Behavioural Aspects of Asset Pricing
Can the triggers of
mispricing be identified through analysis of their causing phenomena?
Masters Thesis of Yordan Krasnaliev s1739883
Abstract
The following paper investigates the stock mispricing due to the overreaction of market players to information. It identifies the several, suggested by literature, psychological phenomena as a main cause for overreaction. Within the
research the phenomena are put in one context and through the level to which they are represented in the tested population and the interaction between them, the environment in which mispricing occurs is recreated. The research offers valuable insights about the mechanism of mispricing and suggests possible further developments of the science in
order to make the mispricing more foreseeable.
Special Acknowledgements for the support of my supervisor dr. Wim Westerman, my co-assessor Mr. Aljar Meesters, the
participants of the university investment organization Risk, headed by Joline Stavasius, the participants of the specialisation course Advanced Financial Management and at last but not least, my personal acknowledgement to Karsten Bloch, Arjan Reinders and Sebastian Wirtz.
Program: Double degree MSc IB&M International Financial Management Faculty: Economics and Business
Supervisor: Dr. Wim Westerman, University of Groningen Co-Assessor: Mr. Aljar Meesters
TABLE OF CONTENTS
I. PROBLEM FIELD AND RESEARCH METHOD 3
A. PROBLEM IDENTIFICATION AND RESEARCH QUESTION 3
B. THEORETICAL IDEA AND RESEARCH METHOD 3
I. STRATEGIC METHODOLOGY CONSIDERATIONS 4
II. QUESTIONNAIRES 5
III. FURTHER DEVELOPMENT OF THE RESEARCH 8
II. LITERATURE REVIEW AND QUESTIONS 8
A. UNDERLYING RESEARCH 8
B. PHENOMENA EXPLAINING MISPRICING & QUESTIONS’ DEFINITION 9
I. LITERATURE RELATED TO PSYCHOLOGICAL CONSTRUCTS 9
II. LITERATURE RELATED TO INFORMATION SIGNALLING 15
III. FINDINGS AND ANALYSIS 17
A. DESCRIPTIVE STATISTICS 17
B. CORRELATION 21
C. OPEN QUESTIONS 25
D. ANALYSIS SUMMARY AND ADDED VALUE 26
IV. CONCLUSION AND IMPLICATIONS FOR FURTHER RESEARCH 28
I.
Problem Field and Research Method
A.
Problem identification and research question
Driven by moral and commercial incentives, portfolio managers have to maximize the profit of the portfolios under their jurisdiction. In addition to the technical portfolio management, the managers have to make strategic choices that would maximize the returns on a short and medium run. In this respect a great volume of academic work offers various solutions. This paper was inspired and motivated by the findings of Jegadeesh and Titman (1993) and Jegadeesh and Titman (1999)1, suggesting that momentum and contrarian strategies (trend following strategies) deliver abnormal returns. The authors attributed the findings to overreaction to stock price information. In essence: one can realize abnormal return on the cost of the mispricing of assets by the market. With the following paper I will analyse the behavioural explanations of this phenomena. The goal and research purpose of this paper is
to identify the most significant properties of the environment which pre-dates the mispricing of stocks. I will identify possible predictors or premises for the evolving of trends
through mispricing, as a result of investor behaviour, through recreating the environment in which mispricing is most likely to occur. The main idea is that through gaining better understanding and insight in how and why momentum and contrarian strategies deliver abnormal returns, in other words, how does the mispricing occur, one would be able to exploit those strategies in a more efficient way. I would like to contribute to the general understanding of trend movements.
B.
Theoretical idea and research method
My main focus will be on mispricing caused by overreaction as the critical force of the movement of stock prices. With the research I would like to qualitatively explore the investor behaviour, isolate the possible causes for overreaction and find possible premises that enable this particular behaviour. In this regard, I will view the stock price fluctuations as a cycle framework on which I will try to indentify the triggers for the upward or downward movement. I will conduct the research under the roof assumption that rationality is relative and not universally applicable. It is relative to the individual’s perceptions and the personal character
1
features of the investors. Thus I believe that each type of behaviour is motivated by good sense and I will try to show it. I will apply Descartes’ method of reasoning which implies that every man/ woman is acting according to his/ her own perception of right and wrong and everyone is in a way right for him/ herself no matter which way they choose.
