Profit warning impact on stock prices

Get Complete Project Material File(s) Now! »

Method

This section describes the chosen methods and discusses possible drawbacks with each of these.

Quantitative vs. qualitative approach

Depending on the purpose of the thesis, different methods, in terms of number of obser-vations, are appropriate (Jankowicz, 1991). The primary two methods to choose from are qualitative and quantitative. The former is characterized by a few, but thoroughly processed, numbers of observations which usually take the shape of in depth interviews. A qualitative study is inductive in nature as general theories are constructed from a small number of observations (Darmer & Freytag, 1995). A typical drawback associated with a study conducted using a qualitative approach is that the inferences drawn based on the em-pirical findings are subject to subjective interpretation of the researcher as well as the inter-viewee. Hence the conclusions can be colored due to the personal characteristics, such as previous experience and knowledge of the researcher. The same can be said about the in-terviewees, which have different perspectives that can cause different interpretations of the questions (Ghauri & Grønhaug, 2005). A possible implication of a qualitative study can thus be that depending on who is conducting the research, different conclusions might be made i.e. the reliability of the study might be compromised.
A quantitative study on the other hand, is based on a large(r) number of observations with the objective to draw inferences applicable to a wider population (Bryman, 2004). Thus, it is necessary that the sample both consists of a sufficient number of observations as well as covering all facets of the population in order to adequately represent the population as a whole. Due to the many observations, a quantitative sampling procedure is analyzed by us-ing statistical measurements. In contrast to a qualitative approach, which is often inductive in nature, a quantitative is usually deductive in the sense that large samples are used to veri-fy or reject a specific theory (Darmer & Freytag, 1995).
As far as this study is concerned, a quantitative approach is most appropriate considering the purpose of this thesis:
“…to investigate if there exist non reported insider activity prior to profit warnings on the Stockholm Stock Exchange”.
Since the objective of the thesis is to draw conclusions that can be applied to a wider group than the chosen observations, the sample must be both large and representative enough to reflect the Stockholm Stock Exchange as a whole. In addition, the authors are of the belief that due to the sensitive nature of the subject this thesis aims to investigate, interviews might both be difficult to arrange since potential interviewees may be reluctant to partici-pate and should an in interview take place, personal emotions regarding insider trading might for example either cause an exaggeration of the insider activity or the complete op-posite, a complete denial. It seems probable that one in favor of insider trading (and per-haps engages in such activity him/herself) might tend to deny its existence in order steer off attention and vice versa. Furthermore, even though the interviewees might be credible by virtue of their profession or position in a company, the authors feel that the delicate na-ture of investigated area needs to be supported by legal documents in order to strengthen the trustworthiness of the information obtained during the interview. As a documented support of the accusations of insider trading discussed with an interviewee might be diffi-cult to obtain, the authors feel the need to circumvent this obstacle. With the, by the authors’, perceived complications of personal interviews in mind, the purpose of the thesis can best be fulfilled by a quantitative approach with which statistical models are used to measure insider trading activity on the Stockholm Stock Exchange, thus avoiding the prob-lems associated with personal interviews.

Conducting our study

The following section aims at in detail explaining how the study will be executed and pro-vides the reader with considerations that were made. It will start off with how the data was gathered and then continue with the event study specifications including abnormal return estimations. Finally some critic and concerns of the chosen method will be discussed.

