JÖNKÖPI NG UNIVER SITY
D o e s S i z e M a t t e r ?
Abnormal Return and Market Efficiency at Stockholm Stock Exchange
Bachelor thesis within Business Administration Authors: Einarsson, Per
Wännerdahl, Hampus Tutor: Wramsby, Gunnar Jönköping December 2007
Acknowledgements
We would like to express our gratefulness to the following persons:
Gunnar Wramsby, tutor.
For his support during the process of conducting this thesis. For his straightforward and constructive criticism which has lead us on the right track in times of confusion.
Thomas Holgersson, Associate Professor Economics at JIBS. For his guidance in the statistical jungle.
Bachelor Thesis in Business Administration
Title: Does Size Matter? Abnormal Return and Market Efficiency at Stockholm Stock Exchange
Author: Per Einarsson
Hampus Wännerdahl
Tutor: Gunnar Wramsby
Date: 2007-12-20
Subject terms: Buy recommendations, market efficiency, abnormal return, capitali-zation value
Abstract
Background and purpose
In Sweden private savings in stocks has experienced a large increase and in year 2006 there were 6.7 million people, or 77 per cent of the population owning stocks. A recent study shows that more than every other Swede has deficient knowledge in trading with stocks. Since small private investors often do not know how to gather and interpret information they must utilize investment advices. The large increase in private savings in stocks, the lack of investment knowledge together with the large increase in Internet usage has resulted in investment advice seeking on the Internet. One of the largest sources of investment ad-vices on the Internet in Sweden today is Avanza.se. The purpose with our thesis is to de-scribe and analyze if, after a buy recommendation issued at Avanza’s website, the effects with respect to abnormal return and market efficiency differ significantly depending on a company’s capitalization value.
Method
We have used a quantitative approach to fulfill our purpose. The secondary data required to do so was gathered from the OMX-Group’s website, where historical prices and Index information was collected, and from the online broker Avanza’s website where the buy recommendations were compiled. In order to conduct statistical tests and calculations we have used the statistical software SPSS.
Frame of Reference
The theories we made use of mainly treated market efficiency and abnormal return.
Conclusions
We have seen that the recommendations’ effect concerning abnormal return differ signifi-cantly depending on capitalization value, where the effect on companies with smaller capi-talization values are larger. We have also found tendencies of market inefficiency at the semi strong level for stocks with smaller capitalization value.
i
Table of Contents
1
Introduction ... 1
1.1 Problem background ...1 1.2 Problem Discussion...1 1.3 Purpose ...3 1.4 Definitions...31.5 Demarcations of the study...3
2
Method ... 4
2.1 Research Approach...4
2.2 Data Collection ...5
2.3 Selection...5
2.3.1 Why beta-value is not taken into consideration ...6
2.4 Data Diminution ...7
2.5 Sign Test ...7
2.5.1 Sign Test of Abnormal Return ...8
2.5.2 Sign Test of Market Efficiency ...9
2.6 Confidence Bounds for testing Abnormal Return ...9
3
Frame of Reference... 11
3.1 Efficient Capital Markets...11
3.1.1 The Weak Form...12
3.1.2 The Semi strong Form...12
3.1.3 The Strong Form ...12
3.2 Testing the efficiency of the market ...13
3.2.1 Weak form test ...13
3.2.2 Semi strong test ...14
3.2.3 Strong test ...14
3.3 Abnormal Returns in Efficient and Inefficient Markets ...14
3.4 Market Reactions and Firm Size ...16
3.5 The Swedish Stock Market...16
3.5.1 OMX Stockholm All-Share Index (OMXS) ...17
3.6 Information and trading with stocks ...17
3.7 Previous Research ...18
4
Empirical Findings and Analysis ... 20
4.1 Abnormal Return ...20
4.1.1 Abnormal Return at t-1 ...20
4.1.2 Sign Test Abnormal Return at t-1 ...21
4.1.3 Abnormal Return at t ...21
4.1.4 Sign Test Abnormal Return at t ...22
4.1.5 Abnormal Return at t+1 ...23
4.1.6 Sign Test Abnormal Return at t+1 ...24
4.1.7 Abnormal Returns at t+7 ...24
4.1.8 Sign Test Abnormal Return at t+7 ...25
4.1.9 Abnormal Returns at t+30 ...26
4.1.10 Sign Test Abnormal Returns at t+30 ...26
ii
4.3 Sign Test of Market Efficiency ...28
5
Concluding Remarks ... 30
5.1 Critique of the study...31
5.2 Suggestions for further research ...31
iii
Appendices
Appendix 1 Statistical Running of the Stock Recommendations Appendix 2 SPSS Graphs Confidence Bounds
Figures
Figure 2.1 Two-tailed test confidence intervals ...8
Figure 2.2 Confidence bounds (X1 and X2) ...10
Figure 3.1 Relationship between the three types of information (Buckley, Ross, Westerfield & Jaffe, 1998) ...13
Figure 3.2 Reaction of stock price to new information in efficient and inefficient markets (Buckley, Ross, Westerfield & Jaffe, 1998)...15
Figure 4.1 Pie Chart Large Cap at t-1 ...20
Figure 4.2 Pie Chart Small Cap at t-1 ...20
Figure 4.3 Sign Test Large Cap at t-1 ...21
Figure 4.4 Sign Test Small Cap at t-1 ...21
Figure 4.5 Pie Chart Large Cap at t ...21
Figure 4.6 Pie Chart Small Cap at t...21
Figure 4.7 Sign Test Large Cap at t ...22
Figure 4.8 Sign Test Small Cap at t ...22
Figure 4.9 Pie Chart Large Cap at t+1 ...23
Figure 4.10 Pie Chart Small Cap at t+1...23
Figure 4.11 Sign Test Large Cap at t+1 ...24
Figure 4.12 Sign Test Small Cap at t+1 ...24
Figure 4.13 Pie Chart Large Cap at t+7 ...24
Figure 4.14 Pie Chart Small Cap at t+7...24
Figure 4.15 Sign Test Large Cap at t+7 ...25
Figure 4.16 Sign Test Small Cap at t+7 ...25
Figure 4.17 Pie Chart Large Cap at t+30 ...26
Figure 4.18 Pie Chart Small Cap at t+30...26
Figure 4.19 Sign Test Large Cap at t+30 ...26
Figure 4.20 Sign Test Small Cap at t+30 ...26
Figure 4.21 Sign Test Large Cap at t-1 ...28
Figure 4.22 Sign Test Small Cap at t-1 ...28
Figure 4.23 Sign Test Large Cap at t+1 ...28
1
1 Introduction
“History will be kind to me for I intend to write it” - Sir Winston Churchill
In this introducing chapter we want to give the reader an insight into the subject and present the background of the research topic. The problem discussion will lead to the research questions and finally the purpose of the study will be presented.