The data will originate form questionnaires and interviews and real case analysis with the findings from the above. Test subjects will be selected students of the University of Groningen who have profound knowledge in finance and are actively investing. The open interviews will be conducted with the help of three knowledgeable private investors, where the questions will reflect the observations form the questionnaires and experiments.
In order to outline the research methodology, I will start with a step-by-step description of my approach. At first, I will send and collect questionnaires in order to test the degree to which various behavioural phenomena, which explain overreaction, are represented in the students’ behaviours. Based on the results, I will find links between those phenomena, with which I will generate picture of the environment where mispricing occurs. I will then summarise this environment and construct open questions towards competent control subjects from the field. With the answers from the questionnaires and the findings from their analysis I will have a picture of the realm pre-dating mispricing. I will then analyse a real case from the financial world in order to test the applicability of the findings.
i. Strategic methodology considerations
Several considerations concerning the testing method are inspired by the paper of Hoffmann et al. (2006) and are going to be of importance for me when creating the questionnaire. Those considerations are of strategic weight since they are precondition to a valid and reliable research. The authors suggest that, in case of an experiment, one should start with conducting a set of rules, for every single experiment, that corresponds/ replicates those which the agents face in the real world2. Next, a population of agents is identified and the experiments are performed. The main idea of the authors is that the movements of the prices on the stock exchange are interplay and a result of multiple agents’ accumulative behaviour. Furthermore, an insight of the approach of Hoffmann et al. is that one can measure the level to which
2
investors stick to the purely technical risk-return considerations when making their investment choice. Thus, after I qualitatively measure the degree to which investors comply with risk-return rules, I will be able to weight the findings of the research on overreaction in order to put them in perspective in the real world. I will combine the results from this experiment with observations from the real world, in order to derive conclusions about the degree, to which the real world stylized factors, suggested by the relevant literature, are replicated by the research.
ii. Questionnaires
In its’ core the research will be qualitative evaluation of relevant literature findings from the psychology and behavioural finance fields combined with a row of interviews. In order to begin my research I will “disassemble” the overreaction that causes mispricing and identify which types of overreaction are available in order to recreate the realm pre-dating mispricing. Some of the reasons for mispricing that I will test, are for example representative heuristics, self-attribution bias, recency, conservatism etc., which will be later discussed in detail in the literature review.
Since the triggers of mispricing which I am going to research are psychological constructs that measure preference or attitude, we can test this fundamental psychological feature on any subject. Thus the interviews and experiments will be made with Financial Management students at the University of Groningen and with the assistance of RISK - the student investment organisation at the above university.
In order to enrich these accumulated findings I will test risk aversion in the subjects as a control feature3, since it is not yet clearly discussed in the literature how subjects’ reaction to uncertainty is linked to mispricing. Next to it I will control for gender, origin and risk attitude, on which I will elaborate later in the progress of the research.
Mispricing occurs due to the cumulative actions of many subjects and not a single person. Thus we can expect that mispricing is due to two reasons; first, overreaction of many people and second, the expectation of many people that the others will over-/ under-react. Respectively, we can add questions asking one subject to predict the reaction of another subject while knowing what information the other received. For this, according Massey and Wu (2004), not only the precision of the signal, but also stability of the environment, influences the degree of over/ under-reaction and this will also be incorporated in the questionnaire through an experiment. Another valuable insight of the above authors is their definition of signal as noisy and precise. This will also be incorporated in the test (questionnaire or experiment).
In order to measure the attitudes towards stock price movements or positive and negative attitude towards financial events, I will construct also questions testing attitudes as an unconscious basis for conscious behaviour such as the one specified by K. Goldstein (1941). Using this information, clear questions can be generated to test the attitudes of the subjects to financial information signals. Here it is essential to mention that only information about well known events will be used or such questions that do not require knowledge about a particular situation.
My goal is to use the numbers derived from this test and to figure out to which extent they correlate or not. Thus I will be able to create a picture of the environment in which misjudgement of the real situation is likely to occur which, expressed alternatively, is also the main purpose of this study. I will identify a pattern in the information type that caused the positive or negative misjudgement, so I can draw qualitative conclusion about the possible trigger and interplay between the triggers.
Finally, I will sum up the findings and generalize them in order to apply them to a real life case of mispricing. The main idea is that depending on if variables correlate or not, one will be able to attribute the mispricing to interaction between them, draw a conclusion about the way they interact and judge about the fundaments of this interaction. Through the above method, I will not answer the research question with ”yes” or “no” but I will rather recreate a model of
3
the environment in which mispricing is likely to occur. In this way I will be able to recreate the behavioural framework in which mispricing occurs.
