GM1460 Master Degree Project in Accounting and
Financial Management
Leveling the Playing Field: Can Private Investors Benefit From
Mimicking Insider Trades?
A quantitative study of insider trading
Marcus Mattsson
Marcus Lyckhage
Supervisor: Taylan Mavruk
Graduate School, Accounting and Financial Management, Spring 2023
Acknowledgements
We would like to express our gratitude to our supervisor, Taylan Mavruk, whose guidance
and support has made this thesis possible. We truly appreciate the time he has taken to help us
tackle any problems encountered along the way.
We would also like to show our appreciation towards our main opponents, Naile Aliti and
Zhaoyang Wen, whose feedback and thoughts has helped us improve the quality of this
thesis.
Abstract
Historically, investors have been searching for strategies to maximize performance in the
stock market. It has been shown that corporate insiders can earn abnormal returns by trading
in their own companies as they possess superior information and, to a certain degree, market
timing ability. As the EU tries to cut down on market abuse, a new regulation was introduced
mandating insiders to report and publish trades within three business days. While many
studies have investigated the possibility for insiders to earn abnormal returns, far fewer have
explored the ability for outsiders to mimic insiders as an investment strategy. This study
contributes to this rather scarce research area by investigating the possibility of earning
abnormal returns by mimicking insider trades in Sweden, and if the magnitude of the returns
depend on the position of the insider that is mimicked. Through the use of an event study, we
examine insider buy and sell transactions and aim to answer the following: Can outside
investors earn abnormal returns by mimicking insiders in the Swedish stock market and does
the position of the insider affect the magnitude of the returns?
We find that outsiders are able to earn abnormal returns in the short term by mimicking
insiders’ buy and sell transactions. While mimicking an insiders sell transaction is found to
be beneficial for up to five days after the event, mimicking buy transactions is only found to
be profitable during a three day event window. When comparing these results to the abnormal
returns earned by the insiders themselves, we find that they are able to earn abnormal returns
for longer as the reaction to the perceived event does not happen during the date of the
publication but closer to, or during, the date of the transaction. We hypothesize that the
market observes abnormal trading activity during the transaction day and trades are made
based on this information. For both buy and sell transactions, company size is found to have
the greatest effect where investors trading in smaller companies are able to earn greater
abnormal returns. A possible explanation for this relationship is that the information
asymmetry present between smaller companies and outsiders is larger than that between
larger companies and outsiders. Additionally, we find that mimicking top executives such as
CEOs and CFOs as well as board members is superior compared to lower level insiders.
While no significant difference is found between top executives and board members,
mimicking CFOs enables outsiders to earn the greatest abnormal returns on average.
Keywords: Insider Trading, Abnormal Return, Information Asymmetry, Outside Investors,
Investment Strategy, Insider Positions
Table of Contents
1. Introduction........................................................................................................................................ 1
1.1 Problem........................................................................................................................................2
1.2 Purpose........................................................................................................................................ 4
2. Theoretical Framework and Hypothesis Development.................................................................. 5
2.1 Theoretical Framework................................................................................................................5
2.1.1 Efficient Market Hypothesis.............................................................................................. 5
2.1.2 Asymmetric Information....................................................................................................5
2.2 Hypothesis Development.............................................................................................................6
2.2.1 The Profitability of Insider Trading................................................................................... 6
2.2.2 Differences In Profitability Between Insider Positions......................................................8
3. Methodology and Data.....................................................................................................................10
3.1 Sample....................................................................................................................................... 10
3.1.1 Data Collection................................................................................................................ 10
3.1.2 Exclusions........................................................................................................................ 11
3.2 Variable Selection...................................................................................................................... 13
3.2.1 Dependent Variable and Study Design.............................................................................13
3.2.2 Independent Variables...................................................................................................... 15
3.2.3 Control Variables..............................................................................................................16
3.3 Managing Outliers..................................................................................................................... 17
3.4 Descriptive Statistics................................................................................................................. 18
3.5 Model Specification...................................................................................................................20
3.6 Robustness.................................................................................................................................21
3.6.1 Transaction Date.............................................................................................................. 21
3.6.2 Market Adjusted Model................................................................................................... 22
3.6.3 Excluding Others..............................................................................................................22
3.6.4 Clustered Standard Errors................................................................................................ 23
4. Results............................................................................................................................................... 24
4.1 The Profitability of Insider Trading...........................................................................................24
4.2 Differences In Profitability Between Insider Positions............................................................. 28
4.3 Summary of Results...................................................................................................................32
4.4 Robustness.................................................................................................................................32
4.4.1 Transaction Date.............................................................................................................. 32
4.4.2 Market Adjusted Model................................................................................................... 33
4.4.3 Excluding Others..............................................................................................................34
4.4.4 Clustered Standard Errors................................................................................................ 35
5. Analysis & Discussion......................................................................................................................36
5.1 The Profitability of Insider Trading...........................................................................................36
5.2 Differences In Profitability Between Insider Positions............................................................ 39
6. Conclusion.........................................................................................................................................41
References
Appendix
List of Tables
Methodology and Data
Table 3.1 Sample Selection.................................................................................................................... 12
Table 3.2 Description of Sample............................................................................................................13
Table 3.3 Descriptive Statistics.............................................................................................................. 19
Results
Table 4.1 Test for the significance of CAR, all insiders........................................................................ 24
Table 4.2 Daily Average Abnormal Return............................................................................................25
Table 4.3 Regression results, including controls....................................................................................27
Table 4.4 Regression results, including roles.........................................................................................29
Table 4.5 Regression results, separated roles.........................................................................................30
Table 4.6 Test for the significance of CAR, separated roles.................................................................. 31
List of Graphs
Results
Graph 4.1 Cumulative Average Abnormal Return - Buy.......................................................................26
Graph 4.2 Cumulative Average Abnormal Return - Sell....................................................................... 26
1. Introduction
Ever since the establishment of the world’s first stock exchange in 1611, investors have been
searching for ways to outperform the market. While some attempts have been successful,
others have left much to be desired. There are many studies in which different investment
strategies and their viability are investigated. According to Jegadeesh and Titman (1993),
investors tend to overreact to information and, thus, stock prices tend to follow. Based on this
notion, they constructed a number of portfolios where stocks that did well in the past were
bought and stocks with a poor historical performance were sold. This is referred to as a
momentum strategy and it generated significant positive returns over a 3- to 12-month
holding period. Another investment strategy is the contrarian strategy. It is based on the
notion that individuals tend to base their future expectations on past data. This can be
exploited by selling stocks with high past growth and high expected future growth and buying
stocks with low past growth and low expected future growth. In a study conducted by
Lakonishok et al. (1994), it is concluded that the low growth stocks outperformed the high
growth stocks.
There are also strategies that are based on the informativeness of insider information, also
known as insider trading. Insider trading occurs when a person discharging managerial
responsibilities, or closely associated persons, perform trades in their associated company
(Finansinspektionen, 2022). Lakonishok and Lee (2001) study the American stock market
between 1975-1995 and find that insiders better predict market movements than simple
contrarian strategies. Further, the ability to predict returns is stronger in smaller firms. The
informativeness seems to stem from insider purchases while sales have no predictive ability.
A more recent study focused on insider trading is conducted by Mazza and Wang (2021) who
find that in the Chinese market, there is a statistically significant positive excess return
stemming from insider trades.
The wide array of investment strategies and constant search for ways to exploit the market
has led to rules and regulations being introduced to the financial market. In 2014, EU
regulation 596/2014 on market abuse was proposed. This was a result of the old directive,
2003/6/EC, being outdated as technological developments resulted in major changes to the
financial landscape. The need for a uniform and strong framework that applies to all EU
member states has grown larger with increasing trade and the number of cross-border
operating firms increasing. This requires full and proper market transparency which can be
compromised if market abuse is exerted. Market abuse is defined as unlawful behavior in
financial markets which can consist of insider dealing, market manipulation and unlawful
disclosure of inside information. Regulation 596/2014 contains many points to prevent such
behavior. Article 19 states that persons with managerial responsibilities as well as persons
that are closely related to them have a responsibility of notifying the issuer, or the emission
1
allowance market participant, as well as the competent authority of every transaction
conducted on their own account relating to shares or other financial instruments. This should
be done as soon as possible, but no later than three business days after the date of the
transaction.
Regulation 596/2014 was implemented in Sweden on July 3 in 2016 and led to some changes.
In the previous regulation, SFS 2000:1087, an insider trade had to be reported within five
business days. Shareholders with a major ownership in the company, which was previously
defined as ten percent, no longer fall under the reporting obligation. This differentiates the
EU member states from the U.S. where a 10 percent stockholder is an insider according to
Rule 10b-5 (SEC, n.d.). Apart from Regulation 596/2014, Sweden has a complimentary law,
SFS 2016:1306, which states that Finansinspektionen is the competent authority tasked with
handling insider trades. Finansinspektionen is the Swedish Financial Supervisory Authority
who authorize, supervise and monitor all companies operating in Swedish financial markets
(Government Offices of Sweden, 2015). In Sweden, it is Nasdaq and NGM (Nordic Growth
Market) that are licensed to act as stock exchanges and these companies are regulated by
Finansinspektionen (Nasdaq, n.d.-a). They are also responsible for maintaining and
publishing the PDMR (persons discharging managerial responsibilities) transactions register
in which all trades that fall under regulation 596/2014 exist. This register is public as stated
by SFS 2018:331.
1.1 Problem
The results found by Lakonishok and Lee (2001) among others have sparked an interest
among investors who believe that mimicking insiders is a viable investment strategy.
Whether insider trading is beneficial or harmful has been debated for many years (Seyhun,
1998). Voices who argue against the phenomenon claim that the information insiders use
ultimately belong to the company and shareholders. By profiting from this information
without the approval of shareholders, the insider violates their fiduciary responsibility. They
further argue that insider trading could lead to incentives to manipulate management in
various decisions in order to maximize trading profits, despite being suboptimal for the
company. Not everyone is convinced that insider trading is harmful and some argue that it
could even be beneficial for long-term shareholders. By allowing insider trading,
shareholders can provide insiders with an efficient compensation contract where both parties
benefit from an increase in stock price. It therefore aligns the interests of both parties.
Managers could also be willing to accept lower wages as they take the benefits of insider
trading into account when evaluating their total compensation (Seyhun, 1998). Regulating
insider trading interferes with the efficient contracting between shareholders and managers
which ultimately leads to a reduced effort and efficiency. According to Seyhun (1998), there
2
is no objective answer to this question as insider trading can have beneficial and deleterious
effects.
Seyhun (1998) investigated whether insider trading predicted future stock returns between
1975 and 1994. He found that insider buying activity signals a greater than average stock
price increase while insider selling signals a less than average stock price increase. On
average, stock prices increased following both purchase and sale transactions. However, the
stocks purchased by insiders outperformed the market by 4.5 percent and the stocks sold by
insiders underperformed the market by 2.7 percent. Insiders therefore earned significant
profits from trading in their own firms. The author also found that the predictive ability of the
trades increase when a transaction is not preceded by conflicting signals. A firm with insider
purchases that were not preceded by sales over the past 12 months appreciated by 30.5
percent on average. It can be concluded that purchases are good news and sales are bad news
for the future stock price performance.
Some actors have tried to utilize the notion presented by Seyhun (1998) and others, namely
that insider trades are informative. By creating equity funds in which the sole strategy is to
invest in companies that are experiencing significant insider trades, they have tried to
outperform the market. While some funds have significantly outperformed the market (Alpha
Wealth Funds, 2022), others have failed to do the same (Catalyst Funds, 2022). This raises
the question whether using this investment strategy is viable. Sha et al. (2020) found that on
average, insider trading leads to a negative trading return on the Chinese market. Mazza and
Wang (2021) concluded the opposite, namely that insiders did in fact generate statistically
significant positive excess returns, despite studying the same market. As the results from
previous studies and ventures are ambiguous, more research in the field is motivated. Adding
to this, there are multiple studies on the American, Asian and European market (De La
Brunière et al., 2020; Mazza & Wang, 2021; Mazza & Ruh, 2022; Eckbo & Smith, 1998;
Ahern, 2017), but no published studies in a Swedish setting. As market settings, legal
requirements and other contexts seem to impact the results, it further motivates the execution
of such a study.