1.1 Problem background
Private saving in stocks has experienced large increases worldwide during the last 20 years. The largest increase has taken place in Sweden. In 1985, 30 per cent of the Swedish popula-tion owned stocks either through direct ownership or through pension funds. In 1995 these numbers had increased to 53 per cent of the population reaching a top of 84 per cent in 2003. In 2006 there were 6.7 million, or 77 per cent, of the Swedish population owning stocks. The large increase in private savings in stocks during the beginning of the 21st cen-tury is due to mainly two reasons; the large increase on the Stockholm Stock Exchange and the introduction of the new pension system, PPM, which gave more responsibility to the individual to place his future pension in funds. (Aktiefrämjandet, 2007) Other reasons to the large increase are thought to be due to technological development where new financial instruments has been developed together with new and easier trading systems and the glo-balization of the securities markets (Lycke, Runesson & Swahn, 2003).
The increase in trading is continuing as of today. The average daily turnover at Stockholm Stock Exchange in August 2007 was approximately SEK 26 billion; an increase of 62 per cent compared to the same month 2006. The average daily number of closing operations was in August 2007 approximately 108,000; an increase of 106 per cent compared to Au-gust 2006. (www.omxgroup.com)1
Lundell (2000) speculates in an article in the Swedish financial magazine, Privata Affärer, that we in the future will see increased availability to trade on the world’s stock exchanges, decreased commission fees, lower taxes and more accessible information which in turn will lead to more actors trading on the security markets. Some of his predictions have already been realized.
1.2 Problem Discussion
According to Hansson (2001) the stock market is efficient in the sense that every single stocks’ pricing is set in accordance to all information available, therefore it does not add any value in making your own analysis. If there would exist an easy way to obtain abnormal return on capital, other investors would soon have knowledge of this method and the pos-sibility of abnormal returns would disappear. The efficient market is further distinguished of that all investors are maximizing profits meaning that no single investor can affect the price of a stock.
2
Since small private investors often do not know how to gather appropriate information and how to interpret it in a correct manner they must utilize investment advices. A recent study in Sweden shows that more than every other Swede has very little or little knowledge of the meaning of owning, selling, and buying stocks and/or funds (Aktiefrämjandet, 2007). Bro-kerage house recommendations, financial publications in press, web sites that present free analysis and other information are all places where the small investor can find low cost in-formation (Muradoglu & Yazici, 2002).
The large increase in private savings in stocks, the lack of investment knowledge together with the large increase in Internet usage has resulted in investment advice seeking on the Internet. One of the largest sources of investment advices on the Internet in Sweden today is Avanza.se with a number of unique web visitors of around 190.000 per week for Octo-ber and NovemOcto-ber 2007 (www.kiaindex.se). However, the recommendations presented on the website are made by the Avanza owned Internet magazine Placera Nu. On the Internet Placera Nu is placed wall-to-wall towards Avanza. (www.avanza.se)
Critical voices to the reliability of stock recommendations have been raised. In particular after the crash of the IT-bubble in Sweden year 2000 when the recommendations were ex-tremely misguiding. The daily newspaper Aftonbladet wrote for instance on their news bill, 1st of April 2000, just weeks before the crash when the Swedish people lost 60 per cent of
their savings and a total amount of approximately SEK 570 billion; “Stake on stocks with-out risk of losing”. (Lidén, 2005)
If the theory of efficient markets should hold it would mean that following the advice of a buy recommendation would not be of any value to an investor, but however, there are re-search showing that this is not the case. According to Sant and Zaman (1996) the effect of a buy recommendation differs depending on how many analysts following a certain firm. As the number of analysts following a firm increases, the market reaction to information decreases. This states that information about well-researched firms is more widely available and is thus already reflected in the price while a stock followed by few analysts does not have a price reflecting all available information and the price is more likely to be affected by a recommendation.
Large private interest in stock savings and research showing diversified results concerning the efficiency of the stock market concerns us about private investors making the best pos-sible investment selections. The fact that Internet as information source to stock recom-mendations is growing at rapid pace and that very little research have been conducted in the area makes us want to shed new light to the subject of matter and the following re-search questions are raised;
• Do buy recommendations presented at the Swedish online broker Avanza’s website have effect on the recommended stocks?
• If the recommendations have effect, is there a difference in the effect depending whether the recommended stock is listed on the Stockholm Large Cap or Small Cap, i.e. whether it is a stock with a large or small capitalization value?
3
1.3 Purpose
Our purpose is to describe and analyze if, after a buy recommendation, the effects with re-spect to abnormal return and market efficiency differ significantly depending on a com-pany’s capitalization value.
1.4 Definitions
Avanza does not produce the recommendations on their own; it is the Internet magazine Placera Nu which produces them. Placera Nu is owned by Avanza and lies like a wall-to-wall website to Avanza on Internet. Due to that Avanza and Placera Nu are integrated to a large extent we will further on refer to our source of the issued recommendations as Avan-za.
1.5 Demarcations of the study
- This study is based upon recommendations issued on the online broker Avanza’s website. Further argumentation is to be found in the 2.3 Selection part.
- Our study focuses only on stocks listed on the Stockholm Stock Exchange.
- We have not made an own classification when it comes to companies’ capitalization value. Instead we have relied upon the list classification Small Cap and Large Cap. Further argumentation is to be found in the 2.3 Selection part.
- There are three different ways to test market efficiency; weak form-, semi strong-, and strong form tests. Due to the purpose and nature of this study, it will be limited to the semi strong test.
4
2 Method
“There seems to be some perverse human characteristic that likes to make easy things difficult” - Warren Buffet
In this chapter we describe how we have proceeded to write this thesis. The proceedings of the data collection and the statistical methods are motivated and outlined. How we selected the data is also described and justi-fied.
2.1 Research Approach
There are two main research approaches, namely the inductive and the deductive approach. The extent to which one is clear about the theory initially in the thesis distinguishes which type of study being practiced. When having a deductive approach one move from theory to data, building up a theory and hypothesis and further on creating an empirical part to test the hypothesis. In the case of the inductive approach one would collect data and develop a theory as a result of the data analysis. (Saunders, Lewis & Thornhill, 2003) It is not always easy to make clear distinctions between the different methods (Ejvegård, 2003). Therefore studies are often mixtures, but for instance more deductive than inductive. In this study we first make use of the frame of references and then move to our empirical study to see if it holds, and more narrowed if it holds on a certain level. The theory will also be used to gain better understanding of the result. Due to this we are having a more deductive approach. The collection of quantitative data use to be referred to as a characteristic for a deductive approach (Saunders et al., 2003). The differences between the quantitative and qualitative approaches lie in how to gather information, how to use it and how to analyze it. To fulfill our problem and purpose this method suits better than the qualitative method due to that it helps us to generalize to some extent which we need to do in order to be able to draw con-clusions about the market efficiency and reactions to the issued recommendations.
We have made use of secondary data in our research. A common threat, or disadvantage, to secondary data is that it have been collected for a specific purpose which could differ a lot from the purpose that we, as a “second-hand” user, have. We find our source of secon-dary data, OMX Group, to be a reliable source and we therefore believe there are no rea-sons to question our secondary data. If one collects data specifically for the research pro-ject one is undertaken, it is called primary data. Advantages related to secondary data are that it is often less expensive to accumulate both concerning time and money since it has already been gathered for another purpose and can also be of higher trustworthiness than primary data. (Saunders et al., 2003)
Methods need to fulfill certain criteria’s to be useful and appropriate, namely the criteria’s of reliability and validity. If these criteria are not met the research results have no scientific value. (Ejvegård, 2003) The definition of reliability involves that the measures are done in a correct manner. This means that if the factor of chance is successfully eliminated, if several researchers which use the same method reach the same result, then the research has high reliability (Thurén, 2007). The definition of validity is that one has examined what one in-tended to examine, and nothing else (Eriksson & Wiederheim-Paul, 2006). Saunders et al.