Considering that institutional investors are strongly bound to following regulations and others’ objectives, I conclude that my findings will be more clearly identifiable if I relate them to the overreaction of individual/ private investors. Thus I will conduct interviews with available individual investors4 asking them about real live cases where they post factum realized that they misjudged and over or undervalued stock. The questions will be created and evaluated based on the findings in the previous step. It also makes sense to pilot test these questions, as also those from the previous part of the study, in advance, in order to acquire the necessary fluency.
It is essential that there are multiple other sources of mispricing of assets which need to be depicted and analysed and also the weaknesses of my method have to be properly identified in order to qualify my research as valid and to determine the degree/ perimeter of validity. After that is done, all the data will be evaluated under the lens of the interview findings, which will result in assigning the right weight to the influence of the different reasons for overreaction. The three experts will be questioned in order - not to confirm or reject the findings of the questionnaires and experiments - but rather to enrich the findings and approximate them to the real world. In addition I will search for weird patterns in the findings that might lead me to results or research direction of which I am not able to be aware of at this stage of the research.
Necessary to mention at this point is that more options to conduct the research were considered and a questionnaire appeared the most practical one. Originally the idea to create a stick exchange simulator was evaluated where the questions could be asked as direct financial information signals. This method would put the subjects in a very realistic situation. The approach was rejected due to the disproportionally large amount of time required to conduct the research. Preparation of the simulation, summary and evaluation of the data would be the smallest of the challenges. Rather problematic would be here to timely finish the simulation in order to maintain the focus of the subjects. Another obstacle would be to physically accomplish a simulation of that kind without a specialized software tool. These would imply to collect the bets of the participants per hand or on paper which would induce technical mistakes in the data collection. Thus a questionnaire which can be sent online appears to be the most purposeful tool for the research
4
iii. Further development of the research
The paper proceeds further as follows: An extensive literature review advising about the origins of the research idea and also summary of the behavioural explanations of mispricing will be provided first in chapter II. Next, the research goal will be made operational in the form of a questionnaire and an experiment and those two items will be pilot tested. The research findings and observations will be summarised. Following in chapter three is an analysis of the findings. It will be used to create a behavioural framework of mispricing. I will continue with showing if the stylized factors, representing the environment in which mispricing occur, are replicated properly comparing to the real world. I will conclude in chapter IV by analysing the strengths and weaknesses of my approach and I will suggest a plot for further research.
II.
Literature review and questions
A.
Underlying research
Jegadeesh and Titman (1993, 1999) analyse momentum relative strength strategies and discover that if one holds a stock for 3 to 12 months, which performed well in the past 12 months, an abnormal monthly return can be realised (if skipping the first week after portfolio formation). They argue that their findings can rather be attributed to overreaction than to delay of the stock price reactions or systematic risk. De Bondt and Thaler (1985) support the view that 3 to 5 years poor performance promises good performance for the same period in the future (contrarian strategy)5. Similarly interesting is the Jegadeesh (1990) and Lehmann (1990) observation that the bad performance in previous week or month can result in good one in the next week or month (again a contrarian strategy). The results of the last 2 contrarian strategies are still debated to be attributable to systematic risk and size effects or to short term price pressure or lack of liquidity in the market. Li and Rosser (2004), who separate the market in Fundamentalists and Noise Traders, summarize that asset market is marginally efficient (on the long run) but in shorter periods it is not efficient, which implies that noise traders cause inefficiency and thus enable the profitability of trend-following strategies (like momentum and contrarian ones). All those studies imply that it is possible to profit while exploiting mispricing
5 Please note the different duration of the observations! In this case a different direction of the wave is exploited.
which unanimously results from the overreacting of agents to information signals. This claim is supported by the opposing finding of Biais at al (2009) that overreaction causes portfolio inefficiency. It is sound since if one realises excessive returns someone else has to loose - arbitrage and speculation are zero sum games.
Further I will concentrate on overreaction, and some other psychological phenomena, and on how they function/ work in order to be able to conclude on what it is caused by. In addition we will also consider what else could be on the market besides overreaction that might cause the wave or bubble.
B.