There are many examples of studies that have taken an insider perspective and focused on the
timing ability and performance of corporate insiders (De La Brunière et al., 2020; Mazza &
Wang, 2021; Mazza & Ruh, 2022). As insiders have access to information restricted from the
public, the average investor cannot match this performance. Far fewer studies have examined
the possibility for an outside investor to capitalize on insider trades. This study aims to fill
that gap by putting the outside investors in focus and examining whether the performance of
corporate insiders can be translated into abnormal profits for outsiders by using the
informativeness of insider trades as an investment strategy. This also gives rise to the
question of which insider, and what type of company, to mimic. Seyhun (1998) found that
3
there were major differences in the returns of insiders depending on their role in the company.
This was explained by top executives having superior information. While the top executives
managed to reach a net profit of 5 percent, large shareholders had to settle for 0.7 percent. He
also found that the insider trading profitability is generally higher in smaller companies and
that other key performance indicators, such as Price-to-Earnings, have implications for the
profitability of insider trading.
While numerous studies have concluded that it is possible to make a profit by mimicking
insiders, it has been found that selectivity is important. Factors such as firm size, the position
of the insider and transaction type all play a part in determining the return for the outsider.
The ambiguity in results, lack of studies in a Swedish setting and discrepancies between roles
makes for an opportunity to conduct a study that brings more clarity to the field.
1.2 Purpose
The purpose of this study is to investigate how corporate legal insider trading affects market
prices and the possibility for outsiders to achieve abnormal returns. In doing so, this study
contributes to the scarce research studying how insider trading information can be utilized by
outsiders. In order to fulfill the purpose, the study aims to answer the following questions:
- Can outside investors earn abnormal returns by mimicking insiders in the Swedish
stock market?
- Does the position of the insider affect the magnitude of abnormal returns earned when
mimicking insiders in the Swedish stock market?
4
2. Theoretical Framework and Hypothesis Development
2.1 Theoretical Framework
2.1.1 Efficient Market Hypothesis
The Efficient Market Hypothesis (EMH) suggests that the prices of assets within a market are
effective and reflect all publicly available information. Thus, when new information is
presented, this should be accounted for in the pricing of the given asset. Fama (1970)
suggests three different forms of EMH, which represent different levels of efficiency. There is
weak form efficiency where all the information, past and present, that is relevant is
represented in the price and it is not possible to predict future returns by looking at the past
returns. There is semi-strong efficiency which also suggests that all relevant information is
present in the price, but also past information. In such a market, making abnormal returns on
publicly available information is deemed to be close to impossible. The third and last form is
strong efficiency which suggests that all information, public and non-public, is represented in
the price. Strong form efficiency suggests that one could not make abnormal returns by using
any kind of information, and that things like insider information are obsolete. Generally
speaking, markets such as the ones in London and New York are deemed to be semi-strong
(ACCA, n.d.).
2.1.2 Asymmetric Information
The Market for 'Lemons': Quality Uncertainty and the Market Mechanism is an article written
by George Akerlof in 1970 that presents the concept of asymmetric information. Using the
market for used cars as an example, Akerlof demonstrated how a market can be influenced by
adverse selection and ultimately collapse. Buyers cannot distinguish between a high-quality
car, a peach, and a low-quality car, a lemon. Because of this lack of information, buyers will
only be willing to pay the average price of a peach and a lemon together. As opposed to
buyers, sellers know the true quality of their car. As sellers with peaches will receive less than
the actual value of the car, only sellers with lemons will stay on the market. This in turn leads
to the high-quality cars disappearing from the market which reduces the price a buyer is
willing to pay. This creates a downward spiral which could ultimately lead to the market
collapsing.
Asymmetric information is not unique to used goods markets. In capital markets, some
participants have access to inside information and can as a result make more informed
decisions regarding trades. This can lead to insiders experiencing greater returns on their
investments which is a notion that can be taken advantage of by outsiders. For outsiders, it
can be hard to determine if a company is a “peach” or a “lemon”. By studying insider trades,
5
outsiders can gain insight into the state of the company and reduce the information
asymmetry, thus making a more informed decision.
Asymmetric information does not only exist between insiders and outsiders but also within
organizations. The informational hierarchy hypothesis suggests that the impact on prices
coincides with the level of information possessed by the insider (Betzer & Theissen, 2009).
The suggestion is that insiders in C-suite positions or members of the board of directors
making trades should affect prices the most. Although some researchers has found this to be
the case (Lin & Howe, 1990; Mazza & Ruh, 2022; De la Brunière et al., 2020; Seyhun,
1986), others (Betzer & Theissen, 2009; Fidrmuc et al., 2006) have not been able to prove
that the level of information is consistent with the level of impact.
2.2 Hypothesis Development
2.2.1 The Profitability of Insider Trading
Whether legal insider trading creates opportunities for earning abnormal returns or not has
been studied in different countries, markets and contexts in the last decades. De La Brunière
et al. (2020) studied insider trades of the 120 biggest French large capitalization stocks
between 2010 and 2020. They found that on average, both insiders and outsiders were able to
beat the market and that the short-term performance was significantly better than the
long-term performance. The authors also conducted t-tests comparing the means of the
insiders’ and outsiders’ returns which showed that there was no significant difference
between the groups. This shows that an outsider can outperform the index in the short run by
mimicking an insider's actions.
Another European study was conducted by Eckbo and Smith (1998) who studied the
performance of insider trades on the Oslo stock exchange between 1985 and 1992. Rather
than conducting an event study as many other studies have done, they used a value-weighted
portfolio return approach which they argued was more realistic. Unlike De La Brunière et al.
(2020), they found that it is almost impossible for outsiders to earn any abnormal return by
simply mimicking insiders.
There are also studies on the Asian market with differing results. Sha et al. (2020) studied
illegal insider trading on the Chinese market between 1999 and 2007 by investigating 328
insider trading cases prosecuted by China Securities Regulatory Commission. They found
that on average, insiders do not earn significant abnormal returns. In fact, the average trading
return was -7.80 percent per case. This result is not in line with studies from American
counterparts (Meulbroek, 1992; Ahern, 2017) who found that illegal insider trading improves
price discovery and generates substantial abnormal returns in the American stock market. Sha
6
et al. (2020) also discuss their findings in light of the Efficient Market Hypothesis. Their
results point toward the strong form of EMH characterizing the Chinese market but this is
conflicted by previous studies in the same location. One of these studies is conducted by
Mazza and Wang (2021) who investigate the performance of corporate legal insiders and find
support for insider transactions being profitable and generating statistically significant
positive excess returns. They do, however, conclude that random investors significantly
outperform insiders on the Chinese stock market. When conducting another study on the
Korean market, Mazza and Ruh (2022) found that corporate insiders possess market timing
skills which allows them to outperform random outside investors and highlighted the
importance of information asymmetry. Their study confirmed that there is a level of
significant information asymmetry not only between outsiders and insiders but between
different insiders as well.
A common approach when researching the profitability of insider trading is to divide
purchase and sale transactions. Seyhun (1998) found that, while both purchases and sales
were followed by an increase in stock price, when insiders purchased a stock it outperformed
the market by 4.5 percent and when insiders sold stock it underperformed the market by 2.7
percent. In more recent years, Tavakoli et al. (2012) investigated if different insider roles
possess information that can lead to abnormal returns. They found that the signal generated
from purchase transactions were stronger than sale transactions. Even though both purchases
and sales were able to predict some level of future return, the former proved to be better than
the latter. This follows the suggestion that some sales do not represent evidence that the
current circumstances within the firm are challenging but rather that the insider was in need
of liquidity for some other reasons beyond the firm.
It is suggested by Lynch & Rothchild (2000) that, while purchasing can be an effective
signal, an insider sale is rarely a sign that things are going wrong. The suggestion is that very
rarely does one purchase a stock without belief that it will perform well in the future. On the
other hand, selling a stock could be done for many different reasons. The reasons for selling
stock could be plenty fold, including satisfying debt, paying for tuition or purchasing a new
house. Degryse et al. (2013) researched the information content behind insider purchasing
and selling and the results corresponded with the arguments laid out by Lynch & Rothchild
(2000). Degryse et al. (2013) found that purchase transactions were followed by abnormal
returns, while selling generated no significant results. This indicates that the reason behind
the sale is because of liquidity or diversification purposes for the insider. While some studies
have found that purchase transactions generate greater abnormal returns (Degryse et al.,
2013; Tavakoli et al., 2012), others have found the opposite (Cheuk et al., 2006; Van Geyt et
al., 2013). The varying results motivate separating the purchase and sale transactions in order
to establish what outside investors should focus on.
7
As highlighted by the studies presented above, there is no clear cut consensus regarding the
abnormal returns as a result of insider trading. Different markets, contexts and transaction
types yield different results which raises questions about the Swedish market. Based on this,
the following hypotheses are formulated:
H1a: Outside investors can earn positive abnormal returns by mimicking insider
purchases in the Swedish stock market.
H1b: Outside investors can earn negative abnormal returns by mimicking insider sales
in the Swedish stock market.
2.2.2 Differences In Profitability Between Insider Positions
The profitability of insider trading has been studied in several different countries where the
possibility of abnormal returns under different market conditions was the focus. Some
researchers have gone a step further by extending the focus onto information asymmetry
within the firm as opposed to the asymmetry between outsiders and insiders. Seyhun (1998)
found that there is a hierarchy of knowledge which has decisive effects on the level of insider
trading performance possible. He found that top executives are able to ascertain the best
performance due to their superior level of knowledge in line with the Informational
Asymmetry Hypothesis.
A study performed by Wang et al. (2012) found that the value of insider trades made by
CFOs were greater than those made by CEOs. During a 10 year period, the excess return
generated by CFOs was 5 percent greater than that of CEOs. Wang et al. (2012) concluded
that CFOs are better at including information about future earnings than their managerial
counterparts. The explanation given is that CFOs are better at financial decision making.
Knewtson and Nofsinger (2013) looked into this phenomenon by exploring two explanations.
One, is there a different level of scrutiny associated with being a CEO resulting in less
willingness to exploit the asymmetry of information? Or, two, is there simply a difference in
financial understanding between CFOs and CEOs? Knewtson and Nofsinger’s (2013)
findings differ from the explanation given by Wang et al. (2012). The difference in excess
return is found to be more correlated with the level of scrutiny being a CEO hindering the
freedom to make investing decisions as compared to the CFO. This is found to be the most
fitting explanation, as the portfolio excess return is drastically lower post introduction of
Sarbanes-Oxley in 2002 making the CFO a more observed figure. Even though the excess
returns from CFO portfolios remain the greatest, the difference is not deemed to be
significant.
8
Looking beyond normal circumstances, there are times where the information that is known
by insiders is highly important. During times of crises, insiders have a greater understanding
than outsiders and company specific information is more valuable under these circumstances.
Tavakoli et al. (2014) explored the credit crunch of 2007-2008 and found that persons in
senior roles, mainly directors, were aware of the extensive leverage on the financial
institutions and were able to sell their shares pre market collapse. They were also able to buy
shares when the market was at its lowest. The authors argue that this insight was only
common to some officers and directors, and not the rest of management. While Tavakoli et al.
(2014) investigated the trading performance of different insiders during the credit crunch,
Davis et al. (2017) explored the performance of different insiders during looming cases of
class action litigation as well as settlement announcements. They find that these
announcements have a significant impact on the share price of the firm in question, and that
some insiders are able to take advantage of this and earn abnormal returns. Comparing
managers and non-managers, the aforementioned are able to sell prior to the announcement of
a lawsuit and make informed decisions prior to the announcement of settlements. The
importance of managerial positions for abnormal returns are substantiated by other studies as
well (Tavakoli et al., 2012; De La Brunière et al., 2020; Massa & Ruh, 2022).
As highlighted by the studies presented above, the role and seniority of the insider seems to
impact if abnormal returns can be generated and if so, the magnitude of these. Insiders are
able to take advantage of the position that they hold during normal circumstances and even
more so during times of volatility. This raises the question whether the profitability for an
outside investor is affected by the choice of which insider to mimic. To understand this
further in a Swedish context, the following hypotheses are formulated:
H2a: There is a significant difference in the abnormal returns of purchase transactions
depending on which insider position is mimicked.
H2b: There is a significant difference in the abnormal returns of sale transactions
depending on which insider position is mimicked.