5
(2003) emphasizes the importance, concerning validity, that the findings in a study really are what they appear to be about.
2.2 Data Collection
To gather information to our frame of references we made use of articles in business peri-odicals found through searches on databases on the Internet to a large extent. We also used resources at the Library at Jönköping’s University to find books and other research, for in-stance earlier theses, within the subject. By penetrating books, business periodicals, papers and earlier theses we have not found any studies which focus on what effect buy recom-mendations presented on the Internet have on the stock prices on Stockholm Stock Ex-change. The key words we made use of worth mentioning were; buy recommendations, stock recommendations, efficient markets, media trends, stock analysis, and name of au-thors.
To the parts of empirical findings we used archives as a basis for data. To find the histori-cal All-Share-index values of the Stockholm Stock Exchange we gathered information from its owner, the OMX group’s website. The same goes with collecting information of histori-cal stock prices at the different dates, t. We see the secondary data at the OMX group’s ar-chives as a neutral source. We also argue that the risk with secondary data, that the data in the first place was collected for another purpose than ours, is eliminated and therefore the validity of the study is not violated. The current buy recommendations at the dates of in-terest, t, we gathered from the archives on Avanza’s website.
2.3 Selection
We have chosen to delimit our study to just focus on buy recommendations. Moreover we have treated recommendations concerning “increase” the same as those of strict “buy”, and we will further refer to both of them as “buy recommendations”. We have excluded sell recommendations in our study due to that it should have been too time consuming and not give the study the corresponding extra value. We wanted to test the impact a source is-suing recommendations on the Internet has on the stock price, therefore it was of impor-tance to choose a source that reached to a large group of recommendation-hungry people. For that reason we have chosen to focus our study on recommendations presented on the online broker Avanza’s website. Avanza.se had 210,198 unique web visitors for week 42, year 2007. This placed Avanza on place number 46 when it comes to Sweden’s most visited websites. (www.kiaindex.se) The number of visitors naturally fluctuates from week to week but on average Avanza.se has a number of approximately 190,000 unique web visitors each week for the period of October and November. What one should have in mind is that Avanza do not produce their recommendations themselves, it is the web magazine Placera Nu which produces them. Placera Nu had a number of 81,597 unique web visitors for week 42, year 2007, which gave them the place 114 on most visited sites in Sweden (www.kiaindex.se). This is not as confusing as it sounds since Placera Nu is owned by Avanza and on Internet the two works as wall-to-wall sites. When visiting the sites it is hard for the visitor to distinguish which of the sites one actually uses due to that they are so highly integrated.
6
To fulfill our purpose when it comes to if market reactions after a recommendation differ depending on the company’s capitalization value, we have looked upon the two different lists Small Cap and Large Cap. It is the size of the capitalization value which decides which list the stock belongs to. We will not make our own calculation about the companies’ capi-talization value; instead we will rely on the different lists’ classifications. By looking at all the buy recommendations at Avanza’s website since the 22nd of November 2006 until the
3rd of October 2007 we have a sufficient amount of recommendations to classify the study as reliable, namely a number of 47 recommendations on Small Cap-listed companies and 70 on Large Cap-listed companies. This amount makes the central limit theorem to take ef-fect which requires a sample larger than 30 elements according to Aczel and Sounderpan-dian (2002).
Furthermore we have chosen to compare the development of the stock in relation to our chosen index, OMX Stockholm All-Share. We argue this is the most suitable index to fulfill our purpose, this in accordance to Ross, Westerfield and Jaffe (2005) who argues that to calculate abnormal return for a specific stock one subtracts the market’s return from the re-turn of the stock for the same day. The Stockholm All-Share index is the index that to the largest extent reflects the market’s return. In the beginning of the study we had discussions about what index to choose and we mentioned indices like sector and list indices. This fall off due to that for instance some sector indices are just based upon a few companies and in that way becomes misleading, while it does not exist list indexes for our lists of interest. This is not in line with the purpose of our study and our ambition to fulfill the validity cri-teria.
2.3.1 Why beta-value is not taken into consideration
After discussions whether or not taking beta-value under consideration we concluded that the extra value of including it did not exceed the additional time and effort. The conclu-sions are in line with the argumentation of Fielding (1989). He argues that there are a num-ber of difficulties in measuring betas of companies stocks. One reason is that the data used is total returns (share price changes plus dividends), and not just stock price movements. Another reason is that many stocks are traded infrequently meaning that their stock price remains unchanged when the market as a whole may have moved. This has the conse-quence that the stock obtains an unrealistically low beta since it appears insensitive to mar-ket price changes. Fielding (1989) further argues that there are difficulties in deciding the best length of period to base the estimation of the beta on. If measuring on a short period recent market conditions is reflected but not the impact on the company’s stock price of longer term stock market movements. On the other hand, measuring on a longer period will include stock price movements when economic conditions probably were different from today’s, and the usefulness when estimating the future risk of the stock can be ques-tioned. The findings that betas have a tendency over time to revert to 1 are very interesting. Low betas are increasing and high betas are decreasing, i.e. risky stocks become less risky and safe stocks less safe. (Fielding, 1989)
7
2.4 Data Diminution
In the cases of a recommendation on a stock where it exists trade in both A and B stocks and when the recommendation does not make clear which one of them it means, we will look at the development of the one with the highest turnover, which exclusively is the B stock.
2.5 Sign Test
Section 2.5 is referring to Aczel and Sounderpandian (2002).
When comparing the means of two populations, statistical tests like the t test requires the assumptions that the populations are normally distributed with equal variance. In many sit-uations one or both of these assumptions are not satisfied, the sign test is then a good al-ternative as a more general test that requires fewer assumptions. The sign test is a test for comparing two populations and is stated in terms of the probability that values of one population are greater than values of a second population that are paired with the first in some way. The null hypothesis will therefore state that the probability that the values of one population are greater than the values of the other population is 0.5. The alternative hypothesis is that the probability is not 0.5.
p is defined as the probability that X will be greater than y, where X and y are the two pop-ulations under consideration. The equation for this probability is thus;
p = P ( X > y )
As discussed earlier the null hypothesis states that it is as likely that X will exceed y as it is likely that y will exceed X, i.e. the probability for either occurrence is 0.5. When conducting a sign test the possibility of a tie, i.e. that X = y, is left out. Every pair of X and y where X is greater than y is denoted with a plus sign (+), and every pair where y is greater than X with a minus sign (-). The test assumes that the pairs of (X, y) values are independent and that the measurement scale within each pair is at least ordinal. As stated above ties are not to be considered, leaving us with a number of plus signs which are used in defining the test statistic which is;
y = Number of plus signs
To calculate the probability we also need the parameters n, which is the sample size minus the tied cases, and p, which as stated in the null hypothesis is set to 0.5. With this data we will calculate a z-value with the following formula due to the central limit theorem (CLT) since our n-value is large;
Where, for binomial variables; E (y) = n*p
8
The z-value will then be compared with the critical value of 1.64 in accordance to a 90 per cent confidence level. For our test we found it most relevant testing on a 90 per cent con-fidence level in accordance with suggestions from statistic literature. As we conduct a two-tailed test, a z-value between -1.64 and 1.64 means that the null hypothesis is supported, and reverse, if the z-value lies outside the critical values the null hypothesis is rejected and subsequently the alternative hypothesis is supported. We are aware of the limitations when we conduct this test; it will not measure abnormal return in absolute terms, only compare proportional differences between the populations Large Cap and Small Cap. Although, these limitations will not violate the fulfillness of our purpose.