Event study specification

This paper aims at studying the extent of insider trading on the Stockholm Stock Exchange rather than effect of profit warnings per se. However in order to do an event must found where insider trading can be isolated i.e. an event where the release date is unknown to the public. For example using earnings announcement and finding pre-announcement drift might indicate insider trading activity (Damodaran, 2002) but, since the date of the earnings announcement is known on beforehand, potential pre-announcement drift can not be ex-plained simply with insider trading. This is due to that investors might be speculating on the outcome on the announcement day. As a result, profit warnings serves as an appropri-ate event since these should be unknown before they are announced to the market. This study will thus use profit warnings to find insider trading prior to them by testing for pre-announcement drift. However, using this method where the population is defined as all profit warnings ever to occur and the approximately 30 observations included in the study constitute the sample, limits the ability to generalize the findings to profit warnings. Thus, should any statistical evidence support the existence of abnormal negative returns which in turn indicates insider trading, those inferences can only be said to hold true in relation to profit warnings. Nevertheless the authors find it reasonable to assume that, if insider trad-ing exist prior to profit warnings it might just as likely occur prior to any other stock price moving information announcement, such as an quarterly report. This notion will be dis-cussed in more depth in the Discussion section.
In order to measure the effects of the event, the profit warning it must be extended into an event window. Since the aim of this paper is to study insider trading by examining pre-announcement drift there is no need to include any days subsequent to the event in the event window and the test period. In order to find the whole effect of the potential insider trading the authors have decided to use a window of 30 days preceding the event. There are two reasons for this length. First of all, insiders, as defined earlier, are not allowed to trade shares in their respective companies later than 30 days before earnings announcements (In-sider law 2000:1087) and second, using a thirty day long event window allows for running statistical tests on the entire period for an individual company since the period includes 30 observations. Even though the data itself may not be strictly normally distributed, the sam-ple size of 30 observations is sufficient enough to assume that the sample parameters are normally distributed under the central limit theorem approach (Azcel, 2002).
Incorporating the event and the estimation period into the model it will look like figure 3-1 below:
Where:
t1 to t2 = the two year period during which beta is estimated
t-30 to t-1 the thirty day period during which the abnormal returns are estimated t0 = the profit warning day

Data gathering

When conducting the data gathering approximately 40 warnings were found for the speci-fied period, 2004-2006. Unfortunately due to the inadequacy of the databases from which stock prices were obtained, less than thirty of the observations remained for statistical test-ing. The effect of this forced an extension of the initial time frame to also include 2003. However only the missing two observations that were required to construct a total sample of thirty were taken from this year. As a result year 2003 is not fully represented in the sample.
The sample do not allow for tests that take into account characteristics such as company size, business field or time period. Unfortunately an involuntary partial division of the ob-servations occurred as exactly 50% of the companies are from the large cap. list while the remaining 50% are spread between the mid- and small cap. lists. Large cap. companies are thus somewhat overrepresented in the sample but no special attention will be paid to this fact though. The uneven distribution of company size however, is not so much due to that larger companies in general tend to adjust forecasts but rather a result of that profit warn-ings for which data were unavailable, and thus had to be abandoned, were almost entirely from small cap. companies. A detailed description of companies included can be found in Appendix 1 and the data and test can be found in Appendix 2.
The sources that are used to gather the relevant data are summarized in the table below.

Beta and abnormal return estimations

Since the objective with an event study is to isolate the effect of the event, profit warnings, effects of general stock movements must be removed. In order to achieve this, CAPM will be used to estimate the expected return which is then subtracted from the actual return – leaving the abnormal return.
In order to estimate the expected return, betas for every stock in the sample are needed. The market term in CAPM is in fact an imaginary creation of all stocks but usually a major stock index is used as a proxy. Thus, in this case, OMXSPI, the all share index of the Stockholm Stock Exchange, is used as a proxy for the overall market return.
To obtain the beta for each stock a number of regressions, using daily returns, will be run. A common procedure (Damodaran, 2002), that will also be used in this study, is to use two years of daily trailing returns to estimate the beta values from the following regression:
Ri = ai + β i Rm  + ε i
Where:
Ri = Return on stock i
α i = Alpha value stock i
β i = Beta value stock i
Rm = Return on OMXSPI
ε i = Random term stock i
Since it is important to separate the estimation period from the event and the event win-dow, the last observation in the time series used to estimate beta, is the last trading day pre-ceding the event window i.e. the 30 trading days leading up to the profit warning. As a re-sult, beta for a firm making a profit warning on 1st September will be estimated using two years of daily returns up until the last trading day before the 30 trading days period prior to 1st September.
After finding betas for all firms in the sample the abnormal returns can be estimated. When performing the estimations, the proxy used for the risk free rate is the twelve month Swe-dish treasury note. The yearly rate for each day was collected and subsequently adjusted to a daily return. A fixed income instrument is more desirable than longer duration bonds to use as proxy for the risk free rate (Damodaran, 2002). The return for the market portfolio is estimated using the OMXSPI index. All abnormal returns are calculated for each of the 30 trading days preceding the event. Thus, the return will be calculated on a daily basis, us-ing the twelve month treasury note adjusted to daily returns, daily return of the stock and the market portfolio and finally the corresponding beta for that particular firm.
Abnormal return can thus be defined as follows:
Equation 3-1 Abnormal return
α i ,t = Ri ,t − E Ri ,t 
Where:
α i ,t = Abnormal return on stock i at time t
Ri ,t Actual return on stock i at time t
E ( Ri ,t )  Required return stock i at time t
And as derived from the CAPM:
Equation 3-2 Abnormal return derived from CAPM
E Ri ,t  = R f ,t + β i Rm,t − R f ,t 
Where:
R f ,t = Daily risk free return at time t
β i = Beta value stock i at quarter q5
Rm,t = Daily return on market portfolio at time t