9
3. Methodology and Data
3.1 Sample
3.1.1 Data Collection
The insider trading data stems from the PDMR transactions register managed by
Finansinspektionen. The register contains data starting from July 3 2016 and features
information about the issuer, the nature of the transaction, the role of the person discharging
managerial responsibilities, trading volume, price and other information. As insiders are
required to report details of their trades, including the transaction category, it makes it
possible to solely focus on transactions where the person discharging managerial
responsibilities buys or sells shares at the market price (Finansinspektionen, 2021). However,
in 2018, changes were made to the reporting format which resulted in a new classification of
trades. This classification is necessary to distinguish between stocks, options, grants and
other instruments. Therefore, only trades conducted after 17 September 2018 are included in
the sample. Further, the sample only includes companies listed on Nasdaq Stockholm, more
specifically OMX Stockholm Large Cap, OMX Stockholm Mid Cap and OMX Stockholm
Small Cap. Nasdaq First North is excluded as it is an alternative trading platform with its own
rules and regulations (Nasdaq, n.d.-b).
Daily stock data is gathered from Yahoo Finance using JupyterHub and the python library
yFinance. yFinance can access all financial data available on Yahoo Finance and allows for a
simplified extraction of large sets of data. The code for using yFinance to extract historical
prices is presented in Appendix 9. After dropping redundant information, the adjusted closing
prices are exported to Excel. Even though hourly data would be preferred as the match
between price data and events would be more precise, technical limitations hinder this option.
As the adjusted closing prices take stock splits into account while the event data from the
PDMR transaction register does not, some adjustments are required. yFinance is used to
identify companies in which stock splits occurred in the period of interest. The event data for
these companies is then adjusted according to the conditions of the splits. This is done by
multiplying (dividing) the number of shares and dividing (multiplying) the price in order to
make the market data and event data comparable. For example: Stock A underwent a stock
split of 2:1 on 21 March 2022 where the price changed from 100 SEK to 50 SEK. The event
data of stock A previous to this date assumes a stock price relative to 100 SEK as it does not
take the stock split into account. If left unadjusted, the return between 20 March and 21
March all other events excluded would be -50 percent which we know is not the case.
Therefore, the price of all transactions before 21 March are multiplied by 2 and the number of
shares in the trade is divided by 2. The total value of the trade therefore remains the same, but
10
the return, all other events excluded, is 0 percent.
Nasdaq OMX Stockholm is open between 09.00-17.30 on regular weekdays but the PDMR
transactions register is open for registration and publication at all hours. Because of this,
insider transactions can be published after the market closes, making it impossible for outside
investors to mimic the trade during that specific day. To adjust for this, transactions that are
published after the market closes are transferred to the next day in an attempt to capture the
full effect of the event.
The Fama French factors used in the Four Factor Model, Small Minus Big (SMB), High
Minus Low (HML) and Up Minus Down (UMD), are collected from AQR Capital
Management and are based on the work of Asness and Frazzini (2013). The risk free rate
used in the calculations is the yield from a Swedish 3 month treasury bill which is retrieved
from the Swedish Central Bank. In order to retrieve fundamental data for all companies in the
sample, Wharton Research Data Services is used. This platform provides access to multiple
databases with varying information and areas of use. In this report, Compustat, a data set
provided by S&P Global Market Intelligence, is used. Variables such as Total Assets, Total
Debt, Net Income and Stockholders Equity are retrieved and used to calculate the control
variables. In this study, OMXS30 is used as a proxy for market return. OMXS30 is a stock
index of the 30 most traded shares on Nasdaq OMX Stockholm and generally reflects the
movements in the Swedish stock market. This data was retrieved from Nasdaq Nordic (2023).
In order to generate abnormal return data for all events, the statistics software STATA is used
along with the eventstudy2 command.
3.1.2 Exclusions
Due to relevance, some trades are excluded from the data set. As Finansinspektionen
provides extensive information about the transactions, it allows for a level of specification.
This specification in turn allows for excluding data which is not deemed relevant. (1)
Transactions that have been revised and/or canceled are excluded. (2) Previous studies (De
La Brunière et al., 2020; Lin & Howe, 1990) have excluded all transactions that are not
open-market. This means that grant and award transactions, stock options and acquisitions of
shares through bonus or rights issues are excluded. Thus, only transactions which include an
insider buying or selling shares are included. (3) Trades that occurred before 17 September
2018 are excluded from the data set as they do not allow for differentiating between open and
non-open market transactions. (4) Trades which happen on Nasdaq’s First North market are
excluded. First North is a Multilateral Trading Facility (MTF) and is, thus, not regulated the
same way that Nasdaq OMX Stockholm is. (5) Trades which are reported more than five days
after the transaction date are excluded as the relevance is reduced. As insider trades are to be
reported within three business days, a limit of five days is set as it takes weekends into
11
consideration. (6) Due to technical limitations and time constraints, roughly 50 percent of the
data is sampled using random selection. (7) Finally, while running the eventstudy2 command
in Stata to calculate abnormal returns, some data is lost due to various reasons such as an
insufficient number of observations in the security return data during the event period.
Table 3.1 Sample Selection
This table shows the exclusions made from the initial sample.
The initial sample consists of 26 121 observations. Revised and canceled trades are removed
from the sample as they lack relevant or timely information. These trades suggest that some,
or all, of the information given by the trade has been altered. Due to the uncertainty, these
types of trades are therefore removed from the sample. Financial instruments not including
shares such as subscription warrants, BTAs and BTUs are also excluded from the sample as
these transaction categories are not always possible for an outsider to replicate. Non-buy or
sell transactions are removed for the same reason as stated above, namely that outsiders
cannot always replicate these trades as they are often traded at a discount. Trades by insiders
in companies which are not listed on Nasdaq OMX Stockholm, such as those listed on First
North, are removed from the sample due to regulatory differences. Further, trades
characterized by a gap of more than five days between the transaction date and the
publication date are excluded. This is because the vast majority of these trades fall outside of
the legal limit of three business days stated in Article 19 in regulation 596/2014. Due to
technical limitations and time constraints, a random sample of roughly 50 percent is utilized.
If a company makes it in the random sample, all trades conducted within this entity are
included. The reason for this is that if there are multiple trades conducted by insiders of a
specific company on recurring days, it may distort the data if only some of those trades are
included as events. These exclusions result in a final sample of 6 550 trades. The random
sampling is done several times to check for any differences in results.
12
Table 3.2 presents the characteristics of the final sample. Of the 6 550 transactions included
in the final sample, 77.69 percent of these are buy transactions and 22.31 percent are sell
transactions. Within the sample, we divide these 6 550 transactions into categories based on
the type of insider: CEO, CFO, Board, or Other. The largest of these groups is Other whose
transactions constitute 53.33 percent of the sample, followed by Board at 23.38 percent, CEO
at 15.49 percent, and finally CFO at 7.05 percent. One potential reason for the uneven
distribution between the groups is that there are multiple board members and “others” in a
company while there is only one CEO and CFO.
Table 3.2 Description of Sample
This table shows a description of the sample with transactions divided into buy and sell as well as insider
position.
3.2 Variable Selection
3.2.1 Dependent Variable and Study Design
In this study, the dependent variable is the cumulative abnormal return. In order to measure
the impact of insider trades on stock prices and in turn abnormal returns, an event study
approach is used. This approach has been utilized by many researchers and is suitable as it
allows for studying specific events and the effects thereof. Previous research on insider
trading has used estimation windows between 100-200 trading days (Van Geyt et al., 2013;
Chang & Suk, 1998; Betzer & Theissen, 2009). This is in line with Armitage (1995) who
stated that an estimation period can be anywhere between 100 to 300 trading days when
working with daily observations. This study utilizes an estimation window of 180 trading
days. According to MacKinlay (1997), the estimation window and the event window should
typically not overlap as it could lead to the event returns heavily affecting the normal return
measure. To avoid that the effects of an event are included in the estimation window, a
trading day gap of 20 days is set. In order to establish the normal return of a security the Four
Factor Model is used.
13
The Three Factor Model for pricing stocks was developed by Eugene Fama and Kenneth
French in 1993, and is commonly referred to as the Fama-French model. The Fama-French
model controls for three factors when pricing stocks: risk, price, and company size. These
factors are included in the model as βmarket for systematic risk, SMB (Small-Minus-Big) to
control for size and HML (High-Minus-Low) to account for the price of the stock. Later in
the 90s, this model was further developed by Mark Carhart who added a momentum factor to
the Fama-French Three Factor Model and it is now referred to as the Fama-French-Carhart
Model or the Four Factor Model. The momentum factor is included in the model as UMD
(Up-Minus-Down) to account for how quickly an individual stock changes its price. The
Fama-French-Carhart Four Factor Model, thus, looks as follows:
𝐸[𝑅 ] − 𝑟 = α𝑐 + β (𝐸[𝑟 ] − 𝑟 ) + β + β + β + ε
𝑖, 𝑡 𝑓 𝑀𝑎𝑟𝑘𝑒𝑡 𝑚 𝑓 𝑆𝑀𝐵 𝐻𝑀𝐿 𝑈𝑀𝐷
(Eq. 1)
where rf is the risk free rate, E[rm] is the expected market return, SMB is Small Minus Big,
HML is High Minus Low and UMD is Up Minus Down.
In order to appraise the impact of an event, the abnormal return can be used. This is the actual
ex post return of a security over the event window minus the normal return of the company.
The abnormal return (ARi,t) is calculated as follows:
𝐴𝑅 =  𝑅  −  𝐸[𝑅 ] (Eq. 2)
𝑖, 𝑡 𝑖, 𝑡 𝑖, 𝑡
where the abnormal return for firm i on day t (ARi,t) is calculated by subtracting the expected
return (E[Ri, t]) from the actual return for firm i on day t (Ri, t). The abnormal return is then
used to calculate the cumulative abnormal return (𝐶𝐴𝑅 (𝑡 , 𝑡 )):
𝑖 1 2
𝑡
2
𝐶𝐴𝑅 (𝑡 , 𝑡 ) = ∑ 𝐴𝑅 (Eq. 3)
𝑖 1 2 𝑖, 𝑡
𝑡=𝑡
1
The cumulative average abnormal return can be used to analyze the total effect of all events:
𝑡
2
∑ 𝐴𝑅
𝑖, 𝑡
𝑡=𝑡
𝐶𝐴𝐴𝑅(𝑡 , 𝑡 ) = 1 𝑁 (Eq. 4)1 2
where N is the total number of observations.
14
Previous studies differ when it comes to the length of event windows (Lakonishok & Lee,
2001; Eckbo & Smith, 1998; Betzer & Theissen, 2009) depending on the type of study. While
some studies focus on the long-term effects of insider trading, others have focused on a
shorter time horizon. As previous studies have found the effects of insider trading to be
greater in the short term (De La Brunière et al., 2020), this study utilizes short event
windows. These event windows are: day t-5 to day t-1, day t-1 to day t1 and day t0 to day t5,
where day t-1 to day t1 is considered the main event window of interest. The reason for
including the other two windows is to observe if any trends can be distinguished before or
after the event. These windows can also act as an indication of the market timing ability of
insiders.
When investigating the market reaction to insider trades by looking at share prices, the event
date can be defined as the transaction date or the publication date. The difference between
these two is that the former is the day the trade is executed while the latter is the day it is
made public, often through publication in a PDMR register (Betzer & Theissen, 2009). As
this study aims to establish whether an outside investor can achieve abnormal returns by
mimicking insider trades, the event of interest that takes place on day 0 is the publication date
rather than the transaction date. This is because it is not possible for an outside investor to act
on the transaction date given that the information is not yet public. This is true for both
hypotheses.
In order to test the hypotheses, a multi-way fixed effects regression model is utilized. This
model suits the data well as it can handle cases where there are multiple observations per time
unit. The model also eliminates singleton groups which reduces the chance of overstating
statistical significance (Correia, 2015).
3.2.2 Independent Variables
In order to determine the impact of the position of the insider and whether it has an effect on
abnormal returns, a set of dummy variables are created. All of these variables take on a value
of 1 if true and 0 if false. A dummy variable will be created for each of the following: CEOs,
CFOs, board of directors, and others. The variable Other includes persons related to the
insider, other management, management of parent company or subsidiary, and companies
with some type of insider relation. When the variables are assigned to a transaction, a
hierarchy is used to determine the position of insiders with two or more roles. The hierarchy
is dependent on the perceived level of knowledge possessed by the insider, where CEOs are
determined to be first, CFOs second, board of directors third, and others fourth. This imposes
that if a person discharging managerial responsibilities is both the CEO and the CFO of a
company, the person will be categorized as CEO.
