis : an event study

(1)

University of Amsterdam

Amsterdam Business School

MSc Business Economics – Finance

Master Thesis

Insider Trading Patterns and Signaling Power of Insider

Purchases before and during the 2007 – 2009 Financial Crisis –

An Event Study

Author: Tsvetelina Alexandrova, 11084219 Supervisor: Dr. Jens K. Martin July 2016, Amsterdam

(2)

Statement of Originality

This document is written by Student Tsvetelina Alexandrova who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

Insider trading records provide manifold examples of CEOs buying large amount of shares after serious companies’ stock price declines during the recent 2007 – 2009 financial crisis. Therefore, the current research focuses on tracking changes in patterns in the transaction size of insider purchases and sales at an individual – daily level as well as on the signaling content and market reaction to large insider purchases, both examined for the periods before and during the recent financial crisis. The analysis reveals an increased fraction of insider purchases relative to sales and total insider trading activity during the crisis period and a tendency for companies’ insiders to engage more in block purchases during the crisis than before. Furthermore, event study results imply that large block insider purchases are connected with sharper stock price decline preceding trade’s reporting and induce a bigger temporary market response, credible up to 10 days after the announcement of the trade, which is statistically higher from the signaling power of smaller insider purchases irrespective of the trading period. These findings support the assumption that insiders willing to show commitment and support the share price of their own company by submitting large block buys can make use of the signal their purchases send to the market. As the analysis shows that insider purchases can not only be regarded as a positive signal in terms of company’s undervaluation, but also may bring noisy signal to the market, it implies a decrease in the degree of superiority of the signaling power of insider purchases over sales. However, since the signaling effect cannot last in the long term, the thesis concludes, that submitting large insider purchase in order to show commitment and support the share price of a company is a credible in the short-term, but very costly signal.

Keywords: Insider trading, block trading, signaling, information asymmetry, financial crisis JEL classification: G14, G01

(4)

1. Introduction

Theories of information asymmetry point out differences in the information content and signaling power of insider purchases and sales: while selling might have very different reasons such as diversification and liquidity needs, because of the costs involved insider purchases are regarded to contain a more precise signal. Furthermore, according to theories of strategic trading, informed investors tend to break up large trades into smaller pieces and spread them over time in order to conceal private information and to reduce the permanent impact of their trades on the stock price and market liquidity (argument goes back to Kyle (1985)).

However, in contrast to the predictions of these theories, company’s CEO of the troubled German bank Hypo Real Estate Holding AG - Georg Funke - buys 20 000 shares in total value of over 440 000 Euro as a large block trade on 15th January 2008 right after company’s shares price collapses. Furthermore, after peaking at 53 Dollar per share in 2007, JPMorgan Chase’ share price falls to less than 20 Dollar by the beginning of 2009 and the company’s CEO Jamie Dimon responds to this by purchasing 500 000 shares. At the beginning of the current year – 2016, Jamie Dimon spends over 26 Million Dollar of his own money to buy again 500 000 shares after a significant 20% decline in JP Morgan’s share price. In the after-hours trading following the announcement of the insider purchase, JPMorgan's share price experiences a 2% jump.

Cases like these raise the questions whether insiders change their trading strategies in times of distress and more importantly, whether large insider purchases can always be regarded as a positive signal in terms of undervaluation or can managers use their signaling power to show commitment, advertise confidence and support the share price of their company.

The main goal of this thesis is to track changes in patterns of the transaction size of insider purchases and sales before and during the 2007 – 2009 financial crisis and to test whether companies’ insiders can make use of the signal they send to the market through their buys. For this purpose, the study hypotheses that first, when in trouble, insider tend to engage more in larger block buy trades and second, large insider purchases are able to send credible and stronger signal to the marker than smaller ones. In order to verify these hypotheses, the current research uses insider trading data from the Thomson Reuters Insider Filing database as well as company and stock price information gathered from CRSP/Compustat Securities Daily merged file available online at Warthon Research Data Services (WRDS) for the period of the recent 2007 – 2009 financial crisis and three years before. By means of statistical methods and an

(6)

- 2 - event study, the thesis reports results that validate the two hypotheses. Insider purchases and sales, aggregated at an individual - daily level, are found to follow different distribution patterns in terms of their transaction size during the recent 2007 - 2009 financial crisis versus during the pre – crisis period 2003 – 2006. Insiders are revealed to engage more in block purchases during times of distress than during normal trading periods. Furthermore, large insider purchases are evidenced to be associated with sharper stock price decline preceding trade’s reporting and to induce a bigger temporary market response, credible up to 10 days after the announcement of the trade, which is statistically higher from the signaling power of smaller insider purchases irrespective of the trading period. Additionally, the baseline results are supported by two robustness checks with respect to the selected measure for size of order and the applied model for estimation of abnormal stock returns.

Taken together, these findings derive an important conclusion about the signaling power of large insider purchases and the market’s belief. As the thesis takes into account additional possible motives for conducting insider purchases next to the possession of positive private information about company’s future financial prosperity useful for the valuation of firms, it shows that purchases may also bring noisy signal to the market, which casts doubt on the degree of superiority and the signaling power of insider purchases over sales. The results from the analysis should help investors willing to understand the dynamics of insider trading or update beliefs about company’s value and contribute to the existing literature regarding insider trading strategies and patterns, market crashes and information asymmetry.

The remainder of this thesis is structured as follows: Section 2 dives into the theoretical background of insider trading and derives the thesis’ hypotheses from theories and empirical findings of related articles in the existing literature; Section 3 introduces the methodology applied to the empirical analysis; Section 4 describes the data and sample selection process; Section 5 presents the results of the data analysis; Section 6 verifies the main findings by means of two robustness checks and finally, Section 7 discusses potential limitations of the research and concludes.

2. Background, Literature Review and Hypotheses Development

Insider trading refers to transactions made by top managers, directors and large shareholders who own at least 10% of a given company’s shares and are therefore considered to have preferential access to private information about the company in comparison to outside market participants. Manifold researches report positive correlation between insider trading and

(7)

- 3 - subsequent stock price movements, implying that insider trading enhances market efficiency, alleviates information asymmetry and contributes to faster price discovery. (Aktas et. al. (2008) Meulbroek (1992); Cornell and Sirri (1992); Chakravarty and McConnell (1997) etc.). Due to these benefits, the legalization of insider trading is a fact, however all corporate insiders are subject to manifold legal and regulatory constraints. Nowadays U.S. insiders are legally obliged to report all their transactions to the Securities and Exchange Commission (SEC) within two trading days after execution1.