Estimating cumulative abnormal returns

The event window in this study extends 30 trading days prior to the warning. Thus, in or-der to investigate the impact on returns prior to the event, daily abnormal returns for any given stock must be summed for the entire period. In general, cumulative abnormal returns are calculated in the following way:
Equation 3-3 Cumulative abnormal return
m
CARi , m = ∑αi , t
t  j
Where:
CARi ,m = Cumulative abnormal return for stock i during the [j,m] time period
However, only looking at whether one specific company’s returns are abnormal does not say if the same negative price fluctuation occurs in relation to profit warnings in general. Therefore it is necessary to include at least thirty warning observations in the study. In ad-dition, only testing for one company’s returns do not circumvent firm specific aspects, such as other price driving news. It can thus be difficult to explain the negative drift exclusively with insider trading as it can have been caused by other factors. Insider activity is still though, the most plausible explanation. Over a large sample, firm specific price moving news, good and bad, cancel each other out – leaving insider trading as the only possible ex-planation for the significant negative abnormal return. The authors nevertheless argue that examining drift on an individual company level could be of interest since if indications of insider trading exist, one has an incentive to conduct further research on the company in question, for example looking at the short interest and/or press releases. However due to the limited sample the value of such a test would, in this context, be low and therefore such a test will not be conducted.
To identify any price drift for the sample as a whole, one looks at cumulative abnormal re-turns on a portfolio level. Since this study is only concerned with profit warnings and not so called reversed profit warnings (announcing higher than expected earnings) and that no division of the announcements according to magnitude will be made, there is no need of dividing the firms into more than one portfolio. If the study aimed at studying negative (profit warnings) as well as positive earning forecast adjustments (reversed profit warnings), dividing them into two portfolios, one negative and one positive, would be necessary in or-der for the results not to offset each other in a statistical test.
The cumulative abnormal return for the portfolio can be calculated using the formula be-low:
Equation 3-4 Cumulative average abnormal return
1 n m
CAˆRp , m  ∑∑αi ,t
n
Where:
ˆ = Cumulative average abnormal return for portfolio p during the [j,m] time period
CAR p ,m
n = Number of stocks included in portfolio p

Statistical testing

To be able to draw a valid inference about a population from a sample, it is important that the sample is created in a random manner, otherwise the sample will not be a true represen-tation of the population and any inference drawn can thus be distorted (Azcel, 2002). As mentioned above, the population in this study is defined as all profit warnings. Thus the observed warnings included in the sample must be randomly selected among profit warn-ings in general. The selection of observations i.e. the sampling was however not done ran-domly since the sample constitutes entirely of profit warnings from a limited period with conditions that may differ from those prevailing historically.
T-tests will be used to accept or reject the hypothesis whether the abnormal return are sig-nificantly different from zero and thus ascertain that the abnormal returns are in fact ab-normal. Since the very nature of trading on profit warning information requires either shortening or selling company shares in order to make a profit, the hypothesis is that the abnormal return is significantly less than zero. Considering the purpose of the study, study-ing insider trading prior to profit warnings, statistically significant positive deviations from zero are of no value since such deviations could not be derived from insider trading but ra-ther from the ordinary random walk characterizing an efficient stock market. Based on this reasoning there is no need to use a two-tailed test where both positive and negative deviations are tested for significance but instead a one-tailed test where only lower t-statistic can reject the hypothesis stated below.
When testing on the portfolio level the hypothesis are:
H ˆ ≥ 0
0 : CAR p ,m
ˆ 
H 1 : CAR p ,m0
The test will be carried out on an 95% significance level and with n-1, in this test, 29 de-grees of freedom, the critical value equals 1.6991 for a one-tailed test. In order to reject the null hypothesis the t-statistic must thus be below the negative of this value.