2.5.1 Sign Test of Abnormal Return
When testing if abnormal return exists we will have pairs of X’s and y’s where X is the change of a stock price at a specific date and y is the index change at the same date. Thus, a stock return larger than the index return will yield a plus sign (+), and reverse, a stock re-turn lower than the index rere-turn will yield a minus sign (-). The days included in the test are t-1, t, t+1, t+7 and t+30 due to that these are the most relevant in order to fulfill our pur-pose.For every test the X’s and y’s are divided in populations of either Large Cap or Small Cap. Our null hypothesis will thus, in a two tailed test, be:
H0: p = 0.50
with the alternative hypothesis: H1: p ≠ 0.50
In this particular test, obtaining a z-value that is positive and larger than 1.64, the alterna-tive hypothesis is supported with the meaning that the population X has larger values than the population y, i.e. the stocks show a significant higher return than the index and abnor-mal return exists. Likewise, a z-value that is negative and sabnor-maller than -1.64 means that the stocks show a significant lower return than index. The two tailed test can be illustrated with this figure;
9
2.5.2 Sign Test of Market Efficiency
When testing whether the market being efficient on a semi strong level we will conduct a sign test giving plus signs (+) for stock returns at the recommendation day, t, with the same return on the previous day, t-1. Minus signs (-) will be given the stock returns on day t with a different return on the previous day. Likewise we will conduct the test in the same man-ner looking at the stock return on day t with respect to the following day, t+1. If in 50 per cent of the cases a stock return would have the same return on the previous (or following) day there would be no systematic pattern that a stock return at day t yields the same return on the previous or following day, neither that it has the opposite effect, i.e. the market is considered to be efficient on a semi strong level. We will thus have the following null hy-pothesis;
H0: the proportion of “neighbouring” stock returns with the same return as on day t=50%
with the alternative hypothesis;
H1: the proportion of “neighbouring” stock returns with the same return as on day t≠50%
If the null hypothesis would be supported, i.e. we obtain a z-value between 1.64 and -1.64, we will conclude that market efficiency on a semi strong level exists. If, on the other hand, a z-value positive and larger than 1.64 would be obtained the null hypothesis will be re-jected meaning that the proportion of “neighbouring” stock returns with the same return as on day t is larger than 50 per cent. The conclusion of such a result would be that the market is inefficient since the effect of the buy recommendation on day t does not only have effect on the stock price on that particular day. If a z-value negative and smaller than -1.64 is obtained, the stock return on day t will in more than 50 per cent of the cases have a different return than on the previous (or following) day.
2.6 Confidence Bounds for testing Abnormal Return
In order to further test whether there exist abnormal returns at the day of recommendation and whether there is a difference between Large Cap and Small Cap we conduct a statistical test using confidence bounds. The idea is eliminating the risk of considering a stock return of 0.0001 per cent as abnormal giving it the same relevance as a return of let say 10 per cent. We will then calculate, for each of the 70 Large Cap and 47 Small Cap recommenda-tions with the help of the statistical software SPSS, the variance for the time frame, t-7 to t+30, excluding the stock price for day t which we expect could contribute to a misguiding large variance. By calculating the mean value and variance for every recommendation and its corresponding time series we will get an upper- and lower bound with the following formula;
X1 represents the upper bound and X2 the lower bound. 1.64 corresponds to a 90 per cent
confidence level, X is the sample mean and S2 is the sample variance. For all data there are
no or insignificant correlation and the time series are considered independent. Further we X
10
will construct, with the help of the SPSS software, a graph for every recommendation with its upper- and lower bounds, plotting the change in stock price, for every day t, from t-7 to t+30. This is an example how it can be displayed:
3 0, 0 0 2 9, 0 0 2 8, 0 0 2 7, 0 0 2 6, 0 0 2 5, 0 0 2 4, 0 0 2 3, 0 0 2 2, 0 0 2 1, 0 0 2 0, 0 0 1 9, 0 0 1 8, 0 0 1 7, 0 0 1 6, 0 0 1 5, 0 0 1 4, 0 0 1 3, 0 0 1 2, 0 0 1 1, 0 0 1 0, 0 0 9, 0 0 8, 0 0 7, 0 0 6, 0 0 5, 0 0 4, 0 0 3, 0 0 2, 0 0 1, 0 0 , 0 0 -1, 0 0 -2, 0 0 -3, 0 0 -4, 0 0 -5, 0 0 -6, 0 0 -7, 0 0 t 2,0 1,5 1,0 0,5 0,0 -0,5 V a lu e DIFF(Acando070206, 1) upper lower
Figure 2.2 Confidence bounds (X1 and X2)
In this way recommendations that have lead to an abnormal return can be sorted out, where abnormal return is considered as a value over the upper bound, thus index is in this test not considered. In order to find out whether there is a difference in abnormal return between Large Cap and Small Cap we set the following null hypothesis;
H0: PSRSC > x1 = PSRLC > x1
with the alternative hypothesis; H1: PSRSC > x1 ≠ PSRLC > x1
Where PSRSC > x1 corresponds to the proportion of Small Cap stock recommendations that
has lead to an abnormal return larger than the upper bound (x1) and PSRLC > x1 corresponds to the proportion of Large Cap stock recommendations that has lead to an abnormal re-turn larger than the upper bound (x1). The proportions will then be analyzed with respect
to the hypothesis to see which one is supported. If the null hypothesis is rejected it has the meaning that it is of statistical significance which according to Ruppert (2004) means that the result can not reasonably be attributed to mere chance and whether this is of practical importance is up to the researcher to decide.
11
3 Frame of Reference
“How is education supposted to make me feel smarter? Besides, every time I learn something new, it pushes some old stuff out of my brain. Remember when I took that home winemaking course, and I forgot how to drive?”
- Homer Simpson
In this chapter we will present selected relevant theories supporting the subject. Initially theories regarding ef-ficient capital market and the relationship between the three types of information are treated. Further on discussions concerning new information, market reactions and firm size will take place. Later the Swedish Stock Market, Information and Trading with Stocks are presented. Finally, previous studies which are re-lated to this thesis are presented.
3.1 Efficient Capital Markets
According to Fama (1970), an efficient capital market is one in which stock prices fully re-flect available information. Further on he argues that in order for the market to be efficient, i.e. the prices are fully reflecting all available information, the following conditions must be fulfilled;
• There are no trading costs in trading securities.