2.1. Patterns of Insider Trading Strategies

According to theories of strategic trading, informed investors competing for the use of the same information typically split large orders into smaller transactions and allocate them over time in order to conceal the private information and reduce the permanent impact of their trades on the stock price and market liquidity (Kyle (1985), Foster and Viswanathan (1994) etc.). In this regard, Lebedeva et.al. (2015) analyze trading strategies of corporate insiders and provide a strong evidence that when facing competition from other insiders, corporate insiders slower their dealings, increasing the length of transaction sequences over time. Besides the tendency to avoid block trading, insiders are also found to adjust to changes in market liquidity, i.e. they prefer trading more on high - liquidity days in order to conceal their information - based trading in higher trading volumes. For the same reason, trade duration is evidenced by the study to rise in markets when liquidity is low. Overall, these findings imply that insiders’ primary concern is the negative impact of their actions on the price and market liquidity. Nonetheless, the theories cited above refer to the competitive trading behavior of multiple informed investors or the behavior of corporate insiders in the aggregate, while the focus on this thesis is on patterns in insider buy and sell transactions at an individual trade level.

In addition, Betzer et. al. (2014) analyze trading and reporting patterns of corporate insiders before and after the implementation of the Sarbanes-Oxley Act (SOX) in the US. The authors assume that the lax regulatory requirements in the pre-SOX era, permitting up to 40 days time period between execution of transaction and reporting, allows insiders to time their trades and take advantage of it. By engagement in “stealth trading”, i.e. splitting up large trades into smaller chunks and trading them on different days, but reporting them jointly at the end, insider

1_{Section 16 of the Securities and Exchange Act of 1934, required insiders in the US to report their dealings to the} SEC within 10 days after the end of the month following execution of the trade. Since 29 August 2002 with the implementation of the Sarbanes-Oxley Act (SOX), reporting deadline is narrowed to 2 business days after the transaction.

(8)

- 4 - traders willing to execute large quantities can reduce the negative price impact of a given transaction on the following trades part of the same trading sequence. The study finds out a clear evidence of insiders’ engagement in stealth trading and detects significant violations in the reporting requirements, especially in the pre-SOX period. Although the researchers do not discover statistically significant difference in the probability of violation of the reporting requirement between purchases and sales, they show that on average, insider purchases are associated with longer reporting delays than insider sales. Surprisingly, the fraction of non-stealth trades is found to be much larger for insider purchases than for insider sales, which is related to the discovery that in general the average dollar trading volume for insider purchases (stealth and non-stealth) is much smaller than for sales and therefore purchases it is less likely for the purchases to be split and thus labeled as stealth trades than sales. Besides that, both the existence of reporting delays and stealth trading are found to be systematically related to firm’s, trader’s and trade’s characteristics and to impede the adjustment of stock prices. However, the cumulative abnormal returns (CARs) experience positive flow regardless of the trading delay. Share prices are also discovered to respond to reporting of insider trades as the reaction is stronger after stealth trades and purchases in comparison to non-stealth trades and sales respectively. All findings are considered to hold also in the post-SOX era, although less evident. Betzer et.al. (2014) is essential for this thesis in terms of databases, sample selection criteria and methodology approach. Nonetheless, the current thesis does not focus on timing of trade reporting and violations of regulatory requirements, but on insider trading strategies regarding execution of big blocks versus smaller transactions.

Similar to the first main scope of this thesis, several studies examine patterns in insider trading strategies. For example, Biggerstaff, Cicero and Wintoki (2015) relates insiders’ trading duration to the duration of their informational advantage. The research finds out that an opportunistic insider will trade in a short window (submitting isolated orders) or spread trades over time (engaging in trading sequences) depending on whether his or her informational advantage is long- or short - lived respectively. Controlling for these patterns, the study evidences that both insider purchases and sales predict abnormal returns, especially when they are reported to the SEC after market closing. The conclusion derived from the study is that insiders tend to deliberately time their dealings and reporting to minimize the market impact of their trades, maximize the duration of their informational advantage and thus maximize profits. Tamersoy et. al. (2013) claim to represent the largest statistical analysis on US data from the whole SEC Form 4 Fillings, tracking patterns in insider trading behavior over time from a

(9)

- 5 - network perspective with respect to insider’s corporate role, industry sector, transaction type and relationships to other insiders. The study accounts for consistently larger number of insider purchases than sales, especially in the period of 2003 – 2008, assumed to be related to a change in a US tax law regarding capital gains. It furthermore reveals higher volatility of the CEOs’ behavior compared to the insiders, shown by the significant number of sales after 2003 and their sharp decline in late 2007. The sector comparison statistics reveal that in terms of the number of transactions, the technology sector is the largest, followed by consumer services, capital goods, healthcare and finance. Insider traders are found to tend to engage in trading sequences as the cases with subsequent transactions of the same type appear more often and with shorter cycle than the opposite cases. Part of the main findings of the study is also the discovery that insiders are able of timing their trades – they “buy low and sell high” compared to market closing prices.

The main distinction of these papers to the current thesis is that Tamersoy et. al. (2013) tracks changes in insider trading patterns over time with respect to the number of transactions, not to transaction volume or size and Biggerstaff, Cicero and Wintoki (2015) connect the decision of trading in blocks or engaging in a trading sequence with the duration of the informational advantage, not considering possible different motives such as commitment and intention to support the share price.

There are only a few studies that relate insider trading behavior to the context of recent traumatic events. Marin and Oliver (2006) examine patterns of insider trading prior to sharp stock moves at an individual stock level and discover a significant asymmetry between sales preceding stock crashes and insider purchases preceding stock price peaks. In particular, the authors evidence that the peak of aggregate sales occurred almost a year before the largest S&P 500 index drop, while the aggregate purchases peaked in the month preceding the jump. Holt (2014) analyses insider trading behavior during 2008 stock market crash and discovers large number of insider buy transactions during the crash, bringing positive abnormal returns to the insiders as well as to the investors mimicking their behavior. Fahlenbrach and Schulz (2011) find that bank CEOs did not reduce their shareholdings during the recent financial crisis in 2007 and 2008 and therefore suffered large wealth losses. Lebedeva et. al. (2015) discover a significant drop in the number of transactions per sequence during the recent financial crisis, where individual transactions are assigned into a transaction sequence when the same insider submits a subsequent individual transaction in the same direction within one week.