Validity and reliability

The concepts of reliability and validity must be taken into consideration by any researcher who wants to make one’s results credible (Ghauri & Grønhaug, 2005). Therefore, a brief discussion on how this thesis encompasses these important concepts will follow.

Validity

Validity measures how well the findings conform to what the researcher intends (Ghauri & Grønhaug, 2005). This means that the observed measure should equal the true measure. Is-sues that can lower the validity are for example misinterpretations of questions by a res-pondent in questionnaire. This particular issue is not a concern for this paper as it solely re-lies on secondary data. Although, problems that might create validity issues as far as this study is concerned can be minor errors when collecting data due to the considerable amount of different information needed. Possible errors include dates for each profit warn-ing, historical stock prices or index prices. Thus, the collection of these must be conducted with care. Other aspects that can lower the validity of the findings in this paper is that the approach of using an event study to measure insider activity might not be one hundred percent certain. However, considering that the method has been applied for a similar study before (Engert, 2005) and through the reasoning that since profit warnings are completely random events, abnormal returns prior to such a release can only be caused by insider trad-ing or news leakage (Damodaran, 2002). However, these two cannot actually be separated since news leakage, if used without being publicly known, automatically becomes insider trading. Considering these explanations for abnormal return prior to profit warnings the potential validity implications caused by choosing an event study, is by the authors deemed to be entirely non existent. A third aspect to consider concerning potential negative validity implications is the model used for calculating the excepted return. CAPM for example, does not capture all the underlying factors that affect the expected return. On the other hand, the use of credible sources for obtaining the data which includes OMX Group and Affärsdata, contributes to an increased validity.

Reliability

The concept of reliability is concerned with the stability of the measures. It means that, re-gardless of who is conducting the study, the same results should be found provided that the exact same method is used (Ghauri & Grønhaug, 2005).
This thesis, as it is based on secondary data that are readily available to anyone, analyzed by standardized statistical models, can be considered to generate reliable results in the sense that anyone hosting the necessary skills could duplicate the results. However, one must use the exact same procedures, including the same model for estimating expected returns to reach the same conclusions. By employing other models for expected as well as abnormal return calculations, like for instance the market model or the APM instead of CAPM, will generate results that will differ from those found in this thesis where CAPM is used.

Criticism of sources and method

Financial event studies in general are sensitive to the models chosen, especially concerning the expected return and in turn abnormal return estimations. Therefore regardless of the method chosen, the research will always be exposed to the inherent imperfections of the model used. The model chosen for the expected return estimations in this paper is heavily reliant upon CAPM’s ability to accurately determine the expected return which unfortu-nately might cause some distortions in the inferences drawn from the results. CAPM has as its main drawback that certain aspects might not be fully reflected when using it to deter-mine the expected return since, for instance, studies have shown that small firms have been somewhat undervalued with CAPM (Smith & Smith, 2004). In addition, the rather strin-gent assumptions of CAPM seldom hold true in reality. Nevertheless, it is the most widely used model to value companies (Smith & Smith, 2004) which provides some insights to the model’s superiority and accuracy despite its critique. Thus the authors feel confident that the results are both valid and reliable regardless of the mentioned problems associated with CAPM.

Table of contents
1 Introduction
1.1 Background
1.2 Problem
1.3 Purpose
1.4 Research questions
1.5 Delimitations
1.6 Literature search
2 Frame of reference
2.1 Insider trading
2.2 Efficient market hypothesis
2.3 Price drift
2.4 Profit warnings
2.5 Event study
2.6 Expected return estimation
3 Method
3.1 Quantitative vs. qualitative approach
3.2 Conducting our study
3.3 Validity and reliability
3.4 Criticism of sources and method
4 Empirical results and analysis
4.1 Initial remarks
4.2 Descriptive results and analysis
4.3 Statistical testing
4.4 Insider trading law effectiveness
5 Conclusion
6 Discussion
6.1 Profit warning impact on stock prices
6.2 Insider trading pattern and possible differences
6.3 Recommendations for further research
References
GET THE COMPLETE PROJECT