15
In addition to this, a dummy variable will be created for purchases and sales depending on the
type of transaction. Some studies have excluded sell transactions as the reason behind a sale
might be unclear. A sale could have numerous explanations, such as liquidity reasons or
diversification into other stocks. There are far fewer reasons as to why an insider would
purchase shares in their firms and according to Lynch and Rothchild (2000), it is because they
believe the stock is undervalued. As this study investigates whether outsiders can earn
abnormal returns by mimicking insiders rather than the performance and market timing of the
insider, the reason behind a transaction is not of importance. Therefore, both purchase and
sale transactions are included.
3.2.3 Control Variables
In order to enhance the internal validity of the study and verify that the result, i.e. the
magnitude of abnormal returns, is not influenced by extraneous variables, a set of control
variables are included. These specific variables were chosen because they serve as proxies for
profitability, leverage, size and growth prospects which covers a wide range of areas that
could have an impact on the results. The variables are presented below:
Return On Assets (ROA) is defined as the Net Income divided by the Total Assets of the
company. ROA is a commonly used KPI in finance to approximate the profitability of a
company. Previous studies (Davis et al. 2017; Betzer & Theissen, 2009; Sha et al., 2020)
have included some type of profitability metric as a control variable. The relationship found
between this variable and abnormal returns is negative entailing that companies that are more
profitable present lesser options for generating abnormal returns.
𝑅𝑂𝐴 =  𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠 (Eq. 5)
Debt-to-Equity (D/E) is defined as the Total Debt divided by the Book Value of Equity. D/E
is commonly referred to as the firm’s leverage as it shows how much the company is funded
by debt versus equity. According to the Pecking-Order Theory (Myers & Majluf, 1984), less
profitable firms are more leveraged as they are less able to fund their growth through internal
capital. Further, companies with high leverage are often required to disseminate more
information to creditors, and therefore the public. This reduces information asymmetry which
is expected to have a negative impact on the abnormal return (Aksu & Kosedag, 2006). This
variable is included to test if the way a company is funded has an impact on the possibility of
returns made by the investor.
𝐷/𝐸 = 𝑇𝑜𝑡𝑎𝑙 𝐷𝑒𝑏𝑡𝐵𝑜𝑜𝑘 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝐸𝑞𝑢𝑖𝑡𝑦 (Eq. 6)
16
Size is defined as the natural logarithm of the Total Assets of the given company. Previous
studies (Seyhun, 1998; Betzer & Theissen, 2009) found that the insider trading profitability is
generally higher in smaller companies. The notion is that in larger firms, a given piece of
information is less likely to drastically affect the stock price. As the sample in this study
consists of companies of various sizes, including such a variable is important for the quality
of the research.
𝑆𝑖𝑧𝑒 =  𝐿𝑛(𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠) (Eq. 7)
Price-to-Earnings (P/E) is defined as the Share Price divided by Earnings-Per-Share (EPS),
further divided by 100. This ratio shows the expectations of the market and is therefore used
to predict relative future stock returns. Seyhun (1998) found that it also has various
implications in the field of insider trading. The P/E ratio has a negative effect on abnormal
returns entailing that lower P/E firms are more likely to generate greater abnormal returns
than higher P/E firms.
𝑃/𝐸 =  𝑆ℎ𝑎𝑟𝑒 𝑃𝑟𝑖𝑐𝑒𝐸𝑃𝑆 𝑥 100 (Eq. 8)
3.3 Managing Outliers
There are different methods that can be used in order to deal with extreme values in a data
set. Within the field of finance, winsorizing, trimming, and dropping variables are common
approaches. According to Adams et al. (2018), a vast majority of finance studies utilize
winsorization where extreme outliers are limited to a selected percentile. This allows the
observation to stay in the sample rather than removing the entire row observation. In this
paper, winsorization is applied to selected variables in order to limit the impact of
extraordinary data points. Table A in Appendix 1 presents the variables at varying percentiles
before the winsorization. As displayed in the appendix, there are several variables where the
minimum value is at a considerable distance from the 1st percentile. Some variables also
display maximum values that are far greater than those at the 99th percentile. An initial
winsorization at the 1 percent and 99 percent level is carried out. This entails that all values
below the 1st percentile are set to the value of the first percentile and all values above the
99th percentile are set to the value of the 99th percentile. Table B in Appendix 1 shows the
variables used in this study with those affected by winsorization. All variables except the
dummy variables along with Risk free rate were winsorized. The reason for excluding Risk
free rate is that it lacks extreme values with the minimum and maximum value being close to
17
the next percentile.
When merging the event data containing the trades and the control variable data, the event
data is lagged one year so that an event from 2022 is matched with control variables from
2021. This is done as key performance indicators are often presented on an annual basis and
matching the observations from 2023 would otherwise not be possible. Lagging the event
data one year does however lead to a number of observations being dropped as a result of
missing control variable data. Several companies in the sample have only been listed on
Nasdaq OMX Stockholm for a few years. As the control variable data builds on input such as
share price, calculations for the first year on the stock market is not possible.
3.4 Descriptive Statistics
Illustrated in Table 3.3 is the descriptive statistics for all of the variables used in the analysis.
Firstly, looking at the dependent variable, Cumulative Abnormal Return, we observe that all
but one of the event windows have a negative skewness. CAR(0;5) Sell is positive indicating
a left skew and a median value which exceeds the mean. Looking at the median and mean
values, both the three day event, (CAR(-1;1), and the five day event, (CAR(0;5), for buy
transactions have positive median values. However, the mean for the five day CAR,
CAR(0;5), for buy is negative. This along with having the largest negative skewness of any
of the event windows indicates that there are some extreme values which affect the mean
negatively while the median value is positive. When examining the kurtosis for each CAR, it
can be concluded that all of them express a kurtosis greater than five. This indicates heavy
tails and significant outliers which are affecting the mean values.
Looking at the winsorized control variables, all but Size have high kurtosis scores indicating
the presence of significant outliers. While Price-to-Earnings and Debt-to-Equity have positive
skews and mean values, the opposite is true for ROA. ROA is negatively skewed and has a
negative mean, but has a positive median. This indicates that there are some periods of
significant loss for the companies in the sample which makes the mean value negative while
the median stays positive. This matches what can be seen in the minimum and maximum
values for ROA. The minimum value in the sample for ROA is -128.11 percent while the
maximum is 31.71 percent. For P/E and D/E, the maximum values are greater than the
minimum in absolute value which makes sense as they are positively skewed. Looking at the
minimum value for ROA, it could be seen as being quite significant and possibly another
outlier. However, when examining the data we observe that there are other data points which
are close to the minimum and the jump between each observation is not large. As the sample
consists of firms of different sizes in different industries, the control variables can differ
significantly in magnitude between the companies. For a newly introduced medtech
company, it is not uncommon to see a negative ROA as this type of company is characterized
18
by large initial investments and R&D costs which usually takes several years to turn into
profit. Similarly, an established company in the manufacturing industry, such as a car
manufacturer, is expected to have a more stable ROA. In our sample, the ROA of -128.11
percent comes from a medical research company and, thus, is a figure that could be
reasonable in that field. The same things can be observed from the P/E ratio, which has a
minimum and maximum value that could be considered quite significant. However, as with
ROA, there are other data points which are close to the minimum and maximum values and
differences between them are considered reasonable. A large P/E ratio, negative or positive,
indicates that a firm has a high price or small returns, or a combination of the two. If the
returns are positive (negative), the P/E ratio will also be positive (negative).
Table 3.3 Descriptive Statistics
This table shows descriptive statistics for variables in the study. This includes the number of observations,
minimum value, mean, median, maximum value, standard deviation, skewness and kurtosis.
When examining all variables in a study, it is also important to take their interaction with
each other into account. According to Shrestha (2020), multicollinearity occurs when a linear
regression analysis includes several variables that are significantly correlated with the
19
dependent variable and each other. Multicollinearity can lead to some significant variables
becoming statistically insignificant. Generally, multicollinearity is said to exist if the absolute
value of a Pearson correlation coefficient is close to 0.8. In order to control for this, a
pairwise correlation test is carried out and the results are presented in Appendix 2. As the
highest correlation between two variables, based on the value of the Pearson correlation
coefficient, in this sample is -0.600, no further considerations have to be made.
3.5 Model Specification
In order to determine if outsiders can generate abnormal returns by mimicking insiders’
trades, cumulative abnormal return is utilized. When conducting tests without control
variables, the following model is used:
𝐶𝐴𝑅 = 𝛽0+ ɛ (Eq. 9)(𝑖,𝑡)
where 𝐶𝐴𝑅 is the cumulative abnormal return and i represents the sample constituting the
(𝑖,𝑡)
dependent variable, where t represents the different event windows, (-5;-1), (-1;1) and (0;5).
𝛽0 represents the intercept of the regression which is the mean of the dependent variable and
ɛ is the error term.
In order to test hypothesis 1a and 1b, the primary hypotheses of this study, Equation 10 is
used along with Equation 9 with the only difference being inclusion of the control variables.
4
𝐶𝐴𝑅 = 𝛽 + ∑ 𝛽 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + ε (Eq. 10)
(𝑖,𝑡) 0 𝑘
𝑘=1
The following regression uses the CAR generated from running the event study for the entire
sample of buy and sell transactions respectively, and regressing all roles excluding the group
Other. Thus, the following model is used:
𝐶𝐴𝑅 = 𝛽 + 𝛽 𝐶𝐸𝑂 + 𝛽 𝐶𝐹𝑂 + 𝛽 𝐵𝑜𝑎𝑟𝑑 + ε (Eq. 11)
(𝑖,𝑡) 0 1 2 3
The first regression is also used for the purposes of answering hypotheses 2a and 2b, which
means that the roles need to be introduced into the model. Thus, the following model is used:
𝐶𝐴𝑅 = 𝛽 + 𝛽 𝑅𝑜𝑙𝑒 + ε (Eq. 12)
(𝑖,𝑡) 0 1
20
The following regression uses the CARs generated from running the study only using the
events from specific roles. Thus, there will be a CAR CEO, CAR CFO, CAR Board, and
lastly a CAR Other. These CARs are generated using the following model:
𝐶𝐴𝑅 𝑅𝑜𝑙𝑒 = 𝛽 + ε (Eq. 13)
(𝑖,𝑡) 0
It is common for panel data studies in this field to cluster standard errors on firms. In this
study however, robust standard errors are used. The reason for this is that there are several
observations per time unit in the data and thus it does not fit into the frame of a traditional
panel data study. The standard errors for all models are calculated using Equation 14 for the
constant terms (𝛽 ) and Equation 15 for the coefficients (𝛽 ). The equations are as follows:
0 𝑖
Σ𝑥2
𝑆𝐸(β ) = 𝑠 𝑖 (Eq. 14)
0 𝑁Σ(𝑥 −𝑥)̄ 2
𝑖
𝑆𝐸(β ) = 𝑠 1 (Eq. 15)
𝑖 Σ(𝑥 −𝑥̄)2
𝑖
where s is the standard deviation of the residuals, 𝑥 is the individual observation, 𝑥̄ is the
𝑖
mean of all observations and N is the sample size.
In addition to the regressions, a number of t-tests are performed in order to determine if there
are significant differences in the means of different variables. As the sample sizes differ,
Welch’s t-test is used. This allows for unequal sample sizes and variances which in turn
produces more robust results (Moser & Stevens, 1992).
3.6 Robustness
3.6.1 Transaction Date
In accordance with Regulation 596/2014, insiders must report their trades in their company
within three business days of the transaction. Thus, there will be a period between the date of
the transaction and the date of the publication where information about the trade is limited.
As is reported in Table 3.1, transactions where the date of the transaction and the date of the
21
publication differ by more than five days are excluded from the sample. Therefore, the
sample can contain data that varies from a difference of zero to a difference of five days.
When executing our study, the publication date is used as the measure for when the event
occurs, as an outsider can only mimic an insider once they are aware of the transaction.