(10)

- 6 - Additionally, the study discovers a decrease in the average trade duration of transaction sequence after implementation of the SOX-rule and an increase during the 2008 financial crisis. Cziraki (2015) focuses on changes in insider trading patterns of U.S. bank executives during the financial crisis. The study discovers 20% higher probability for insiders of banks with high risk exposure to housing markets to sell a stock during the crisis period compared to banks with low risk exposure and no significant differences in the insider trading activity between the two groups before US housing market weakened. By including controls for additional characteristics influencing insider trading other than private information, the study manages to relate and explain the differences in insider trading behavior with the private information of the companies’ insiders. Two measures of insider trading are used in Cziraki (2015): insider trading in the aggregate, measured as the ratio of the number of insiders that increased ownership in bank stocks to the number of all insiders in the current year and insider trading as net dollar value of open market trades, i.e. value of purchases minus sales for the whole year.

In contrast, this thesis aims to trace changes in the distribution of the transaction size of individual insider purchases and sales, aggregated at an individual – daily level, using a direct comparison of the periods before and during the 2007-2009 financial crisis, which derives the first hypothesis:

Hypothesis 1: Insiders tend to engage more in block trading during times of distress than versus

during normal trading periods.

2.2. Information Asymmetry, Signaling Content and Price Impact of Insider Trades

Various studies analyze the credible signals that directors’ dealings send to the market. Fidrmuc et. al. (2006) emphasize the asymmetric signaling power of insider purchases and sales. While sales can be due to manifold reasons such as insiders’ confidence of firm’s poor future performance, liquidity needs, diversification or rebalancing of portfolio etc., because of the costs involved insider purchases are found to result only from the insider’s confidence about firm’s future financial prosperity. Consequently, insider purchases are regarded as a leading indicator of firm’s future good performance, while sales are considered to send rather a noisy signal. These conclusions go in line with findings by Lakonishok and Lee (2001) and Tavakoli et. al. (2012), which also analyze the information content of insider trading and market reactions to it. In addition, Chan and Li (2013) focus on understanding the different market price and volume responses to these main two types of insider trades. The authors reach the conclusion that insider purchases are connected to larger price changes and stronger market volume

(11)

- 7 - reactions than sales after trading news are released. Besides that, the study evidences that an improvement of a firm’s information environment is more associated with insider purchases than with sales. Furthermore, due to the superior signal insider buying sends to the market, Chan and Li (2013) find out that investors’ responses to the two types of transactions are also asymmetric – insider buying behavior enhances investors to mimic and buy stock more than insider selling does.

However, not all insider trades are informative about company’s future. Cohen et. al. (2012) distinguish between „routine” insider trading – foreseeable insider transactions with no predictive power towards future stock returns, leading to zero abnormal returns – and “opportunistic” insider trading containing viable, profit carrying information. Examples for routine insider sells are transactions conducted by the same insider at the same time each year, most often driven by diversification and liquidity needs, whereas routine buys are associated with insider receiving a bonus in the form of a discount on the own company stock.

In contrast to the studies focusing on insiders’ buying and selling activities, Gao et.al. (2015) examine the signaling content of “insider silence”, i.e. the information revealed by the lack of insider trading activity in a given company over a period of one year. The study provides evidence that insider silence is systematically connected to litigation risk – one of the main reasons restraining insiders from selling shares, when they have significant negative information about the future of their firm. Due to that, rational insiders are found to tend to remain inactive prior to large stock price drops or shareholder litigations and to sell shares when they do not foresee serious danger. Therefore, the study concludes that the lack of insider trading signals worse news and significantly lower future stock returns compared to the case when insiders sell.

As distinguished from the above cited articles, this thesis reviews the information content of insider trading before and during the recent financial crisis, as the main emphasize is on the content and strength of the signal of insider purchases. Since the records of some significant insider transactions as those conducted by the CEOs of the Hypo Real Estate AG and JP Morgan Chase suggest that, instead of worrying about the stock price information they reveal by trading in a big block, insiders of troubled companies may wish to exploit the signaling effect of their trades for opportunistic reasons by submitting larger buy orders than usual in order to signal commitment and support the share price of their own firm in a time of distress. In particular, this thesis assumes following cases in which insiders of troubled companies may have incentives to buy in large blocks during crisis: 1) buying in order to show good leadership,

(12)

- 8 - confidence and commitment for the future of their own company and 2) buying in order to temporary support the share price and enhance other investors to follow. The first case is supported also by Halverson et. al. (2004), who highlight the positive effect of self-sacrificial leadership behavior on followers’ perception during situational crisis. The experimental study provides examples of CEOs that decrease voluntarily their salaries in times of financial crisis as important component of charismatic and transformational leadership. The second case partly refers to the insiders’ manipulative strategy, discussed in Chakraborty and Yilmaz (2004), who reveal that in some situations insiders deliberately trade in the wrong direction to confuse market participants and make short-term losses but long-term profits. These motives are considered as additional reasons to purchase, possibly bringing additional benefits for to company in situation of distress, without simultaneously violating the assumption that insiders purchase in order to make profits. This can create noise in the signal of insider purchases to the market and thus weakens the degree of superiority of insider purchases over sales.

Furthermore, investigating the effect of large block trades on the temporary and permanent stock prices of New York Stock Exchange companies, Holthausen, Leftwick and Mayers (1990) discover correlation between block size of sales and their temporary price impact, but the permanent price impact of block purchases is found to not distinguish from the one observed in purchasing 100 shares. In contrast, this thesis focuses on the signaling power and market response to large insider purchases, aiming to reveal their role as a possible bullish indicator. This derives the second hypothesis:

Hypothesis 2: Large insider purchases send stronger signal to the market than smaller ones.

3. Research Design

The research objectives of this thesis consist in the examination of specific patterns in the insider trading behavior around the recent financial crisis 2007 – 2009 and the signaling content and market reaction to large insider purchases. These objectives are investigated applying a two - step methodological approach: 1) a statistical analysis on the insider trading behavior followed by 2) an event study of the market impact of block insider purchases, both conducted for the two periods of interest.

3.1. Insider Trading Patterns

First objective of the analysis is tracking changes in patterns in the transaction size of insider purchases and sales during the recent financial crisis versus precrisis period on an individual

(13)

-- 9 -- daily level. As a suitable indicator for size of the order is selected the absolute dollar amount of the individual insider purchases and sales, measured by the multiplication of the number of shares traded by a given insider on a given date with the respective price per share. Since several transactions can be made by the same insider on the same day, all individual trades which are in the same direction need to be aggregated in order to receive a total daily individual transaction volume. This requires the sum of all shares purchased or sold by a given insider of a given company to be multiplied by the average price per share of his daily purchases or sells. This can be expressed with the following formula:

𝑇𝑟𝑎𝑛𝑠. 𝑉𝑜𝑙𝑢𝑚𝑒_𝑗𝑡= ∑ 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑠ℎ𝑎𝑟𝑒𝑠 𝑝𝑢𝑟𝑐ℎ𝑎𝑠𝑒𝑑 _𝑗𝑡 𝐽 𝑗=1 ∗∑ 𝑃𝑟𝑖𝑐𝑒 𝑝𝑒𝑟 𝑠ℎ𝑎𝑟𝑒 𝑗𝑡 𝐽 𝑗=1 𝑛

Where 𝑗 is the company insider, 𝑡 is the trading day and 𝑛 is the number of the insider’s individual buy or sell transactions for this day.