• All available information is costless available to all market participants.
• All agree on the implications of current information for the current price and distribu-tions of future prices of each security.
Wramsby and Österlund (2005) argue that it also exists a fourth condition which needs to be fulfilled in order for the market to be efficient, namely that the investors are price takers. This implies that no private trader on the market is big enough to affect the price on for in-stance a stock.However, Fama (1970) argues that these conditions are sufficient for market efficiency, but not necessary, and in practice the markets do not work this way. For in-stance, as long as investors take all available information into account, despite large transac-tion costs, when the transactransac-tion do take place the security is argued to be priced “fully re-flecting” all available information according to Fama (1970). Ross et al., (2005) illustrates the efficiency arguing that a company releasing an announcement in the morning will im-mediately have affect on the stock price. There will be no possibility for an investor to buy the stock in the afternoon and make a profit the day after since the efficient market hy-pothesis predicts that the share price will be adjusted immediately after the announcement in the morning. They further argue that the efficient market hypothesis has implications for investors and firms. An investor should only expect to obtain a normal rate of return be-cause information is reflected in prices immediately, before the investor has time to trade on it. Firms selling securities should expect to receive a fair value, i.e. the present value, meaning that they can not fool investors by receiving more than that value. These are as-sumptions concerning an efficient market, i.e. all available information is immediately re-sponded to by the market. However, in reality different kinds of information may affect stock prices more quickly than other information. After further investigating the findings
12
of Fama (1970), Ross et al. (2005) separates information into three different types depend-ing on market response rates.
3.1.1 The Weak Form
A capital market is weakly efficient when all information being used are past prices. His-torical information is the easiest kind of information and if it would be possible to make extraordinary profits by looking only at historical patterns, everyone would do it thus eliminating this possibility. (Ross et al., 2005)
3.1.2 The Semi strong Form
A capital market is semi strong efficient when prices are reflecting all publicly available in-formation. Publically available information is publicly released company reports, press re-leases, newspaper articles and historical price information. In a semi strong market the price of a stock should adjust immediately after new information has been released. (Ross et al., 2005)
3.1.3 The Strong Form
If a capital market is strongly efficient not only the above mentioned publicly available in-formation is included in the price but also private inin-formation. In a strong form efficient market an insider with private information about a company would not be able to profit from this information, i.e. any information with a value to the stock price known to at least one investor is fully incorporated into the present price. Empirical evidence shows that strong market efficiency does not exist in reality and the semi strong efficiency seems to be most common. (Ross et al., 2005)
13
Figure 3.1 Relationship between the three types of information (Buckley, Ross, Westerfield & Jaffe, 1998)
3.2 Testing the efficiency of the market
3.2.1 Weak form test
As stated above, in a weakly efficient market one would be able to predict future stock prices by looking at historical price patterns. The random walk hypothesis is used to test whether a market is efficient in its weak form. The hypothesis implies that a stock’s histori-cal price movement is unrelated to future movements. Efficiency in its weak form can be tested mathematically with the following formula;
Pt = Pt-1 + Expected return + Random errort
The formula states that the price of today equals the sum of the last observed price plus the expected return for the specific stock plus a random error component occurring over the interval. The random component is depending on new information on the stock and can thus be either positive or negative. The random component in one period is unrelated to the random component in any past period – it is thus unrelated to past prices and if stock prices follow this equation they follow a random walk. There is a consensus in previous lit-erature that this hypothesis holds meaning that markets are efficient in its weak form, meaning that there is no predictability of future stock prices from past prices. (Ross et al., 2005)
14
3.2.2 Semi strong test
After extensive research supporting efficiency on the weak level, Fama (1970) extended his research to testing the speed of price adjustment to publicly available information, i.e. test-ing semi strong efficiency. This is usually done with an event study, testtest-ing that the released new information (annual earnings, change of CEO, new issues etc.) only effects the price of the stock at time t, and not before (t-1) or after (t+1) the new information is being re-leased. The abnormal return of a stock for a particular day is according to Ross et al. (2005) calculated by subtracting the market’s return from the return of the specific stock for the same day;
AR = R - Rm
Ross et al. (2005) further argues that in an efficient market previous information will al-ready be incorporated into the stock price, and what the market does not yet know, i.e. fu-ture information, can also not be reflected in the price today. If efficient, new information at time t will cause abnormal return at time t only. Event studies examines whether the re-lease of information causes abnormal returns only at day t or at other days as well, before or after the releasement. Fama (1970) points out that previous research on the semi strong form made on various types of public announcement’s affects on abnormal stock returns are all showing on efficiency on the semi strong level.
3.2.3 Strong test
To test strong market efficiency means testing whether access to not publicly available in-formation will yield an abnormal return. Fama (1970, p.398) concludes that “Thus it seems that the specialist has monopoly power over an important block of information, and, not unexpectedly, uses his monopoly to turn a profit. And this, of course, is evidence of market inefficiency in the strong form sense.” Ross et al. (2005) shows on various researches re-garding insider trading where official records from the Securities and Exchange Commis-sion in the U.S.A has been evaluated. The results are clear; the insider trading was abnor-mally profitable.
3.3 Abnormal Returns in Efficient and Inefficient Markets
If you, as a potential investor, are eager to learn about a potential investment object, an enormous amount of information is available to you. Not only that it exists a large set of information to gather, the motive for doing so is also large; the profit motive. (Buckley, Ross, Westerfield & Jaffe, 1998)
A market is efficient with respect to information if there is no way to make abnormal re-turns or excess profits by using that information. Without knowing anything special about a stock, an investor acting in an efficient market expects to earn an equilibrium required re-turn from an investment. (Buckley et al., 1998)
If the market is efficient the price will respond to new information as the solid line in the figure below illustrates. However, this is not always the case. If the market reacts as the early response line shows one can expect that inside information exists on the market. The broken line illustrates an overreaction and then correction back to its true price. In contrast the
15
dotted line pictures a delayed reaction. Here the market acts slowly and it takes 30 days for it to fully absorb the information. The two latter cases illustrate paths that the stock price will take under an inefficient market. If it takes several days for the market to react to the new information about a stock, for instance a buy recommendation on the stock, trading profits will be available to investors buying at the date when new information is released and selling when the price is back at equilibrium. (Buckley et al., 1998)
Figure 3.2 Reaction of stock price to new information in efficient and inefficient markets (Buckley, Ross, Westerfield & Jaffe, 1998)
Efficient-market response: The price directly adjusts to and fully reflects new information. Early response: New information is leaked and price responds before public dissemination
Delayed response: The price partially adjusts to the new information; it takes the market 30 days to completely absorb the new information.