Previous studies (De La Brunière et al. 2020; Mazza & Ruh, 2022) have found that insiders
can generate abnormal returns and that they have market timing ability to do so. As a
robustness check, a regression using the transaction date instead of the publication date will
be performed. If no abnormal return is found in the events using the publication date, we
want to observe if this abnormal return is captured using the transaction date. Therefore, the
difference in means between the CARs for the transaction and publication date are compared
and a regression is run where the event date is set to the transaction date.
3.6.2 Market Adjusted Model
As a robustness check, the calculations of abnormal returns are run using the Market
Adjusted Model as opposed to the Four Factor Model. While the Four Factor Model includes
factors that control for systematic risk, size, price and momentum, the Market Adjusted
Model uses the market’s return on a specific date as the expected return for the stock. The
abnormal return using the Market Adjusted Model is calculated as follows:
𝐴𝑅 =  𝑅  −  𝑅 (Eq. 16)
𝑖, 𝑡 𝑖, 𝑡 𝑚, 𝑡
where the abnormal return for firm i on day t (ARi, t) is calculated by subtracting the return of
the reference market m on day t (Rm, t) from the actual return of firm i on day t (Ri, t). The
cumulative abnormal return and cumulative average abnormal return is then calculated using
Equations 3 and 4.
3.6.3 Excluding Others
This study aims to understand if an outsider can mimic the trades of an insider and generate
abnormal returns for themselves. In our group of insiders, we have CEOs, CFOs, board
members, and lastly a group which have been characterized as Other. In the group Other,
there are several different types of insiders included which do not meet the criteria of the
remaining groups. These insiders range from other companies, employees in lower level
managerial positions, as well as persons related to insiders which have an obligation to report
their trades. Previous research (Mazza & Ruh, 2022; Seyhun, 1998) has found that there are
differences in the level of abnormal return that can be obtained based on the hierarchical level
of the insider. Insiders which are higher up in the hierarchy, i.e. CEOs, CFOs, and board
22
members, are better able to generate abnormal returns partly due to the increased access to
important information. The group Other has differing levels of insider knowledge within the
group but also in comparison to the other groups. As the group Other constitutes roughly 53
percent of the total sample and therefore contributes greatly to the total result, it is warranted
to run the events while excluding this group to check for differences. Therefore, as a
robustness check, we are testing for the difference in means between the CARs for the data
set with the group Other and the one without.
3.6.4 Clustered Standard Errors
The structure of the data of buy and sell transactions includes the year of the transaction and
the ticker of the associated company which is then matched to yearly control variables. Due
to the fact that there are some companies which have more than one transaction in a given
year, the sample can not be assumed to be panel data. Therefore, we use robust standard
errors in our main regressions instead of clustered standard errors. As a robustness check, a
regression using clustered standard errors on firm instead of robust standard errors is run and
the results are compared.
23
4. Results
4.1 The Profitability of Insider Trading
The first test in the results section is to determine if the cumulative abnormal return (CAR)
for buy and sell transactions for events t-5 to t-1, t-1 to t1, and t0 to t5 are significantly different
from zero. Table 4.1 illustrates that all CARs are significantly different from zero, except the
CAR for sell transactions five days prior to the event; model (4). The main event window, t-1
to t1, is significant at a 1 percent level for both buy and sell transactions. Looking more
specifically at the CARs for buy transactions, five days prior to the event the CAR is found to
be -0.60 percent, while the three day CAR, model (2), is found to be positive at 0.29 percent.
However, looking at model (3) the CAR is negative again at -0.35 percent. The
announcement of the event coincides with a positive abnormal return, but it is not significant
enough for CAR to be positive the five days after the event. For the sell transactions, the
CAR five days prior to the sale (model (4)) is not statistically significant. The three day CAR
for sell transactions, model (5), is statistically significant at -0.66 percent. The five days
following a sell transaction, the CAR is even more negative at -0.99 percent.
Table 4.1 Test for the significance of CAR, all insiders
This table shows the regression results from Equation 9 where it is tested if CAR is significantly different from
zero.
In Appendix 3, the annualized cumulative abnormal return is calculated. For the three day
main event window, the annualized CAR for buy transactions is 27.65 percent and -42.87
percent for sell transactions. When annualizing the CAR for day 0, the corresponding values
are -14.04 percent and -47.98 percent. These differences are explained by Table 4.2 where the
only day with positive CAR is t-1 for buy transactions and where the value of t1 for sell
transactions lower the returns in absolute terms.
Table 4.2 shows the average abnormal return (AAR) for buy and sell transactions day by day.
For the buy transactions, the only day with a positive AAR is t-1 at 0.44 percent. As the main
event window is significantly positive, it means that the full effect comes from this day. The
results from the sell transactions are somewhat different. AAR is positive for each day
24
between t-5 and t-2, although only significant for half of them. The return changes to negative
at t-1 and this trend continues until t5, with the exception of t3 being positive. Given that
several days in the post-event window are significantly negative, it entails that the sell
transactions can be utilized for a longer period of time compared to the buy transactions
where only one day constitutes a profitable trade. The fact that AAR Sell changes between
being significant and not during the event window could potentially be explained by the fact
that there are not as many observations in this sample.
Table 4.2 Daily Average Abnormal Return
This table shows the average abnormal return on a daily basis for buy and sell transactions.
Graph 4.1 displays the cumulative average abnormal returns (CAARE) from t-20 to t20 for the
buy transactions in the sample. CAARE reaches its peak of 0 at t-19 and declines until t-2
where the direction changes. After the small increase in CAARE between t-2 and t-1, the
downward trend continues. As illustrated in the graph, the downward trend is followed by a
jump in the price. This jump happens two days prior to the publication of the transactions,
which coincides with the average number of days between the date of the transaction and the
date of the publication. Thus, the positive abnormal return found in Table 4.1 for model (2), is
more likely to stem from the transaction rather than the publication. The downward trajectory
before t0 illustrated in Graph 4.1 indicates that, on average, insiders buy shares when their
company is performing worse than expected.
25
Graph 4.1 Cumulative Average Abnormal Return - Buy
Graph 4.2 displays the cumulative average abnormal returns (CAARE) from t-20 to t20 for the
sell transactions. CAARE reaches its peak at t-3 before starting to decline. Similarly to the buy
transactions, t0 is right of the turning point illustrated in the graph. The downward trajectory,
which was preceded by an upward trajectory a number of days prior, comes two days prior to
the date of the publication at t0. Similarly to the buy transactions, the difference in days
between the date of the transaction and the date of the publication is roughly two days. Thus,
the negative CAR illustrated in Table 4.1 for the three day sell transactions is more likely to
stem from the negative reaction to the transaction rather than the publication. The positive
CAARE before t0 in Graph 4.2 indicates that, on average, insiders sell shares when their
company is performing better than expected.
Graph 4.2 Cumulative Average Abnormal Return - Sell
26
In Table 4.3 the regression results from Equation 10, where control variables are included, are
presented. The first control variable, D/E, shows no significance for five out of six models,
indicating that it has no effect on the magnitude of cumulative abnormal returns. For model
(5), the significance is partial at a 10 percent level. The next variable, ROA, is significantly
negative at a 1 percent level in model (1) and an increase of one standard deviation entails a
decrease in CAR of 3.67 percentage points. Model (6) shows a negative significant
relationship at a 5 percent level. The relationship indicates that after the event, an increase by
one standard deviation in ROA is associated with a decrease of 4.57 percentage points in
CAR. This implies that higher ROA firms generate greater abnormal returns in absolute terms
after the event. The control variable for size (Size) is negative and significant at a 1 percent
level for models (1) and (2) while positive in model (3). For the sell transactions, Size is not
significant before the event but models (5) and (6) are significant at a 1 percent level. The
results for models (1) and (2) indicate that the CAR is negatively affected by an increase in
firm size where an increase of one standard deviation decreases CAR by 0.42 percent and
0.12 percent respectively. For sell transactions, an increase in Size by one standard deviation
increases CAR by 0.17 percent and 0.38 percent for models (5) and (6) respectively. This
implies that smaller companies generate greater abnormal returns in absolute terms for both
buy and sell transactions. The last variable, Price-to-Earnings (P/E), shows no significance
for the buy transactions in model (1) through (3). For the sell transactions, model (6) shows a
partial negative significant relationship at a 10 percent level.
Table 4.3 Regression results, including controls
This table shows the regression results of Equation 10 where control variables are included.
27
4.2 Differences In Profitability Between Insider Positions
Table 4.4 shows the results for a test of the significance of CAR, using the regression model
in Equation 11, where each insider's contribution to the total CAR for the different event
windows is illustrated while excluding the group Other. The first regression that includes all
insiders looks at the effect of each insider on the total CAR while adjusting for each of the
other groups and holding Other as constant. Firstly, looking at the regression run using CEO,
CFO, and Board we observe that the CFO is significant in model (1) at a 5 percent
significance level. Thus, the CAR for CFO transactions prior to the event is contributing
significantly to the total CAR and is associated with a 0.98 percent decrease in CAR. For the
remaining insider positions, no significance is found prior to the event. Turning to model (2),
all insiders have a significant impact on the total CAR at a 1 percent level. While CEOs and
board members’ buy transactions contribute to the total CAR at 0.60 percent and 0.55 percent
respectively, CFO transactions seem to have the largest impact at 0.91 percent. Finally,
observing the results for model (3) for buy transactions it is only board member transactions
that are contributing significantly to the total CAR and it is significant at a 1 percent level.
The remaining insider positions show no significance.
When examining the sell transactions for insiders, it is only model (4) where significance is
found in one of the independent variables. Board is contributing 1.03 percent to the total
CAR at a 5 percent significance level. Looking at models (4) and (5), only CFO sell
transactions are associated with a negative CAR. Thus, prior to the event a CFO sale is
negatively associated with CAR and is associated with a 1.03 percentage point decrease in
CAR. Both for sell transactions for CEOs and board members, the announcement of the sale
of shares does not mean that the CAR for that period will be negative.
28
Table 4.4 Regression results, including roles
This table shows the regression results of Equation 11 where the dependent variable is CAR for the entire buy
sample and sell sample respectively.
Illustrated in Table 4.5, is the regression results for Equation 12 where each insider group is
regressed by itself. CEO buy transactions are significant at a 10 percent level in models (1)
and (2), but show no significance for model (3) or the sale transactions illustrated in models
(4), (5) and (6). For CEO buy transactions, the association with CAR is positive for model (1)
and model (2). CEO transactions five days before the announcement are associated with a
0.54 percent increase in CAR and CEO transactions during the three day window are
associated with a 0.33 percent increase in CAR.
In Table 4.5, looking at the results for CFOs, models (1) and (2) are significant at a 5 percent
level and model (5) is significant at a 10 percent level. When the regression is run for CFOs
holding other roles constant, the effect on CAR is larger than the effect recorded in Table 4.4.
Comparing the results for model (5) in the first and second regression, the coefficient is
insignificant in the first regression and significant at a 10 percent level in the second
regression. The opposite is true for model (2) where the coefficient for the first regression
illustrated in Table 4.4. is larger at 0.91 percent than the second regression illustrated in Table
4.5 at 0.64 percent.
Turning to the board member transactions, model (2) for Board is significant at a 10 percent
level but is found to be greater in the first regression with a 1 percent level significance. In
both regressions, Board is significant at a 1 percent level in model (3) with a 0.72 percent
effect in the second regression. Board member transactions are also found to be significant in
29
model (4) at 1 percent significance level while only a 5 percent significance level in the first
regression. Finally, the reference group Other is significant at a 1 percent level in two of the
three models, (2) and (3), for the buy transactions but no significance in the models for the
sell transactions. Interestingly, unlike the other insider groups, both coefficients for the group
Other in the second regression are negative when holding the other groups constant. Thus,
buy transactions made by Other seem to be negatively associated with a 0.62 percent
decrease in CAR for model (2) and associated with a 0.51 percent decrease in model (3).
Table 4.5 Regression results, separated roles
This table shows the regression results from Equation 12 where each insider group is regressed with the
dependent variable being the CAR for the entire buy sample and sell sample respectively .