Thus, as in the case of Georg Funke, who bought on 15.01.2008 subsequently 13 849, 4 202 and 1 949 shares of Hypo Real Estate Holding AG at 22.08 Euro, 22.09 Euro and 22.10 Euro respectively, applying the formula, the absolute volume of the total daily transaction is calculated to approximately equal 441 800 Euro, the amount that was also reported to the BaFin2_.

𝑇𝑟𝑎𝑛𝑠𝑎𝑐𝑡𝑖𝑜𝑛 𝑉𝑜𝑙𝑢𝑚𝑒_𝑖𝑡 = (13 849 + 4 202 + 1 949) ∗(22.08 + 22.09 + 22.10) 3

= 20 000 𝑠ℎ𝑎𝑟𝑒𝑠 ∗ 22.09 𝐸𝑢𝑟𝑜 𝑠ℎ𝑎𝑟𝑒⁄ ≈ 441 800 𝐸𝑢𝑟𝑜

For the rare cases in which the same insider is trading on the same day in both opposite directions, transactions are counted separately. If the transactions are carried out by the same insider, on the same day and in the same direction, but reported on different days, they are also not aggregated.

The test of changes in the insider trading patterns of insider purchases and sales on an individual - daily level consists in the comparison of the mean, median and especially the values indicated by the upper quartiles and the 95th percentiles of the trading volume in both periods of interests. This is shown by the descriptive statistics as well as by means of boxplots depicting the

(14)

- 10 - distribution of the transaction volume for insider purchases and sales before and during the recent financial crisis for all industries in the sample.

The descriptive statistics as well as the outliers depicted by the boxplots should derive implications about insider’s tendency to split orders into smaller chunks and engage in longer transaction sequences or their increased engagement in block trading. I expect the latter to be more pronounced for insider purchases during the crisis period and especially in the financial and technology sectors since they are typically more volatile. Furthermore, if an insider is willing to exploit the signaling effect of his purchases in order to signal commitment and to support the share price of his own company, he should have incentives to report his block purchase almost immediately after execution. Therefore, I separate all insider purchases and sales into quartiles depending on their transaction volume and calculate the average days between reporting and transaction date for each quartile within both trading groups during the crisis and during the period before that, by the following formula:

𝐷𝑎𝑦𝑠 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 = 𝑅𝑒𝑝𝑜𝑟𝑡𝑖𝑛𝑔 𝑑𝑎𝑡𝑒 − 𝑇𝑟𝑎𝑛𝑠𝑎𝑐𝑡𝑖𝑜𝑛 𝑑𝑎𝑡𝑒

I predict a tendency for the larger purchases to be reported quicker than those in the lower quartiles and the sales.

3.2. Event Study Methodology

The second part of this thesis focuses on insider purchases only. It analyzes the market before and after insider purchases are reported to the SEC and aims to check whether the information signal of large insider purchases is credible. For this purpose, I apply a standard event study approach with daily data and the market model to report the cumulative abnormal returns (CARs) over time for different event windows by means of the functions in the Eventus software available in Warton Research Data Services (WRDS).

First, in order to avoid double - counting of observations, all insider purchases of the same company, that are reported on the same day are aggregated by their dollar amount. Thus, the transaction volume for company 𝑖 at reporting day 𝜏 can be expressed by the following formula as the sum of the dollar amount of the individual insider trades reported by company’s insiders on that day:

𝑇𝑟𝑎𝑛𝑠𝑎𝑐𝑡𝑖𝑜𝑛 𝑉𝑜𝑙𝑢𝑚𝑒𝑖𝜏= ∑ 𝑇𝑟𝑎𝑛𝑠𝑎𝑐𝑡𝑖𝑜𝑛 𝑉𝑜𝑙𝑢𝑚𝑒𝑗𝜏 𝐽

(15)

- 11 - This is similar to the aggregation method presented above in Section 3.1. but now the purchases in the sample are aggregated on a firm - daily level. Afterwards the insider purchases are classified by size and divided in groups depending on their transaction volume – Small and Large using the lower and upper quartiles of the distribution as cutoff points as well as the period in which they were executed - crisis versus pre - crisis.

In the next step, this Thomson Reuters insider trading data is merged with data on daily returns and company characteristics taken from CRSP/Compustat merged file on Securities daily, available in WRDS by their specific company identifier CUSIP code and the announcement day. The announcement day, i.e. the day on which a given insider purchase is filed with the SEC, is chosen as an event day. The number of observations is the upper and lower quartile subsamples is not equal because after the merge some observations are automatically dropped due to missing CUSIP, filing date outside of the given period interval or too few estimation period date. Furthermore, as done in Betzer et.al. (2014), the market model is selected over a 255-day estimation window ending 46 days prior to the event. This represents the estimation period, i.e. the period far away from the event in which the relevant normal performance model parameters are estimated without being impacted by the event itself. Furthermore, I also select the CRSP value - weighted index, Patell Z significancy test, available at Eventus, as well as different event windows ranging between -30 and +30 days before and after the event to capture the market reactions to the SEC Fillings in short and longer-term perspective. The timeline below illustrates and summarizes the inputs:

The program calculates the abnormal returns for every company for every event in the sample using the market model as it retains only those firm-event observations for which the number of valid returns in the estimation window is sufficiently high. The abnormal return is defined as the difference between the actual ex - post return on a security and the expected return without conditioning on the event taking place. Thus, for company i and event date 𝜏, the abnormal return is:

𝐴𝑅_𝑖𝜏 = 𝑅_𝑖𝜏− 𝐸[𝑅_𝑖𝜏|𝑋_𝜏]

t = -30

t = -301 t = -46

t=0, Event date

t = +30

(16)

- 12 - where 𝐴𝑅_𝑖𝜏, 𝑅_𝑖𝜏 and 𝐸[𝑅_𝑖𝜏|𝑋_𝜏] are the abnormal, actual and expected returns respectively for day 𝜏. The market model assumes a stable linear relationship between the market return and the firm’s return:

𝑅𝑖𝜏 = 𝛼𝑖 + 𝛽𝑖𝑅𝑚𝜏+ 𝜀𝑖𝜏 where 𝐸[𝜀𝑖𝜏 = 0] and 𝑉𝑎𝑟[𝜀𝑖𝜏] = 𝜎𝜀𝑖𝜏 2

where 𝑅𝑖𝑡 and 𝑅𝑚𝑡 are the security and market returns respectively, 𝜀𝑖𝜏 is the zero mean error

term and 𝛼𝑖, 𝛽𝑖 and 𝜎𝜀𝑖𝜏

2 _{are the parameters of the market model. Substituting 𝑋}

𝜏 with the market

return in the market model, the sample AR for the stocks in the event window is computed: 𝐴𝑅_𝑖𝜏 = 𝑅_𝑖𝜏− 𝛼_𝑖− 𝛽_𝑖𝑅_𝑚𝜏

Under the Null hypothesis, the event is assumed not to have an impact on the behavior of the stock returns (mean and variance) in the event window:

𝐴𝑅𝑖𝜏 ~ 𝑁(0, 𝑉𝑎𝑟(𝐴𝑅𝑖𝜏))

which should be rejected. The abnormal returns are aggregated by the software across securities to report the mean abnormal returns across the 𝑖 = 1, … , 𝐼 cases for each day of interest:

𝐴𝐴𝑅_𝜏 = ∑ 𝐴𝑅_𝑖𝜏

𝐼

𝑖=1

In the next step, I aggregate the average abnormal returns across securities over time in order to get the sample cumulative abnormal returns (CAR) for every event window (𝜏₁, 𝜏₂) of interest:

𝐶𝐴𝑅_𝑖(𝜏_1,_𝜏₂₎ = ∑ 𝐴𝑅_𝑖𝜏

𝜏2

𝜏1

Finally, CARs are averaged for the whole sample, resulting in the cumulative average abnormal returns: 𝐶𝐴𝐴𝑅_𝑖(𝜏_1,_𝜏₂₎= ∑ 𝐴𝐴𝑅_𝑖𝜏 𝜏2 𝜏1 = 1 𝐼∑ 𝐶𝐴𝑅𝑖(𝜏1,𝜏2) 𝐼 𝑖

which are tested for statistical significance. The CARs and CAARs are reported over six different event windows, namely (-30, -1), (-3,-1), (-2,0), (0,+2), (+1,+1) and (+1, 30) to capture

(17)

- 13 - the market reactions to the SEC Fillings in short and longer - term perspective separately for the crisis and the pre - crisis period.

Then the results are compared between the two groups – Small (insider purchases in the 1st

quartile according to their transaction volume distribution) and Large (insider purchases in the 4th quartile according to their transaction volume distribution) as well as between the two periods in order to derive a conclusion about the signaling content and power of insider purchases before and during the recent financial crisis. This is done by means of a simple t-test for equality means. The difference is regarded as significant at 95% confidence level, if the corresponding p-value is smaller than 0,05.

The essential part of the analysis is to see whether share price reaction is stronger after larger insider block purchases than after smaller ones and what happens with the market before and after the announcement, i.e. whether the market underreacts and block purchasing is a useless signal. I predict higher than expected returns after the disclosure of both - small and large insider purchases. However, I expect the larger purchases to be associated with larger stock price decline prior to the event and to have temporary higher stock price impact after trading news are released compared to the purchases in the smaller subgroup. I furthermore expect these patterns to be more pronounced during the crisis period.

4. Data and Sample Selection

The current section provides a description of the databases and of the sample selection process for the empirical analysis.

The analysis described in Section 3 requires collecting data on daily insider transactions, stock returns and firm characteristics. The latter two are extracted from the Center for Research in Security Prices (CRSP)/Compustat merged file. Insider trading data is obtained from Thomson Reuters Table One File, which captures all U.S. insider non-derivative transactions and holdings information filed on Forms 3,4 and 5 as reported to the Securities and Exchange Commission (SEC). The database contains information on company’s name, CUSIP identification code, transaction code, transaction amount, transaction date, transaction price, announcement date, insider’s name and role code, resulting number of shares held, industry code, cleanse indicator, showing whether errors in the reported data were corrected by means of external sources etc. The data is drawn for the period of January 2003 until December 2009, which captures the time after the implementation of the changed regulatory SOX-rule, which is evidenced to have induced significant changes in insider trading behavior, and at the same time

(18)

- 14 - creates almost equal length of observation periods before and during the recent financial crisis. Because the negative effects of the financial crisis spread also on other industries in the economy and are still visible months after its peak in the third quarter of 2008, when the biggest investment bank Lehman Brothers fails, as crisis period is defined the trading period from January 2007 until end of 2009. The initial insider trading dataset includes 4 590 742 observations for 12 966 firms and 131 070 insiders.

In order to avoid controlling for other regulatory differences between countries that may have impact on the insider trading patterns, I keep only companies located in the US, subject to same regulatory requirements during the period of interest. Furthermore, I select only the two highest levels of insider role codes in Thomson Reuters3, because they comprise the insiders who are typically considered to have more information about the situation of the company. Besides that, they are usually more connected to the company, hold higher stakes and thus, stand to lose more is a stock price declines. This creates higher incentives to try to show commitment during bad times, especially by the CEOs. The analysis is based only on all open market or private insider purchases or sales, whose value can be verified with high degree of confidence, i.e. excluding all options and other derivatives transactions and selecting cleanse indicator R or H. As done in Lakonishok and Lee (2001) and other papers, I furthermore exclude transactions with less than 100 shares to focus on more meaningful events as well as transactions with missing number of shares, execution price, transaction date, reporting date to the SEC, sector field or insider’s position in the company. This results in a sample of 1 553 098 individual insider transactions, conducted by 60 681 insiders from 7 497 companies.

After aggregating the transactions induced by the same insider in the same day in one, if they are of the same type as explained in the first part of the previous section, the variables shares traded, transaction price, transaction volume and days between transaction and reporting days are winsorized at 1% level. Finally, I get a sample consisting of 415 787 observations on individual - daily level and logically, the same number of firms and insiders as before. Regarding the event study sample, after this aggregation of the trades at a firm – daily level as explained in Section 3.2., the sample contains 280 310 firm – daily observations, only 68 389 of which are purchases. 35 295 are conducted in the pre – crisis period and the rest 33 091 purchases – during the financial crisis.

(19)

- 15 -

5. Results

The current section reports and discusses the results from testing both hypotheses derived in section 2, after applying the statistical and empirical approaches, explained in Section 3.

5.1. Insider Trading Patterns

First objective of the analysis is tracking changes in insider trading patterns before and during the 2007 - 2009 financial crisis by using the descriptive statistics as well as by comparing the distributions of the absolute dollar amount of the individual - daily insider purchases and sales in the two periods of interest.