16
3.4 Market Reactions and Firm Size
Sant and Zaman (1996) found that the number of analysts following a stock has a large im-pact on a stock’s abnormal return after a favorable recommendation in Business Week’s ‘Inside Wall Street Column’. The strongest abnormal return was shown for stocks followed by less than five analysts whereas stocks followed by more than twenty analysts showed no abnormal return. Further there is a relationship between the number of analysts following a firm and the speed of price adjustment to new information. According to Brennan, Jegade-esh and Swaminathan (1993) and Lo and MacKinlay (1990), stocks followed by many ana-lysts respond more rapidly to new information than stocks followed by few anaana-lysts. Thus, both abnormal return and speed of price adjustment are related to the number of analysts following a stock. As stated earlier, our intention is not to look at number of analysts fol-lowing a stock but rather at the stock’s capitalization value, and it is therefore interesting and highly relevant to look at the relationship between the number of analysts following a stock and the stocks’ capitalization value. Bhushan (1989) argues that number of analysts following a firm is an increasing function of firm size. This, he argues, is due to that an in-vestor is likely to find a piece of private information about a large firm more valuable than the same piece of information about a small firm since he can trade a greater amount of money in the large firm without authorities getting suspicious of insider trading. He also argues that since analysts strive to generate transactions for their companies, analyzing large companies stimulates the interest of a larger number of investors and therefore creates a greater potential of creating transactions. Merton (1987) means that size is positively asso-ciated with the number of individuals who are interested in a firm, thus creating a higher speed of price adjustment. We can thus see that firm size is highly related to the number of analysts following a firm, which in turn has impact on both abnormal return and the speed of price adjustment.
3.5 The Swedish Stock Market
In October 2006 a Nordic Marketplace was created. The owner, OMX Group, listed com-panies from Sweden, Denmark and Finland on the same market place in order to make it more attractive for investors. This meant that the former separate markets now have the same rules, increased trading hours and are traded with the same trading system. (www.omxgroup.com)2 The Nordic Exchange offers access to around 80 per cent of the
stock market trading in these markets. Since year 1990 trading on the Stockholm Stock Ex-change is fully automated, meaning that there is no trading on the floor since this date. All orders are matched in the electronic trading system SAXESS (www.riksbank.com). By the end of 2006 the Nordic Exchange had 791 listed companies, an increase by 50 companies compared to 2005. These companies had a total capitalization value of SEK 8,435 billion. There are 159 members (intermediaries) of which the most important ones are SEB, Car-negie, Handelsbanken, Nordea and Morgan Stanley. The total turnover of stock trading reached SEK 9,731 billion making the Nordic Exchange the fifth largest in Europe (www.omxgroup.com)2. As the Nordic Exchange was created in October 2006 new listings
of the companies were also created. With respect to a company’s capitalization value it is listed on either Large Cap, Mid Cap or Small Cap and these lists are being revised semi-annually in May and November. The criteria for the listings are:
17 • Large Cap – Market capitalization > € 1,000 million • Mid Cap – Market capitalization € 150 – € 1,000 million
• Small Cap – Market capitalization < € 150 million (www.halvarsson.se)
3.5.1 OMX Stockholm All-Share Index (OMXS)
The All-Share-index of the Stockholm Stock Exchange includes all stocks listed on the Nordic exchange in Stockholm. The purpose with the index is to mirror the current status and development on the market. The reference date, the date of base, for this index is De-cember 31th 1995, with the base value of 100. (www.omxgroup.com)3
3.6 Information and trading with stocks
The Swedish media market has undergone a large change since the beginning of the 1980’s and which accelerated during the 1990’s (Sundin, 2006). A new media landscape is emerg-ing and it is the Internet and broadcastemerg-ing which mostly have contributed to this develop-ment (Harrie, 2006).
The Internet is a part of daily life in Sweden. Most people have access and use it frequently. The rapid expansion of broadband services and a strong response among subscribers have accelerated the trend (Harrie, 2006). It exists more than 50 million applications of band in the EU, and the Nordic countries have reached far in their expansion of broad-band (Bengtsson, 2006). In 2005 almost three out of four households had access to the In-ternet in their homes. Households use consists primarily of e-mail, household tasks, par-ticipation in discussion such as chat groups, newspaper reading and information searching (Harrie, 2006). The information search on Internet also holds when it comes to investment decisions. Here investors can find the information needed at a very low cost (Muradoglu & Yazici, 2002). The majority of all facts that an investment decision is based upon is second-hand information which have been received from sources with diverse credibility. This second-hand information could be both known facts and predictions. Different actors have diversified opinions about facts which are not verified and different skills in interpreting available information which give differences in their investment behavior. Since all actors trade under the same basic limitations, the final price of the stock will reflect the total sum of the value of the forecasts for them all. (Samuelsson, 1991)
In seven years the Internet has changed the foundation of stock trading. Then, we are not having IT bubbles in mind, instead that every fifth order on the Swedish Stock market is traded through pure online brokers. Lower commission fees and larger availability are seen as explanations behind this development which in turn have made it easier for private per-sons to buy and sell stocks. (Almström, 2005)
18
3.7 Previous Research
We will here present previous studies in the area of our research subject will be presented. This is done with the reason to give you, as a reader, deeper understanding in the area and to see where we place our study comparing to earlier studies.
Eugene F. Fama (1970). Efficient Capital Markets: A Review of Theory and Em-pirical work.
Journal Article
In this article, Fama presents both empirical and theoretical literature concerning the effi-cient market model. His ideal scenario how a capital market should look like is a market in which prices present correct signals for resource allocation. According to him this is when investors can choose among securities with the underlying assumption that prices at any time “fully reflect” all available information. Such a market is called “efficient”. Fama di-vides the market into three forms depending on how much of the information being re-flected in the market. The market is said to be weak form if the pricing totally is based on historical price movements. The semi strong form tests whether prices efficiently adjust to in-formation that is publicly available. Such inin-formation could for instance be press releases and annual reports. Finally, strong form tests whether given investors or groups have mo-nopolistic access to any information relevant concerning pricing of the stock. Fama´s con-clusion is that this model stands up well, with a few exceptions.
Mattias Hedenryd and Peter Sandén (2000). The influence of buy recommendations on stock prices - ‘A survey of four business magazines.’
Master Thesis
Hedenryd and Sandén looks at the effect of buy recommendations published in the four mayor Swedish business newspapers/magazines. They found that after a recommendation was published in the largest business newspaper, Dagens Industri, the stock showed on av-erage a positive abnormal return of 6.18 per cent compared to index. The avav-erage for all recommendations were a positive abnormal return of more or less two per cent at the day of the publishment of the recommendation, but a negative return at a time from a week to a month after the recommendation, to showing again a positive return after six months. Their results show that abnormal return exists due to buy recommendations and classifies the market as being semi strong according to the classifications of Fama. Critics to their re-sults are possibly due to the fact that the study was conducted in the year 2000 during the IT-bubble and their study shows that those newspapers publishing the most recommenda-tions on companies from the “new economy” also had the largest abnormal return.
Rajiv Sant and Mir A. Zaman (1996). Market reaction to Business Week ’Inside Wall Street’ column: A self-fulfilling prophecy.
Journal Article
Sant and Zaman looks at the market reaction to stock recommendations published in Busi-ness Weeks ‘Inside Wall Street’ column, i.e. secondary information targeting a large public. They found that a stock which has been rewarded with a favorable recommendation in Business Week earn significant positive abnormal returns at the time of distribution of the Business Week. But, a significant positive abnormal return was only true for stocks
fol-19
lowed by less than 21 analysts and the positive abnormal return was decreasing as the number of analysts following the stock was increasing. In fact, a stock followed by more than 20 analysts had no abnormal return. And reversely, the fewer analysts following a stock, the larger the abnormal return. The strongest results were shown for stocks followed by five or fewer analysts. These stocks usually have lower average trading volume, com-pared to stocks followed by more analysts, and experienced the largest increase in trading at the distribution day of Business Week.