The results from testing the significance of CAR for all roles is presented in Table 4.6. Model
(2) shows that the CEO group has a positive CAR of 0.49 percent at a 1 percent significance
level. This is followed by a negative CAR of -0.69 percent for the days after the event. For
30
the sell transactions, model (6) is the only one showing significance with a CAR of -2.55
percent. The buy transactions for the CFO group show similar results but the CAR in model
(2) is higher at 0.92 percent. This group also shows significant results for the sell transactions
in model (5) at -1.58 percent. Similarly to CEO, the return for the five days following the
event is significantly different from zero with a CAR of -1.62 percent. When examining the
results for the Board, it can be seen that for the three day buy event displayed in model (2), a
CAR of 0.52 percent is reached at a 1 percent significance level. For the corresponding sell
transaction, the CAR is -0.57 percent. The last group, Other, is the only group showing no
significance for the three day buy window displayed in model (2). The group does however
yield a negative CAR of -0.63 percent for the corresponding sell event in model (5). For all
buy transactions, model (1) shows that the CAR is negative during the days leading up to the
event. For the sell transactions, an opposite trend can be seen in model (4) with the exception
of the CFO group. This indicates that for the days leading up to a buy event, the cumulative
abnormal return is lower than expected. Similarly, it shows that for the days leading up to a
sell event, the CAR is higher than expected.
Table 4.6 Test for the significance of CAR, separated roles
This table shows the regression results from Equation 13 where it is tested if the CAR for each role is
significantly different from zero
In order to determine the difference in abnormal return between the different insider groups, a
Welch’s t-test is carried out between each of the groups and the results are illustrated in
Appendix 4. Some observations are dropped which is a result of singleton observations being
removed. When comparing the difference in means between CEO and CFO, there is a
31
significant difference between two of the six CARs. A significant difference at a 5 percent
level can be found between the CAR for buy transactions five days before the event and a
significant difference at a 10 percent level can be found for CAR five days after the event.
When comparing the difference in means between CEO and Board, a significant difference at
a 1 percent level is found for both buy and sell transactions five days after the event. When
examining the buy transactions, the CAR for CEO is negative at -0.70 percent for the five
days after the publication, while the CAR for board members is positive at 0.20 percent in the
same time span. Comparing the difference in means between CEO and Other, a significant
difference is found at a 5 percent level for buy transactions during the three day event and for
the sell transactions five days after the event. Comparing the difference in means between
CFO and board members buy and sell transaction, there is very little significant difference to
be found. Only the CAR for buy transactions five days prior to the event and the five days
after the event for sell transactions are found to be significant at a 10 percent level.
Comparing the difference in means between CFOs and Other, there is a significant difference
at a 5 percent level for buy transactions five days prior to the event and a significant
difference at a 1 percent level for buy transactions during the three day event. Finally, when
comparing board members’ transactions versus that of others, a significant difference is
found in three of the six event periods. There is a significant difference at a 1 percent level for
buy transactions during the three day event and the five days after the event. There is also a
significant difference at a 5 percent level during the five days after the event for the sell
transactions.
4.3 Summary of Results
The results in section 4.1 support hypothesis 1a and the null hypothesis is rejected as the
main event window, CAR(-1;1), is significantly positive. For sell transactions, hypothesis 1b
is supported and the null hypothesis is rejected as the main event window is significantly
negative. The results in section 4.2 support hypothesis 2a and the null hypothesis is rejected
as there is a significant difference in CAR found for the main event window. However, no
such difference can be found for sell transactions between the roles in the main event window
and, thus, the null hypothesis cannot be rejected and no reliable support is found for
hypothesis 2b.
4.4 Robustness
4.4.1 Transaction Date
Illustrated in Appendix 5 is a test of difference in means of CAR between the publication date
sample and the transaction date sample. Looking at the amount of observations included in
each sample, we observe that seven more observations are now included in the sample of buy
transactions while five more observations are included in the sample of sell transactions using
32
the transaction date sample. This could be a result of the fact that some observations in the
publication date sample lack actual return data for the entire event window. Excluding these
observations from both samples yielded no difference in the significance of the results. When
examining the results for the test of difference in means, there are some significant
differences in the buy and sell transactions. For the five day window leading up to the event,
there are significant differences in both buy and sell transactions. For the sell transactions, the
publication date sample generated a mean CAR of 0.20 percent while the corresponding
number for the transaction date sample was 1.05 percent. For the remaining sell event
windows, there were no significant differences. For the buy transactions, there are significant
differences in both the five day event window before the event as well as the five day
window after the event. The most notable difference here is that CAR(0;5) is negative at
-0.04 percent for the publication date sample while the transaction date sample is positive at
0.24 percent.
The regression results from Equation 9 with the transaction date sample is presented in
Appendix 5. We find that the main difference between this result and that of the main
regression is that the CAR for all event windows in the transaction date sample are
significantly different from zero. In Table 4.1, the CAR for sell transactions five days prior to
the event, no significance is found. Meanwhile, the CAR for the same event using the
transaction date sample is significant at a 1 percent level. Another relevant difference
between the two results is the CAR from five days after the event. In Table 4.1, this CAR is
significantly different from zero but negative. For the transaction date sample, this CAR is
positive and significantly different from zero at a 5 percent level. This, in conjunction with
Graph 4.1 and 4.2, supports the notion that stock price reactions from the insider trade are
more likely to be closer to the date of the transaction rather than the date of the publication.
These results prompt the question whether it is more appropriate to use the publication date
or the transaction date for this type of study. While the transaction date generates more
significant results and lies closer in time to the perceived market reaction, this study attempts
to answer whether it is possible to earn abnormal returns mimicking insiders. An outside
investor could trade solely on the observation of abnormal trading activity. However, before
the transaction is published the outside investor cannot know who performed the trade. It is
therefore not possible to know if you are mimicking an insider, and what position this insider
holds. This entails that utilizing the publication date is the appropriate method for this study.
4.4.2 Market Adjusted Model
Appendix 6 shows the result of comparing the means of CAR for all event windows using the
Market Adjusted Model as well as the Four Factor Model. For the buy transactions, the only
statistically significant difference is that the CAR five days prior to the event using the Four
33
Factor model results in a mean of -0.60 percent and using the Market Adjusted Model
generates a mean of -0.20 percent. This results in a significant difference between the two at a
5 percent confidence level. For the sell transactions, the means are significantly different for
all variables. Worth noting is that all means that are positive (negative) in the Four Factor
model are also positive (negative) in the Market Adjusted Model.
Appendix 6 shows the regression results using Equation 16, the Market Adjusted Model. All
variables are significantly different from zero which is not true for the main regression model
presented in Table 4.1. When using the Four Factor Model, model (4) depicting CAR(-5;-1)
for sell transactions was not significantly different from zero. Apart from this, there are other
differences in the level of significance obtained. As the only variable separating the Four
Factor Model from the Market Adjusted Model is the expected return where the former
estimates normal return using a number of variables while the latter only uses the actual
return of the market, the difference could be explained by this variable.
4.4.3 Excluding Others
Illustrated in Appendix 7 is a test of difference in means comparing the CAR of the total
sample with the CAR of a sample excluding the group Other. When excluding the group
Other, 2 432 out of 5 089 buy transactions are removed from the sample. For the sell
transactions, 1 061 out of 1 461 observations are removed. The difference in the amount of
buy transactions removed from the sample versus the amount of sell transactions that are
removed shows the difference in the frequency that top executives like CEOs, CFOs, and
board members purchase and sell their shares. These groups are not as keen to sell their
shares as they are to buy. Moreover, observing the difference in means between the two
samples we see that for the most part, there is not a significant difference between them. The
only significant difference is the three day event window for buy transactions. The mean
CAR generated by purchasing shares in the sample excluding Other is 0.57 percent, while the
mean CAR generated by purchasing shares in the total sample is 0.29 percent. Looking at the
five day CAR after the event for buy transactions, the mean CAR including Other is -3.5
percent while the mean CAR excluding Other is -1.5 percent which could seem like a
significant difference. However, due to the size of the standard error, there is no significant
difference found in those means.
Illustrated in Appendix 7 is the result from the regression run on the sample excluding the
group Other. In comparison to the main regression, where others are included, there is less
significance. CAR is found to be significantly different from zero at a 1 percent level for buy
transactions during the five days prior to the event and the three day event. CAR is also found
to be significantly different from zero for sell transactions at a 1 percent level for the three
day event and a 5 percent level for the five days after the event. While there are less events
34
where significance is found, CAR is found to be greater in absolute value for buy transactions
five days prior to the event and during the three day event window. Unlike the main
regression, no significant difference from zero can be found for buy transactions during the
five days after the event. This is because the CAR is greater in the regression excluding the
group Other, but is still very close to zero. During the three day CAR for sell transactions, the
CAR is found to be greater in absolute value in the three day event window but less in
absolute value for the five days after the event. Finally, even though there are some
differences between the samples, they are deemed to be small enough to not alter the results
of the study.
4.4.4 Clustered Standard Errors
Illustrated in Appendix 8 is the results from regressing CAR while clustering standard errors
on firms as opposed to utilizing robust standard errors which is the method used throughout
the study. The results show that model (1) is significant at a 10 percent level, model (5) at a 1
percent level and model (6) at a 5 percent level. This can be compared to the regression
model presented in Table 4.1 where it is concluded that all CAR except model (4) is
significantly different from zero at a 1 percent significance level. The choice of standard
errors can therefore be said to affect the results of this study. As previously discussed, using
clustered standard errors is not the preferred choice in this study as a result of the structure of
the data where several observations per time unit can occur.
35
5. Analysis & Discussion
5.1 The Profitability of Insider Trading
In line with previous studies defining the event date as the publication date (De la Brunière et
al., 2020; Chang et al., 1998; Wang et al., 2012) cumulative abnormal returns can be found in
the short term for both buy and sell transactions. This indicates an ability for outside investors
to mimic insiders as an investment strategy to generate abnormal returns. The strong-form
efficient market hypothesized by Fama (1970) does not hold as both insiders, and to a certain
degree outsiders mimicking insiders, are able to generate abnormal returns with the use of
inside information. It is rarely the case that markets are completely strong-form efficient. Sha
et al. (2020) found strong form efficiency in their study of the Chinese market but Mazza and
Wang (2021) found that insiders were able to generate abnormal returns in the same market.
Our research falls in line with De la Brunière et al. (2020) who found semi-strong form
efficiency. Both insiders and outsiders were able to generate abnormal returns, but outsiders
were only able to take advantage of this for a limited period of time as the cumulative
abnormal return was positive during the three day event window but negative, or very close
to zero, the five days following the event. This suggests that, even though insiders and
outsiders were able to take advantage of the situation, the market corrected itself quickly.
Some previous studies (Degryse et al., 2013; Tavakoli et al., 2012) have found greater
abnormal returns for buy transactions than for that of sell transactions. Our findings are in
line with another group of studies (Cheuk et al., 2006; Van Geyt et al., 2013) who show that
sell transactions generate greater abnormal returns in absolute value for outsiders. One
potential explanation to this phenomena, proposed by Cheuk et al. (2006), is that there are far
fewer sell transactions than buy transactions in the period studied. A sell transaction might
therefore stand out more to outside investors who believe that the reason behind the sale is
valuable private information.
The sample of buy transactions finds significant negative abnormal returns prior to the event
and the opposite for sell transactions, indicating a level of market timing on that of the
insider. This means that the insider demonstrates an ability to buy on a low and sell on a high.
Our results are consistent with other studies (Mazza & Ruh, 2022; De la Brunière et al.,
2020) who conclude that insiders do in fact possess a market timing ability. In our case, the
insider purchase generates a slight upward trajectory and is able to generate abnormal returns
for up to five days after the event. However, this is not true for outsiders. During the date of
the publication, there is a perceived reaction to the announcement and significant cumulative
abnormal returns can be found in the short term. However, looking at outsiders' ability to
mimic insiders over five days after the event, negative cumulative abnormal return is found.
This indicates that the market reaction happens prior to the date of the publication and closer
36
to the date of the transaction. A potential explanation to this is that outsiders observe an
abnormal trading volume or a significant buy (sell) order during the transaction day and react
to this by buying (selling) shares within the given company. This allows the insider to earn
abnormal returns while the outsider can only benefit in the short term.