Table 1 provides summary statistics on insider trading activity at an individual - daily level for 60 981 insiders from 7 497 firms in the sample during the selected time period 01.01.2003 – 31.12.2009 and reports them separately for insider purchases, sales and total sample. Insider purchases include all open market or private purchases executed by a given insider on a given day, and insider sales – all open market or private sales transactions conducted by a given insider on a given day. Shares represent the total number of shares exchanged in insider buy or sell transactions initiated by a given insider on a given day. Price refers to the average price per share exchanged in an insider transaction initiated by a given insider in a given day. Volume is the absolute dollar amount of the individual - daily transactions, created by multiplying the number of traded shares by a given insider on a given day with the respective execution price. Days to report represent the number of days between execution and reporting day of the insider transactions in the sample aggregated at an individual – daily level. All variables of interest are winsorized at 1% level, except days to report, which is winsorized at 5% level. N refers to the number of observations for each subcategory. Panel A summarizes the data for the whole sample time span from 2003 until 2009, whereas Panel B and Panel C respectively provide information for the time periods before and during the 2007 - 2009 financial crisis separately.

[Insert Table 1 here]

The statistics in Table 1 reveal that in both crisis and pre - crisis period the number of insider purchases is a lot smaller than the number of sales. During the pre - crisis period the individual – daily insider purchases represent approximately 21% of the number of sales and 18% of the number total insider transactions at that period. This pattern changes during the crisis period, when the fraction of insider purchases to sales increases to ca. 43% and represents ca. 30% of the total transactions. Furthermore, the comparison of the total number of observations in both

(20)

- 16 - time spans reveal a decline in the insider trading activity during the period 2007 - 2009. While the number of insider purchases remains almost the same during the financial crisis as during the period before, the number of sales drop by more than a half in the crisis period.

Regarding the distribution of number of shares exchanges in individual daily transactions, Panel B and C report differences in patterns of purchases and sales. The descriptive statistics of the number of shares purchased in individual daily transaction show higher mean, median and standard deviations during the crisis period than prior to it. Especially visible are the differences of the values of the 75th and the 95th percentile. In contrast, number of shares sold in individual daily transactions reveal very similar patterns in both time periods with slightly higher standard deviation during the financial crisis. It can be also derived from the table that during the financial crisis the transaction prices for insider purchases and sales appear to differ from the transaction prices of their pre-crisis counterparts – buy prices in Panel C appear to be lower than the buy prices in Panel B and sell prices in Panel C appear to be higher than the sell prices in Panel C regarding all examined density parameters, except minimum and maximum value. Despite the lower purchase prices during the crisis, the statistics of the absolute dollar transaction amount - Volume - of the insider purchases and sales in the sample follow similar patterns to those of the number of shares bought or sold, respectively. During the crisis the absolute dollar amount of insider purchases show higher mean, higher standard deviation, lower values in all density parameters until the second quartile, but higher 75th, significantly higher 95th_{and 99}th_{percentiles than in the earlier examination period, which reveals a tendency of}

companies’ insiders to engage more in block purchases during the financial crisis. Regarding the absolute dollar amount of insider sales during the crisis the table reports higher mean, higher median, higher standard deviation, slightly higher 75th and 95th_{percentiles and no difference}

in the 99th percentile than in the period prior to the crisis. In terms of total transaction volume, it can be derived from the table that the volume prior to the crisis appear to be bigger in all statistical criteria except the standard deviation.

The variable days between transaction and reporting days does not account for significant difference between the two periods of interest. The mean and median values close to 2 days indicate that on average, companies’ insiders obey the changed SOX-rule obligating them to report insider transactions up to two business days after execution. It is furthermore notable, that on average, it takes more time for the information of an insider purchase to get to the market than the information of an insider sale, regardless of the examined period.

(21)

- 17 - Graphic 1 visualizes the distribution of the absolute dollar amount of the examined individual – daily insider transactions (expressed in 10.000$) separately for insider purchases and sales and for both periods of interest – pre - crisis 01.01.2003 – 31.12.2006 and financial crisis period 01.01.2007 – 31.12.2009. The box plots reveal significantly larger transaction volumes and much larger outliers for insider sales than for insider purchases during both of the reviewed periods. Regarding the distribution of insider purchases, it can be derived from the graphic that during the financial crisis more observations are concentrated in the upper tail. This again indicates a tendency of the companies’ insiders to engage more in block insider purchases during the crisis period.

[Insert Figure 1 here]

Figure 2 plots the distribution of the absolute dollar amount of the examined individual – daily insider transactions (expressed in 10.000$) separately for insider purchases and sales and for each sector in which the companies in the sample are operating during both periods of interest – pre - crisis 01.01.2003 – 31.12.2006 and financial crisis period 01.01.2007 – 31.12.2009. The industry sector codes are according to the Thomson Reuters Table One File classification and refer to finance, healthcare, consumer non-durables, consumer services, consumer durables, energy, transportation, technology, basic industries, capital goods, public utilities and miscellaneous. The results show significant higher transaction volumes for sales than for purchases for all industries. According to the picture, the highest volatility industries appear to be consumer durables, the technology sector, finance and consumer services as they account for the largest outliers for both transaction types in both periods of interest. In these industries are observed also the largest insider purchases, reaching absolute amounts of ca. 40 million dollars. In almost all industries the transaction volume of insider purchases reaches higher values during the crisis period than in the period before.

[Insert Figure 2 here]

Table 2 reports the average days between transaction and reporting day for all 92 225 insider purchases and 323 565 insider sales in the sample, aggregated at an individual – daily level and separated in quartiles on basis of their transaction volume distribution for both periods of interest – pre - crisis 01.01.2003 – 31.12.2006 and financial crisis period 01.01.2007 – 31.12.2009. The statistics provide an important finding that the average days between execution of the transaction and reporting to the SEC tend to decline with the increase of the transaction size in both time periods. Interestingly, insider purchases in the fourth quartile account for the

(22)

- 18 - smallest difference between reporting and transaction date, while regarding insider sales smallest values are observed for the transactions in the third quartile. Overall, it can be inferred that insiders executing large transactions tend to report them also quicker.

Taken together, the findings in this subsection are in support of Hypothesis 1 by revealing different distribution patterns in terms of the transaction size of insider purchases and sales aggregated at an individual – daily level and examined before and during the recent 2007 - 2009 financial crisis.

5.2. Event Study Analysis

The second part of this thesis aims to check whether the buy signal of large insider purchases is credible, i.e. whether insiders indeed can make use of the buying signal they send to the market in order to show commitment and “support” the share price of their company. Investigating the market reaction around announcement of large insider purchases requires conducting a standard event study with different event windows to capture the short and long – term effect by applying the methodology discussed in Section 3.2.