Eurico J. Ferreira and Stanley D. Smith (1999). Stock price reactions to recommen-dations in the
Wall Street Journal
“Small Stock Focus” column.Journal Article
Ferreira and Smith study the information content of the Small Stock Focus in the Wall Street Journal, i.e. a source of secondary information readily available on a daily basis. The column tends to focus on unlisted stocks that have very large price changes on the day be-fore the publication in the Wall Street Journal. When all stocks with bad news announce-ments were assumed to be sold short were combined with the stocks with good news an-nouncements assumed to be bought, the results showed an abnormal return on the day be-fore and on the day of the announcement and then a decrease on the following days. This was possibly reflecting an adjustment to the large response on the day before and of the announcement. They found that the increase before and at the announcement was larger for less actively traded, hence more actively traded stocks was less affected by an an-nouncement in the Wall Street Journal. Their results are consistent with the above men-tioned study by Sant and Zaman who found positive abnormal returns for favorable re-ports for stocks followed by less than 20 analysts.
Michael J. Brennan, Narasimhan Jegadeesh and Bhaskaran Swaminathan (1993). Investment Analysis and the Adjustment of Stock Prices to Common Information. Journal Article
Brennan, Jegadeesh and Swaminathan looks at the relationship between the number of in-vestment analysts following a stock and the speed of price adjustment of the stocks’ price as new information is released. They looked at data for all firms from the CRSP NYAM-NASDAQ tapes from January 1977 to December 1988, looking at firm size and number of analysts following each firm, and then running statistical analysis of the speed of price ad-justment as new information was released. They found support for their hypothesis that many analyst firms react faster to common information than do the few analyst firms.
20
4 Empirical Findings and Analysis
“It is clear our nation is reliant upon big foreign oil. More and more of our imports come from overseas.” - George W. Bush
In this chapter we will present the results of the study as clear and specific as possible. The results in relation to our theory underlie the analysis. As we will present a large amount of data, the results and analysis will be presented in an integrated manner. This will increase the readability due to that the reader can follow the analysis meanwhile referring to the results presented in the figures.
4.1 Abnormal Return
In this section the results and analysis concerning abnormal returns will be presented. It will be demonstrated by the use of a more descriptive method, pie charts, and also through a statistical sign test. The pie charts will present the proportion of Large Cap and Small Cap stocks which have yielded an abnormal return. Those stocks’ that have shown abnormal re-turn belongs to the positive side. The sign test is done with the aim to statistically secure the study’s results. Still, the pie charts and the sign test are based upon the same data. The results will help to partly answer the question if the recommendations at Avanza’s website have effect and mainly if there is a difference in the effect of the recommendations whether the stock is listed on the Large Cap or Small Cap.
4.1.1 Abnormal Return at t-1
Figure 4.1 Pie Chart Large Cap at t-1 Figure 4.2 Pie Chart Small Cap at t-1
Here Large Cap shows a proportion of 53 per cent of the stocks that yield abnormal return, while Small Cap shows a result of 55 per cent. The proportion of abnormal returns therefore differ with 2 per cent when comparing the two lists with each other.
The two populations show on almost identical abnormal returns and when comparing them with the abnormal returns of the day after redommendation, t+1, the results are rather similar. The recommendation is therefore expected to have effect at recommendation day, t.
21
4.1.2 Sign Test Abnormal Return at t-1
Figure 4.3 Sign Test Large Cap at t-1 Figure 4.4 Sign Test Small Cap at t-1
Large Cap has a z-value of 0.48 which is found within the confidence interval, supporting the null hypothesis, at the same time as Small Cap has a z-value of 0.73 which is also found within the confidence interval, supporting the null hypothesis.
As the results show, both the z-values are close to 0, one could therefore argue that no ab-normal return is present. This is not in line with the study of Ferreira and Smith (1999) which found tendencies of abnormal returns for the day before recommendation.
4.1.3 Abnormal Return at t
Figure 4.5 Pie Chart Large Cap at t Figure 4.6 Pie Chart Small Cap at t
Here Large Cap shows a proportion of 57 per cent of the stocks that yield abnormal return, while Small Cap shows a result of 66 per cent. The proportion of abnormal returns therefore differ with 9 per cent when comparing the two lists with each other. When comparing this day’s abnormal return with the previous day, t-1, one can see that the proportion has increased with 4 per cent for Large Cap and 11 per cent for Small Cap.
22
The difference of 9 per cent between the lists is too large to be ignored, meaning that Small Cap recommended stocks show larger abnormal return than those belonging to Large Cap. These results are all in line with previous conducted research in the area that shows that stocks with smaller capitalization value are to a larger extent affected by a recommendation. Bhushan (1989) argues that there is a relation between the firm’s size, i.e. capitalization value, and the number of analysts following a stock. Therefore the results above are in line with the study of Sant and Zaman (1996) which found that stocks’ which are followed by a larger amount of analysts to a larger extent reflect all available information, and therefore the stock price is not showing noteworthy effect after being recommended. Small Cap stocks are more affected by a recommendation since they are analyzed by fewer analysts and therefore have a stock price which is not to the same extent reflecting all available information. The results also show an increased amount of stocks showing abnormal returns at t comparing to the day before, t-1, both in the case of Small Cap and Large Cap. This is in line with the research by Hedenryd and Sandén (2000) which showed on abnormal return, due to the recommendation, of on average 6.18 per cent after recommendation on stocks listed on Stockholm Stock Exchange. Yet, they did not distin-guish between Large Cap and Small Cap. However, concerning Large Cap the difference is only 4 per cent which can not be considered as significant while in the Small Cap case the difference reach a level of 11 per cent which could be seen as significant. One could see this highly related to the recommendation.
4.1.4 Sign Test Abnormal Return at t
Figure 4.7 Sign Test Large Cap at t Figure 4.8 Sign Test Small Cap at t
Large Cap has a z-value of 1.2 which is found within the confidence interval, supporting the null hypothesis, while Small Cap has a z-value of 2.19 which is found in the right-hand side rejection region and therefore rejecting the null hypothesis. Compared to the z-values for t-1, Large Cap experienced an increase of 0.5 while Small Cap experienced an increase of 1.5.
The fact that Large Cap supports the null hypothesis and that Small Cap rejects it gives stronger evidence that stocks listed on Small Cap are affected by the recommendations to a significantly higher extent in comparison to the Large Cap stocks which can not be statisti-cally considered to show an abnormal return after recommendation. These results answer our research questions both regarding whether recommendations have effect and whether there is a difference of the effect depending on the stocks capitalization value.