Previous studies (Seyhun, 1998; Betzer & Theissen, 2009; Cheuk et al., 2012; Davis et al.,
2017; Lakonishok & Lee, 2001; Sha et al., 2020) have found that controlling for other firm
specific characteristics has had an effect on the abnormal return that can be generated by
insider trading. Several studies (Seyhun, 1998; Betzer & Theissen, 2009; Cheuk et al., 2012;
Davis et al., 2017; Lakonishok & Lee, 2001) found that size has a significant impact on the
abnormal return, where a negative relationship exists. Smaller companies are more likely to
experience abnormal returns for both buy and sell transactions, partly due to the greater level
of information asymmetry between insiders and outsiders. The argument made is that bigger
companies are under greater scrutiny and have a harder time masking their actions and
performance. For smaller companies receiving less attention, this is fairly simple. Thus, the
reactions to trades in smaller companies are more severe as it might be unexpected.
Our results show that size has a significant impact on the cumulative abnormal return in all
event windows for buy transactions. However, unlike previous studies, our findings show that
the relationship between cumulative abnormal return and size is negative for the window
before the event and during the three day event window, but that it is positive during the five
days after the event. This suggests that smaller companies in our sample have greater
reactions to the buy transaction prior to, and during, the event. The result is in line with
previous studies (Seyhun, 1998; Betzer & Theissen, 2009; Cheuk et al., 2012; Davis et al.,
2017) and could be explained by the level of information asymmetry present between smaller
companies and outsiders. However, unlike the studies mentioned we find a significant
positive effect for size five days after the event. This would suggest that while the effect for
smaller companies is greater in the short run, the effect for bigger companies is less volatile
and more sustained. When the bigger companies have abnormal returns, it is less likely that
investors are able to sell quickly at a profit as the prices return to the expected level rapidly.
These results, while different from previous studies, could possibly be explained by the same
thing, i.e. information asymmetry. In smaller companies where the level of transparency is
lower, outsiders value the information received from insider trades higher. A purchase or sale
can be interpreted as a signal of a major forthcoming event. For larger companies, events of
this size are generally harder to mask as the scrutiny they are under is much greater. If an
outside investor purchases a share in a given company based on the actions of insiders, then
the hope is that the purchase by the insider is based on things trending upwards. If the outside
investor now finds out that this is not indeed the case or other investors believe that the price
has reached its peak, they might be inclined to dispose of their holdings. On the contrary, this
37
would rarely be true for larger companies as the transparency and level of visibility for these
companies is way larger. For sell transactions, however, a negative relationship exists before
the event but a significant positive relationship is observed during and after the event. This
implies that investors are also able to earn the greatest abnormal returns trading smaller firms
for sell transactions. This is likely due to the same reasons as for the buy transactions, i.e.
information asymmetry. What could also be contributing to the greater abnormal returns in
absolute terms for smaller companies is the sensitivity of investors. Investors generally have a
stronger relationship with larger companies that have formed over many years. These
companies are also known to be less volatile and more sustainable than smaller companies
are. This may result in investors being more likely to sell when an insider in a smaller
company is selling as compared to a larger company as other factors are considered.
Seyhun (1998) found that the Price-to-Earnings ratio has a negative effect on abnormal
returns for insider trading. Lower P/E firms are more likely to generate greater abnormal
returns than higher P/E firms. The natural inference in this case is that it is likely that outside
investors are going to believe that firms that are already perceived as being “overpriced” are
going to be growing a lot more. Unlike Seyhun (1998), our study finds that the P/E ratio has
no significant impact on the cumulative abnormal returns for buy transactions. However, for
sell transactions the relationship is significant and positive five days after the event. This
implies that for this event window, higher Price-to-Earnings firms are associated with lower
abnormal returns. This indicates that our results do not follow that of Seyhun (1998).
Turning to leverage, previous studies (Sha et al., 2020; Betzer & Theissen, 2009) have found
that leverage and cumulative abnormal return have a negative relationship for purchases and
positive for sales. In our sample, the buy transactions have no significant relationship with
Debt-to-Equity, which is the proxy used for leverage in this study. For sell transactions, a
partially significant relationship is present during the three day event window. This
relationship is positive during the event, which indicates that the more the company is funded
by debt the less likely a shareholder is going to be able to generate abnormal returns by
selling shares in the given company. This supports the notion of Aksu and Kosedag (2006),
namely that higher leverage is associated with lesser abnormal return. However, as only one
event window is partially significant, we conclude that it lacks effect in most cases.
For Return on Assets, significance is found for both buy and sell transactions. Previous
studies (Davis et al. 2017; Betzer & Theissen, 2009; Sha et al., 2020) have found a negative
relationship between profitability metrics, such as ROA and Return On Equity (ROE), and
abnormal returns generated by insider buy and sell transactions. This is not the case in our
study, where the only negative significant relationship that is found is the five days before the
event for buy transactions. This suggests that prior to the event, higher ROA stocks were
performing worse than lower ROA stocks. A high ROA is often related to a company in a
38
growth phase as the more seasoned companies often have a lower but more stable ROA. For
sell transactions, ROA and cumulative abnormal return has a negative significant relationship
after the event. This means that when insiders sell, higher ROA stocks experience a more
negative reaction than that of lower ROA stocks. This could be due to the fact that higher
ROA companies are often smaller and less established and thus more volatile. This suggests
that investors in these types of companies take more consideration to an insider trade and
consider it as a warning signal.
5.2 Differences In Profitability Between Insider Positions
Previous studies investigating whether the role an insider holds within a company, and the
access to information that follows, affects the ability to generate abnormal returns differ in
their results. While some found this to be true (Lin & Howe, 1990; Mazza & Ruh, 2022; De
la Brunière et al., 2020; Seyhun, 1986), others have found no support for this notion (Betzer
& Theissen, 2009; Fidrmuc et al., 2006). Our findings show that some insider groups do in
fact perform better than others when engaging in buy transactions. This is in accordance with
the Informational Hierarchy Hypothesis which states that greater access to information can
lead to greater abnormal returns. For sell transactions, our results indicate that there is no
significant difference between the roles. In contrast to the Informational Hierarchy
Hypothesis, the difference in access to critical information seems to have no impact.
When examining the results of each role in isolation, CFO has the highest cumulative
abnormal return for the main event window for buy transactions followed by Board and CEO.
The CFOs also experienced the most negative abnormal return for the three day sell window.
This shows that on average, CFOs in this sample perform better than the other groups. There
is however no statistically significant difference in the means of cumulative abnormal return
between CFOs, CEOs and board members which entails that it could be a coincidence, or that
other factors are influencing this difference.
When comparing the means in cumulative abnormal returns between CEOs, CFOs and board
members, there are no significant differences in the three day event window for buy or sell
transactions. There are however some significant differences in the period leading up to the
event as well as the following five days. One group that stands out is Other, where lower
level employees, persons related to the insider and other companies are included. This is the
only group who did not experience any significance for the three day buy window. When
comparing the means in cumulative abnormal returns of CEOs, CFOs and board members to
the group Other, there are significant differences between all of them. Further, when
excluding the group Other from the sample, the cumulative abnormal return for the three day
buy event window was significantly higher. This result could be explained by the
Informational Hierarchy Hypothesis, namely that differences in access to vital information
39
impacts the possibility to generate abnormal returns. As CEOs, CFOs and board members can
be assumed to have greater insight in the company than the parties included in the group
Other, this explanation seems plausible.
Some authors have argued that the scrutiny top executives are under can restrict them in their
trades and lower the excess returns generated (Knewtson & Nofsinger, 2014; Fidrmuc et al.,
2009). This is not entirely in line with the results in this study where top executives such as
the CEO and CFO on average perform equally well, if not better, than board members who
presumably receive less attention. There is however a possibility that the top executives
would perform even better if not monitored as closely, which was the case for CFOs before
the introduction of the Sarbanes-Oxley Act. One result that is in line with Knewtson and
Nofsinger (2014) is that the CFOs generate the greatest excess return on average. This can
however not be said with statistical significance as no such significance was found when
comparing the means in cumulative abnormal returns of CFOs to that of CEOs and board
members.
40
6. Conclusion
Ever since the opening of the first stock market, investors have been trying to find ways to
outperform the market. In a time where this is becoming increasingly hard, investors try to
utilize all the information they can get their hands on in order to make more informed
decisions. As the current EU regulation on market abuse requires insiders to report and
publish all transactions within three business days, it presents an opportunity for outside
investors to reduce the information asymmetry between the parties.
Our sample consists of 6 550 trades performed by insiders, and related persons, in Swedish
firms listed on OMX Stockholm between 17 September 2018 and 3 February 2023. Using an
event study methodology, we aim to investigate whether outside investors can earn abnormal
returns by mimicking insiders and if the position of the insider affects the magnitude of the
returns.
Our results indicate that on average, it is possible for an outside investor to generate abnormal
returns by simply mimicking an insider’s buy or sell transactions. For buy transactions,
outsiders are able to earn abnormal returns in the short term but are unable to keep this up for
the days following the event. However, for sell transactions the outsider is able earn abnormal
returns both in the short term and five days after the event. We find that the market reaction to
the event occurs during the transaction date and not during the publication date. We argue
that outside investors observe abnormal trading activity, and take action accordingly. An
outsider mimicking an insider on the date of the publication only earns a fraction of the return
generated by the insider. These results are somewhat affected by the control variables. Size is
found to have the most impact where smaller companies generated greater abnormal returns.
Additionally, our results show that the position of the insider does impact the ability to
generate abnormal returns when buying shares. The CEOs, CFOs and board members
experienced significantly higher returns than the group Other which was the only group that
did not generate a positive abnormal return. These results can be explained by the
Informational Hierarchy Hypothesis as top executives and board members generated greater
abnormal returns. For the sell transactions, there is no significant difference between the
groups meaning that for these transactions, the difference in access to critical information
seems to lack impact.
While we have shown that it is possible for outside investors to generate abnormal returns by
mimicking insiders, the viability of using it as an investment strategy remains to be discussed.
Our results show that for buy transactions, only one out of eleven days experienced a positive
average abnormal return. This imposes that the investor would have to be very active and
closely monitor the market in order to time this window of opportunity which is time
consuming. Active trading strategies are often associated with a higher risk. A high frequency
41
of trades also imposes high transaction costs in the form of commissions and the bid-ask
spread. Whether this constitutes a viable investment strategy depends on the investor’s risk
profile and time constraints. An option to this active strategy could be investing in a fund
with the strategy of mimicking insider trades. These types of funds have seen mixed results,
but could be an option for investors who believe in the sentiment but strive for a more passive
alternative.
Our findings contribute to the existing literature on insider trading. While there are many
studies exploring the market timing ability and performance of corporate insiders, there are
far fewer who approach the topic from an outsider perspective. This study shows that there
are significant differences between trading on the transaction day and the publication day.
Despite the return being negatively affected by the time between these days, outsiders were
still able to capitalize on the events in the short run. In addition to this, no published studies
investigating outsiders’ ability to earn abnormal returns by mimicking insiders in a Swedish
setting were found. Our findings further contribute to prior research by showing that there are
differences in performance between insider groups. This constitutes a vital piece of
information for investors seeking to adopt insider mimicking as an investment strategy.
One limitation of this study is the lack of hourly stock price data. The result of this is that a
trade occurring midday in an open market becomes hard to match to the correct price.
Measures were taken to correct for this but it is likely that the effects were diminished. One
aspect that this study does not answer is the relationship between lower level employees and
top executives as the category Other includes related persons, executives from subsidiaries
and more. Therefore, a suggestion for further research is to examine this relationship more
closely. Another suggestion for further research is to perform a similar study but using
holding periods of several months. While our study is focused on the short-term effects, a
study observing the abnormal returns over several months would add to the understanding of
the viability of using insider mimicking as an investment strategy. As company structures
vary across different industries, both in terms of financing and assets, future research could
also include an industry variable to control for differences between them.
42
References
ACCA Global. (n.d.). Market matters. Accessed 2023-02-12 from:
https://www.accaglobal.com/gb/en/student/exam-support-resources/fundamentals-exams-stud
y-resources/f9/technical-articles/market-matters.html
Adams, J. C., Hayunga, D. K., Mansi, S., Reeb, D. M. & Verardi, V. (2019). Identifying and
Treating Outliers in Finance. Financial Management, 48(2), 345-384.
https://doi.org/10.1111/fima.12269
Ahern, K. R. (2017). Information networks: Evidence from illegal insider trading tips.