Table 3 presents the market model abnormal and cumulative abnormal stock returns around public announcements of insider purchases, divided in small and large based on the 1st and the 4th quartile of their transaction volume respectively for the period of the financial crisis 01.01.2007 - 31.12.2009 along with the associated Patell Z-test and the corresponding P-values. As mean abnormal return (AR) is defined the sample average abnormal return for the specified day in reviewed time interval and the cumulative abnormal return (CAR) is the sample average cumulative abnormal return for the period (-30,+30) relative to the announcement date. Numbers that are significantly different from zero at the 0,1% level are boldfaced.

For the subsample of 8095 large insider purchases, i.e. insider purchases in 4th quartile according to their transaction volume during the financial crisis, all abnormal returns in the interval (-30; -3) are negatively and statistically significant different from zero at 1% level. Interestingly, the average daily abnormal returns turn positive two days before the reporting day. This may be connected to the discrepancy between transaction and reporting day, i.e. the fact that companies’ insiders report their trades to the SEC on average two days after execution. This distortion of prices between trading and reporting day is also evidenced in Betzer

(23)

- 19 - et.al.(2014) as well as other studies on insider trading. However, it is noteworthy that the abnormal returns peak at 1,07% on the first day after the announcement and the vast majority of the daily average abnormal returns in this subsample of insider purchases in the 4th_quartile

remains positive and statistical significant different from zero approximately until day 10 after the event.

The second subsample of 7060 small insider purchases, i.e. purchases in the 1st quartile according to their trading volume, also depicted in Table 3, show a slightly different pattern during the crisis period. Although almost all abnormal returns in the time interval (-30,-2) days before the event are also negative, only a small fraction is statistically significant different from zero at the 0,1% level, while the rest of the abnormal returns are either significant at a higher level or insignificant. The abnormal returns turn positive one day prior to the announcement, peak on the announcement day at 0,44% and remain positive and statistically significant different from zero at 0,1% level approximately until day 5 after the announcement. The lower percentage value on the peak day as well as the fewer number of days after the event on which results remain statistically significant can be an indicator that the signal smaller insider purchases send to the market induces smaller temporary stock price change than larger ones during the 2007 - 2009 financial crisis.

While Table 3 provides a good sense of the statistical significance for the return patterns of small and large insider purchases during the recent financial crisis, Figures 3 and 4 visualize the examined trends and gives more insights into the economic significance of the results.

[Insert Figures 3 and 4 here]

Figure 3 plots the market model sample average abnormal returns for the insider purchases, separated in 1st and 4th quartile according to their transaction volume distribution during the financial crisis for the period of 30 trading days before through 30 trading days after the transaction is reported to the SEC. As it can be seen from the illustration, insider purchases in the 4th quartile appear to have much bigger effect on the share price on the first day and shortly after the announcement of the trade in comparison to the insider purchases in the 1st quartile. However, in the long term the differences in the abnormal return between the two groups almost vanish. Graphic 4 plots the market model sample average cumulative abnormal returns for the insider purchases, separated in 1st and the 4th quartile based on their transaction volume during the 2007 – 2009 financial crisis from 30 trading days before through 30 trading days after the transaction is reported to the SEC. The chart depicts an almost constant decline in the average

(24)

- 20 - stock price of the subsample of small insider purchases starting at least 30 trading days before the announcement day and reaching their lowest point at -3,33% two days before the event. From day -2 on they continue rising with decreasing speed and turn positive on day 17. Regarding the large insider purchases subgroup, the cumulative abnormal returns account for a much sharper and more rapid stock price drop in the time span (-20,-2) prior to the event. Immediately after an insider purchase of the 4th quartile is reported to the SEC, the stock prices tend to increase, first very quickly and then more gradually.

Finally, Table 4 summarizes these trends for the cumulative abnormal of both subgroups during the crisis period, reporting the cumulative average abnormal returns (CAARs) over six different event windows and reports the p-values from the t-test for equality of means between the two groups.

According to the table, the cumulative average abnormal returns for the insider purchases in the 4th quartile are -5,62% from day -30 through day -1, 1,07% on day 1 and 2,21% in the window (0,+2), while the cumulative average abnormal returns for insider purchases in the 1st quartile are -4,70% from day -30 through day -1, 0,41% on day 1 and 0,94% in the event window (0,+2). Values of the t-test reveal significantly different results between insider purchases in the 1st and the 4th_{quartile for the event windows (-30,-1), (0,+2) and (+1,+30). I interpret these results as}

follows: Large insider purchases go in line with a sharper stock price decline preceding the trade’s reporting (and the trade itself) and induce a bigger temporary market response in comparison to small insider purchases conducted during the crisis.

The same analysis is repeated for the sample of insider purchases during the pre – crisis period 2003 – 2006 and the results are reported in Table 5, Table 6, Graphic 5 and Graphic 6.

Table 5 shows the market model abnormal and cumulative sample average abnormal stock returns around public announcements of insider purchases from the 1st and the 4th quartile based on their transaction volume for the pre – crisis period 01.01.2003 - 31.12.2006 along with the associated Patell Z-test and P-values for the null hypothesis that the average abnormal returns equal zero, reported from 30 trading days before through 30 trading days after the transaction is reported to the SEC. Like in the financial crisis period, the abnormal returns for the large purchases subgroup in the pre-crisis period also appear to be significantly negative for the whole period from day -30 to day -3 prior to the trading announcement. However, only

(25)

- 21 - small fraction of the numbers is significant at 0,1% level, while the rest is significant at higher level. The table furthermore shows that the sample average abnormal returns for the insider purchases in the upper quartile turn positive at day -2 prior to the event and peak at 0,90% on the first day after the event, remaining positive and statistically significant at 0,1%, 1%, 5% or 10% level approximately until day 7. The abnormal returns for the purchases in the 1st quartile are significant only on few days around the event and exhibit their highest peak on the day on which the trade is reported to the SEC.

Figure 5 visualizes the abnormal return trend between the large and small insider purchases in the pre-crisis period, which appears to look quite similar to the trend experienced in the period of the financial crisis, but with lower absolute numbers on the days of the return peaks for both groups.

[Insert Figures 5 here]

Figure 6 plots the 2-month evolution of the cumulative abnormal returns around announcements of insider purchases, submitted in the period prior to the financial crisis and separated in 1st and 4th quartile based on their transaction volume.

[Insert Figures 6 here]

While the cumulative abnormal returns for the smaller purchases subgroup evolve relative constant close to zero over time, the cumulative abnormal returns for the purchases in the upper quartile experience a relative constant decline, reaching their lowest point on day -3 at -3%, followed by a rapid jump during the first few days of the announcement and a more moderate further evolution after that.