23
4.1.5 Abnormal Return at t+1
Figure 4.9 Pie Chart Large Cap at t+1 Figure 4.10 Pie Chart Small Cap at t+1
At this day, t+1, Large Cap shows a proportion of 51 per cent of the stocks that yield abnormal return while Small Cap shows a proportion of 57 per cent of the stocks that yield abnormal return. One can here see that the possible affects that occurred at the day of recommendation, day t, now have decreased and moved back to almost the same proportions as the day before the recommendation, t-1. When comparing the two populations respectively with the day of recommendation, t, one can see that the proportion of Large Cap decreased with 4 per cent and Small Cap decreased with 9 per cent.
As we saw at day t, the results clearly showed abnormal return for Small Cap, but for this day the proportion has fallen back to where it was at t-1. As no abnormal return could be found for t+1 and neither for t-1, the result shows that the abnormal returns is concentrated at day t, the day of recommendation. The pattern is similar for Large Cap stocks, but it is too weak and small to distinguish any clear pattern.
24
4.1.6 Sign Test Abnormal Return at t+1
Figure 4.11 Sign Test Large Cap at t+1 Figure 4.12 Sign Test Small Cap at t+1
Large Cap has a z-value of 0.24 which is found within the confidence interval, supporting the null hypothesis, while Small Cap has a z-value of 1.02 which is also found within the confidence interval, and also supporting the null hypothesis. For Large Cap we see that the z-value has experienced a decrease with 1 and for Small Cap the z-value has experienced a decrease with 1.2 compared to the previous day, t.
As discussed above this statistically secures that no abnormal return for either Large Cap or Small Cap is found for t+1.
4.1.7 Abnormal Returns at t+7
Figure 4.13 Pie Chart Large Cap at t+7 Figure 4.14 Pie Chart Small Cap at t+7
At this day, t+7, Large Cap shows a proportion of 51 per cent of the stocks that yield abnormal return while Small Cap shows a proportion of 32 per cent of the stocks that yield abnormal return, a difference of almost 20 per cent when comparing the two. In comparison to the results of t, the proportion of Large Cap has reduced with 6 per cent, while for Small Cap the proportion has reduced with 34 per cent.
25
Large Cap shows a stable development, here at t+7 the stock return follows index just as in the previous measured days. The difference from day t is only 6 per cent which is very small. Small Cap on the other hand shows a remarkeably large difference, from a large abnormal return at day t (66 per cent) to here showing a proportion of only 32 per cent of the recommended stocks which have larger stock return than index. The positive effect of the recommendation has now turned to a negative return in seven days. This shows tendencies of what Buckley et al. (1998) argues is overreaction, pictured in Figure 3.2, where the stock price experience a large increase at the day of recommendation and later suffer a negative development as it adjusts to what probably is its accurate price with respect to the new information. The result of the test is also in line with the study by Hedenryd and Sandén (2000) who found abnormal return for recommended stocks at Stockholm Stock Exchange at day t which then turned to a negative development.
4.1.8 Sign Test Abnormal Return at t+7
Figure 4.15 Sign Test Large Cap at t+7 Figure 4.16 Sign Test Small Cap at t+7
Large Cap has a z-value of 0.24 which is found within the confidence interval, supporting the null hypothesis, while Small Cap has a z-value of -2.48 which is found in the left-hand side rejection region, thus rejecting the null hypothesis. In Large Cap we see that the z-value has not experienced any change but for Small Cap the z-z-value has experienced a de-crease with 3.5 when comparing to the previous measured day, t+1.
The results show clearly that Large Cap has experienced a stable development as it has ex-actly the same z-value as at the previous measured day, t+1, and is supporting the null hy-pothesis with distinction. Small Cap shows on large changes in the proportion of positive abnormal return, and is at t+7 strongly rejecting the null hypothesis on the left-hand side. In many cases where the recommendations seemed to be good at t has now turned out, in a time period of seven days, to show negative results.
26
4.1.9 Abnormal Returns at t+30
Figure 4.17 Pie Chart Large Cap at t+30 Figure 4.18 Pie Chart Small Cap at t+30
At this day, t+30, Large Cap shows a proportion of 47 per cent of the stocks that yield abnormal return at the same time as Small Cap shows a proportion of 36 per cent of the stocks that yield abnormal return. When comparing the two lists with each other there are 11 per cent more of the Large Cap stocks that shows abnormal returns than in Small Cap. In relation to the previous measured day, t+7, there is no difference worth considering. Here we see once again that it is approximately the same proportion of Large Cap stocks which shows abnormal returns as in the two previous measured days, t+1 and t+7. Small Cap continues to show the same result as in t+7, meaning that 30 days after the
recommendations the development of the recommended stocks are overwhelmingly negative. The tendencies for this day are similar to the discussion of overreaction held in 4.1.7. The results for both t+7 and t+30 further answers the reseach questions whether there is a difference of the effects of the recommendations depending on the stock’s capitalization value.
4.1.10 Sign Test Abnormal Returns at t+30
Figure 4.19 Sign Test Large Cap at t+30 Figure 4.20 Sign Test Small Cap at t+30
Large Cap has a z-value of -0.48 which is found within the confidence interval, supporting the null hypothesis, while Small Cap has a z-value of -1.9 which is found in the left-hand
27
side rejection region, thus rejecting the null hypothesis. In Large Cap the z-value has ex-perienced a change of -0.7 and for Small Cap the z-value has exex-perienced a decrease with 0.6 in comparison to t+7.
In this period the null hypothesis is supported with distinction concerning Large Cap. Therefore, over 30 days there is no added value in following the recommendations than tossing a coin. Since for Small Cap the null hypothesis is rejected the result of the recom-mendations after 30 days is clearly showing on a negative development, thus tossing a coin would in this case be a better option. We are aware of that other factors might have influ-enced the individual stock’s price development but as we have a satisfying large amount of observations we believe such factors will not influence our results.
4.2 Confidence Bounds for testing Abnormal Return
In this section the results and analysis concerning abnormal returns in relation to the calcu-lated upper and lower bounds will be presented. Abnormal return is considered as a value over the upper value, thus index is in this test not considered. By this the number of stocks which, at recommendation day, exceeds its upper value could be counted. Consequently this makes it possible to compare the two lists’ proportion of stocks which show abnormal return. This was tested with the null hypothesis;
H0: PSRSC > x1 = PSRLC > x1
and the corresponding alternative hypothesis; H1: PSRSC > x1 ≠ PSRLC > x1
The results for Large Cap were a proportion of 22.9 per cent with values exceeding the up-per confidence bound and for Small Cap 25.5 up-per cent. Small Cap is thus having a propor-tion 2.6 per cent larger than Large Cap.
As this is another way of looking at abnormal return one could expect the results to differ somewhat from the test of abnormal return in 4.1. The results in this test do not show on differences between Large Cap and Small Cap to as large extent, however there is still a dif-ference between the two populations where the recommendations have larger impact on Small Cap stocks. The differences are in this test again confirmed which gives additional support to the tests made in 4.1, thus our research questions are responded once again. The result of this test also gives additional support to the study made by Sant and Zaman (1996). Their study showed that as the number of analysts following a stock decreases, which is linked to a smaller capitalization value (Bhushan, 1989), the abnormal return in-creases. The result here is not strong enough to say that the recommendations on Small Cap stocks are affected to a larger extent compared to Large Cap, but as this is in line with both the results from our earlier tests and other studies we indeed see this as a strong indi-cation.