Journal of Financial Economics, 125(1), 26–47.
https://doi.org/10.1016/j.jfineco.2017.03.009
Akerlof, G. (1970). The Market for “Lemons”: Quality Uncertainty and the Market
Mechanism. The Quarterly Journal of Economics, 84(3), 488-500.
https://doi.org/10.2307/1879431
Aksu, M. & Kosedag, A. (2006). Transparency and Disclosure Scores and their Determinants
in the Istanbul Stock Exchange. Corporate Governance An International Review, 14(4),
277-296.
https://doi.org/10.1111/j.1467-8683.2006.00507.x
Alpha Wealth Funds. (2022). The Insider Fund. Accessed 2023-03-10 from:
https://www.theinsidersfund.com/wp-content/uploads/2023/01/Insiders-Fund-Summary_2001
-to-Dec2022.pdf
Armitage, S. (1995). EVENT STUDY METHODS AND EVIDENCE ON THEIR
PERFORMANCE. Journal of Economic Surveys, 9(1), 25–52.
https://doi.org/10.1111/j.1467-6419.1995.tb00109.x
Betzer, A., & Theissen, E. (2009). Insider Trading and Corporate Governance: The Case of
Germany. European Financial Management, 15(2), 402-429
https://doi.org/10.1111/j.1468-036X.2007.00422.x
Carhart, M. M. (1997). On Persistence in Mutual Fund Performance. Journal of Finance,
52(1), 57-82.
https://doi.org/10.2307/2329556
Catalyst Funds. (2022). Catalyst Insider Buying Fund. Accessed 2023-03-10 from:
https://catalystmf.com/docs/factsheets/INS/factsheet.pdf
Chang, S., & Suk, D. Y. (1998). Stock Prices and the Secondary Dissemination of
Information: The Wall Street Journal’s “Insider Trading Spotlight” Column. The Financial
Review, 33(3), 115–128.
https://doi.org/10.1111/j.1540-6288.1998.tb01386.x
Cheuk, M., Fan, D., & So, R. (2006). Insider trading in Hong Kong: Some stylized facts.
Pacific-Basin Finance Journal, 14(1), 73-90.
https://doi.org/10.1016/j.pacfin.2005.06.002
Asness, C. & Frazzini, A. (2013). The Devil in HML‛s Details. Journal of Portfolio
Management, 39(4), 49-68.
https://doi.org/10.2139/ssrn.2054749
Cohen, L., Malloy, C., & Pomorski, L. (2012). Decoding Inside Information. The Journal of
Finance, 67(3), 1009–1043.
https://doi.org/10.1111/j.1540-6261.2012.01740.x
Correia, S. (2015). Singletons, Cluster-Robust Standard Errors and Fixed Effects: A Bad Mix.
Davis, F., Taghipour, B., & Walker, T. J. (2017). Insider trading surrounding securities class
action litigation and settlement announcements. Managerial Finance, 43(1), 124–140.
https://doi.org/10.1108/MF-05-2016-0129
Degryse, H. de Jong, F., & Lefebvre, J. (2014). An Empirical Analysis of Legal Insider
Trading in the Netherlands. De Economist, 162(1), 71-103.
https://doi.org/10.1007/s10645-013-9222-4
De La Brunière, S.N., Haye, J. C., & Mazza, P. (2020). The performance of corporate legal
insiders on the French stock market. International Review of Law and Economics, 61(C).
https://doi.org/10.1016/j.irle.2019.105880
DIRECTIVE 2003/6/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL.
https://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2003:096:0016:0025:en:PDF
Eckbo, B. E., & Smith, D. C. (1998). The Conditional Performance of Insider Trades. The
Journal of Finance, 53(2), 467–498.
https://doi.org/10.1111/0022-1082.205263
Fama, E. F. (1970). Efficient Capital Markets: A Review of Theory and Empirical Work. The
Journal of Finance, 25(2), 383-417.
https://doi.org/10.2307/2325486
Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds.
Journal of Financial Economics, 33(1), 3–56.
https://doi.org/10.1016/0304-405X(93)90023-5
Fidrmuc, J. P., Goergen, M., & Renneboog, L. (2006). Insider Trading, News Releases, and
Ownership Concentration. The Journal of Finance, 61(6), 2931–2973.
https://doi.org/10.1111/j.1540-6261.2006.01008.x
Finansinspektionen. (2021). Choice of transaction category when reporting PDMR
Transactions. Accessed 2023-03-18 from:
https://www.fi.se/contentassets/395fd56d8ecc4dda968f17e073f572a6/lathund-transaktionskat
egorier-2021-02-17-engn.pdf
Finansinspektionen. (2022). PDMR transactions. Accessed 2023-03-18 from:
https://www.fi.se/en/markets/investors/pdmr-transactions/
Government Offices of Sweden. (2015). Swedish Financial Supervisory Authority
(Finansinspektionen). Accessed 2023-01-24 from:
https://www.government.se/government-agencies/swedish-financial-supervisory-authority-fin
ansinspektionen-/
Jegadeesh, N. & Titman, S. (1993). Returns to Buying Winners and Selling Losers:
Implications for Stock Market Efficiency. The Journal of Finance, 48(1), 65-91
https://doi.org/10.2307/2328882
Knewtson, H.S., & Nofsinger, J.R. (2014). Why are CFO insider trades more informative?.
Managerial Finance, 40(2), 157-175.
http://dx.doi.org/10.2139/ssrn.1800202
Lakonishok, J., & Lee, I. (2001). Are insider trades informative? Review of Financial Studies,
14(1), 79–111.
https://doi.org/10.1093/rfs/14.1.79
Lakonishok, J., Shleifer, A., & Vishny, R. W. (1994). Contrarian Investment, Extrapolation,
and Risk. The Journal of Finance, 49(5), 1541–1578.
https://doi.org/10.2307/2329262
Lamba, A.S., & Khan, W.A. (1999). Exchange listings and delistings: the role of insider
information and insider trading. Journal of Financial Research, 22(2), 131–146.
https://doi.org/10.1111/j.1475-6803.1999.tb00719.x
Lin, J., & Howe, J. A. (1990). Insider Trading in the OTC Market. The Journal of Finance,
45(4), 1273–1284.
https://doi.org/10.1111/j.1540-6261.1990.tb02436.x
Lynch, P., & Rothchild, J. (2000). One Up On Wall Street: How To Use What You Already
Know To Make Money In The Market. Penguin Books.
MacKinlay, A. C. (1997). Event studies in economics and finance. Journal of Economic
Literature, 35(1), 13-39.
Mazza, P., & Ruh, B. (2022). The performance of corporate legal insider trading in the
Korean market. International Review of Law and Economics, 71.
https://doi.org/10.1016/j.irle.2022.106076
Mazza, P., & Wang, S. (2021). Corporate legal insider trading in China: Performance and
determinants. International Review of Law and Economics, 68.
https://doi.org/10.1016/j.irle.2021.106024
Meulbroek, L. K. (1992). An empirical analysis of illegal insider trading. The Journal of
Finance, 47(5), 1661–99.
https://doi.org/10.2307/2328992
Moser, B.K., & Stevens, G.R. (1992). The American Statistician, 46(1), 19-21.
https://doi.org/10.2307/2684403
Myers, S. C. & Majluf, N. S. (1984). Corporate financing and investment decisions when
firms have information that investors do not have. Journal of Financial Economics, 13(2),
187-221.
https://doi.org/10.1016/0304-405X(84)90023-0
Nasdaq. (n.d.-a). Rules & Regulations. Accessed 2023-01-25 from:
https://www.nasdaq.com/market-regulation/nordics/stockholm
Nasdaq. (n.d.-b). First North Growth Market. Accessed 2023-03-07 from:
https://www.nasdaq.com/solutions/nasdaq-first-north-growth-market
Nasdaq Nordic. (2023). Historical Prices. Accessed 2023-03-08 from:
https://www.nasdaqomxnordic.com/index/historiska_kurser?Instrument=SE0000337842
REGULATION (EU) No 596/2014 OF THE EUROPEAN PARLIAMENT AND OF THE
COUNCIL. https://eur-lex.europa.eu/eli/reg/2014/596/oj
SEC. (n.d.). Insider Trading Policy. Rule 10b-5 Prohibition on Insider Trading. Accessed
2023-01-25 from:
https://www.sec.gov/Archives/edgar/data/1164964/000101968715004168/globalfuture_8k-ex
9904.htm
Seyhun, H. N. (1986). Insiders’ profits, costs of trading, and market efficiency. Journal of
Financial Economics, 16(2), 189–212.
https://doi.org/10.1016/0304-405X(86)90060-7
Seyhun, H. N. (1998). Investment intelligence from insider trading. Cambridge, Mass. MIT
Press.
SFS 2000:1087. Lag om anmälningsskyldighet för vissa innehav av finansiella instrument.
https://www.riksdagen.se/sv/dokument-lagar/dokument/svensk-forfattningssamling/lag-2000
1087-om-anmalningsskyldighet-for-vissa_sfs-2000-1087
SFS 2016:1306. Lag med kompletterande bestämmelser till EU:s
marknadsmissbruksförordning.
https://www.riksdagen.se/sv/dokument-lagar/dokument/svensk-forfattningssamling/lag-2016
1306-med-kompletterande-bestammelser_sfs-2016-1306
SFS 2018:331. Lag om ändring i lagen (2016:1306) med kompletterande bestämmelser till
EU:s marknadsmissbruksförordning.
https://svenskforfattningssamling.se/doc/2018331.html
Sha, Y., Zhang, Z., & Liu, L. (2020). Is Illegal Insider Trading a Sure Thing? Some New
Evidence. Emerging Markets Finance and Trade, 56(12), 2929–2944.
https://doi.org/10.1080/1540496X.2020.1813469
Shrestha, N. (2020). Detecting Multicollinearity in Regression Analysis. American Journal of
Applied Mathematics and Statistics, 8(2), 39-42.
https://doi.org/10.12691/ajams-8-2-1
Tavakoli, M., McMillan, D., & McKnight, P. J. (2012). Insider trading and stock prices.
International Review of Economics & Finance, 22(1), 254–266.
https://doi.org/10.1016/j.iref.2011.11.004
Tavakoli, M., McMillan, D., & McKnight, P. J. (2014). The Credit Crunch and Insider
Trading. Financial Markets, Institutions and Instruments, 23(2), 71–100.
https://doi.org/10.1111/fmii.12015
Van Geyt, D., Van Cauwenberge, P., & Bauwhede, H. V. (2013). The Impact of the Financial
Crisis on Insider Trading Profitability in Belgium. Journal of Business Economics and
Management, 14(2), 364-385.
https://doi.org/10.3846/16111699.2011.652980
Wang, W., Shin, Y. C., & Francis, B. B. (2012). Are CFOs’ Trades More Informative Than
CEOs’ Trades? Journal of Financial and Quantitative Analysis, 47(4), 743–762.
https://doi.org/10.1017/S0022109012000257
Appendix 1 - Winsorization
Table A Pre-winsorized Variables
Table B Winsorized Variables
Appendix 2 - Correlation Matrix
Appendix 3 - Annualized CAR
The cumulative abnormal return in this appendix is calculated using the following equation:
𝐴𝑛𝑛𝑢𝑎𝑙𝑖𝑧𝑒𝑑 𝐶𝐴𝑅 =  (1 + 𝐶𝐴𝑅)(𝑇𝑟𝑎𝑑𝑖𝑛𝑔 𝐷𝑎𝑦𝑠/𝐷𝑎𝑦𝑠 𝑖𝑛 𝐸𝑣𝑒𝑛𝑡 𝑊𝑖𝑛𝑑𝑜𝑤) − 1
where there are an estimated 252 trading days in a year.
Appendix 4 - Differences In Means Between Insiders
Appendix 5 - Transaction Date
Table C Test for difference in means, Transaction vs. Publication
Table D Test for the significance of CAR, Transaction date
Appendix 6 - Market Adjusted Model
Table E Test for difference in means, Four Factor Model vs. Market Adjusted Model
Table F Test for the significance of CAR, Market Adjusted Model
Appendix 7 - Excluding Others
Table G Test for difference in means, Including vs. Excluding Other
Table H Test for the significance of CAR, Excluding Other
Appendix 8 - Clustered Standard Errors
Appendix 9 - Python Code