Table 6 summarizes the event study results for the two examined subgroups during the pre - crisis period, reporting the cumulative average abnormal returns (CAARs) over six different event windows, namely, (-30,-1), (-3,-3), (-2,0), (0,+2), (+1,+1) and (+1,+30).

The corresponding values to the event windows (+1,+1), (0,+2) and (+1,+30) are 0,90%, 1,99% and 2,09% for the purchases in the 4th quartile and 0,25%, 0,65% and 1,57% for the purchases in the lower quartile, respectively. Based on a t-test for equality of means, the numbers are found to statistically differ between the two groups at 5% level. These results confirm the previous finding that the share price reaction is stronger after large insider purchases than after smaller ones. The CARs over the 30-day event window preceding the event day (-30,-1) are

(26)

- 22 - negative and statistically significant different from zero at 1% level for both large and small purchases, but the large purchases subgroup show higher negative values.

Summary results in Table 6 regarding the pre-crisis period are consistent with the findings from Table 4 regarding the financial crisis period and both provide clear evidence that market participants attribute higher information content to large insider purchases than to smaller ones. Large insider purchases are preceded by sharper stock price decline and the temporary stock price reaction after larger insider purchases is stronger than after smaller insider purchases, irrespective of crisis or pre - crisis trading period and the length of the event window. The discussed differences between both subgroups are statistically significant based on a simple t-test for equality means.

The findings imply that it is possible for managers to exploit the signaling effect of their buys, as the signal is credible in the short - term. However, since few days after the block trade is reported prices adjust to a long - term equilibrium, achieved also by conducting insider purchases of a smaller volume, it can be concluded that submitting large insider purchase is a temporary credible, but very costly signal.

6. Robustness Checks

In order to test the validity of the results, the thesis applies two robustness checks with respect to the transaction volume and the market model specifications. In the first one, the number of shares bought relative to the total shares held by firm’s insiders is used as an alternative measure for size of order; in the second, the market model is replaced by the market adjusted model is applied using the same sample and following the same event study methodology discussed in Section 3.2.

6.1. Check with Respect to the Size of Order Specification

The main analysis uses the absolute dollar amount of insider transactions, calculated by multiplying the number of shares with the corresponding price as an indicator for size of order. However, although a given trade may appear costly and large in absolute terms, it may be negligible compared to the insider’s total wealth. To address this, I construct a new measure for size of order – the ratio of number of shares bought by a given insider on a given day to the total shares held by the insider on that day. As done before, I aggregate all purchases conducted by insiders of the same firm on a given day and based on their distribution I allocate the resulting

(27)

- 23 - 64 823 firm - daily observations into Small (purchases in the 1st_{quartile) and Large (purchases}

in the 4th_{quartile) for the period of the financial crisis and the three years before.}

Tables 7 and 8 present the robustness check results for the insider purchases during the 2007 – 2009 financial crisis period, and Tables 9 and 10 - for the pre-crisis period.

[Insert Tables 7,8, 9 and 10 here]

The evolution of the abnormal returns and cumulative abnormal returns for both subsamples of small and large insider purchases during both examined time periods are illustrated on Figures 7 and 8 respectively.

[Insert Figures 7 and 8 here]

The graphically depicted trends show similar patterns to those from the main analysis. The most important results are summarized in Tables 8 and 10 which report the cumulative average abnormal returns for small and large insider purchases in different event windows during the crisis and in the period before, respectively. In both periods of interest, the t-test reveals statistically significant differences in the means of the cumulative average abnormal returns of insider purchases in the 1st and 4th quartile for the event windows of 30 days before and 30 days after the event, implying that large insider purchases go hand in hand with sharper stock price decline prior to the event day and affect the market price of the issuing company’s securities more than smaller insider purchases. In the short – term window of three days before the announcement, results for the two subgroups do not significantly differ which may be explained by the fact that the stock prices are shortly distorted in the time span between transaction and reporting day. In the short – term window of up to three days after the event, the abnormal returns and cumulative abnormal returns of insider purchases from the 4th_{quartile experience a}

significantly higher jump than those of the purchases in the 1st_{quartile irrespective of the}

trading period.

These findings are consistent with the main conclusions regarding the signaling strength of insider purchases derived from the main analysis in Section 5.2.

6.2. Check with Respect to the Market Model Specification

The main analysis calculates the abnormal returns by applying the market model presented in Section 3.2. Although the market model is widely accepted in the event study literature and regarded as a standard method for conducting event studies, the model has also some limitations. For example, it assumes that the risk-free interest rate included in the α factor is

(28)

- 24 - constant, which contradicts to the presumption that market returns can vary over time. To address this issue, I repeat the main analysis with the market adjusted model as a model variation of CAPM and the market model, derived from the market model by replacing 𝛼 with zero and β with 1:

𝐴𝑅_𝑖𝜏 = 𝑅_𝑖𝜏− 𝐸[𝑅_𝑖𝜏|𝑋_𝜏] = 𝑅_𝑖𝜏− 𝑅_𝑚𝜏

Table 11 provides summary of the event study results for insider purchases allocated in the 1st and the 4th quartile based on their transaction volume during both periods of interest – financial crisis 2007 - 2009 and pre - crisis period 2003 – 2006 for the same event windows as in the baseline analysis: (-30,-1), (-3,-1), (-2,0), (0,+2), (+1,+1) and (+1,+30) relative to the public announcement of the trade.

Last column of the table – mean difference t-test - presents the most important findings as it tests for statistically significant difference between the CAARs of the insider purchases of both size for each event window during the financial crisis and the three years before. The model variation does not seem to account for any difference to the baseline results.

7. Discussion and Conclusion

The theory relates insider buying solely to the possession of positive private information about the future financial prosperity of a company and predicts that informed investors such as companies’ insiders will rather spread trades over time in order to conceal this private information and to reduce the permanent impact of their dealing on the stock price and market liquidity. However, insider trading records provide manifold examples of CEOs buying large amount of shares after serious companies’ stock price declines during the recent 2007 – 2009 financial crisis. Therefore, the current research concentrates on tracking changes in patterns in the transaction size of insider purchases and sales at an individual – daily level as well as on the signaling content and market reaction to large insider purchases, both examined for the periods before and during the recent financial crisis. The methodology is grounded on a statistical and a standard event study approach comprising data from two databases on insider trading filings and companies’ stock price information.

In summary, the results indicate that individual insider purchases and sales follow different distribution patterns in terms of their transaction size during the recent 2007 – 2009 financial crisis versus the pre – crisis period 2003 – 2006. In particular, the analysis recognizes an