Phenomena explaining mispricing & questions’ definition
i. Literature related to psychological constructs
Some behavioural models e.g. Edwards (1968) and Barberis et al (1998) claim that “individuals tend to underweight new information in updating their priors” which causes the so called “representative heuristics”: uncertain events or samples are identified by the degree to which they are similar to the parent population. In finance that is - firms that perform well will continue to perform well. That kind of believes, naturally, cause momentum. The questions in this section will resemble the questions used by Kahneman and Tversky (1973) in a research on representativeness. Thus those questions will not require further in depth explanation but just a brief comment.
The first one questions food poisoning and what might have caused it. Since in every developed country the food industry underlies strict hygiene norms, a Bayesian probability would suppose equal chance. Nevertheless in line with representativeness most students will attribute the poisoning to Kebab or dirty Beer glasses. In the second I will ask the students to identify a futurity member offering pictures of decently dressed gentlemen. The only futurity member is not dressed in line with the futurity dress code and one of the pictures is of a Nigerian student. The other pictures are of futurity-like dressed students which are in fact not futurity members. I expect that people will vote for the futurity-dressed-like persons. The third question is related to simple probability calculation6. I expect that if there is deviation from the Bayesian probability calculation it can be attributed to representativeness and is a misjudgement of the situation7.
A very strong explanation offers the overreaction/ under-reaction hypothesis (Jegadeesh and Titman 1999) which on its turn has also multiple explanations. Consistent with the over- under-reaction view is, as suggested by Ross et al (2008), conservatism - people adjust slowly their beliefs to new information. Barberis et al (1998) address the issue extensively and suggest that moreover conservative people would be endangered by under-reaction e.g. the new earning data will not be efficiently enough incorporated in the investors’ decision making process. To test conservatism we refer to the insights of Petty et al (1995) that conservatism is an attribute of attitude strength which will help to generate a scale of measurement. That insight will make it easier to be tested since now we can test attitudes and the extent to which they hold. In addition Maltby (1997) distinguished between three attributes of conservatism which will make the questions more directed.
6 Here Wikipedia offers comprehensive exemplification of a Kahneman and Tversky (1973) question and its
solution which I simplify further in order to make the questionnaire more fluent.
7
First one is linked to the degree to which the subjects inherit their parental values and the second one relates age to conservatism, where older people tend to be more conservative. Thus I will ask subsequently: “To which extend is your family traditional in their behaviour” and “Do you share the values of your parents”. The answers of the first one will vary form “to a large extent” to “not at all”. Thirdly, people that find the world threatening and challenging tend to be more conservative - thus linking conservatism with uncertainty. In order to discover that I will follow the example of Maltby (1997) who asked the subjects if they like one of two pictures - the one complex, the other simple. In order to avoid the aesthetic preferences of the subjects I will introduce concepts rather than pictures. I assume that those who are more conservative will prefer the simple picture/ concept. For the purpose I chose two organization charts. The main idea is that none of them can be “better” since no one will know what type of organization they belong to. Thus only the preference/ strength of attitude will influence the subjects’ choice.
Offerman and Sonnemans (2004) suggest that the reason for over- or under-reaction might be explained by recency, the tendency to overweight new information, or in the hot hand
hypothesis, where traders rely on the latest past record of the firm to forecast what will follow.
Interesting is that on the first sight the last findings contradict to the conservatism. It is then of importance how or under which circumstances those theories function in order to isolate the causing forces/ factors in the further research8. My suggestion is that in different cases people stick to different rules of thumb and thus in the one case the rather take recent information into account and in others they prefer to stick to their old believes. In general the paper of Offerman and Sonnemans (2004) find that people tend to overweight recent information in a way that makes them pessimistic about losers and too optimistic about winners, in the case of Recency. And in the case of the hot hand they, in their endeavour to recognize a trend in the past stock return, overestimate the autocorrelation in the series.
According the summarized information conservatism occurs when people tend to avoid complexity and uncertainty, meaning that Recency or Hot Hand is likely to occur mostly in times of a stable economy. Analysing the issue, I could argue that Recency or Hot Hand might also occur in the other extreme side - crisis. The desperation that one’s old methods/ information do not deliver good results any more and the quick changing environment would
8 Authors’ remark: related are properties of the businesses (stocks), market agents and markets (regulatory,
drive people to prefer more recent one. Anyways these considerations do not underlie the purpose of my research.
Easy to test appears to be the Hot Hand hypothesis since it is practically the endeavour of people to see trends/ series even if there aren’t any - I will ask directly the question if the subjects are inclined to see trends and patterns.
Recency consists of underestimating recent losers and overestimating recent winners, thus attitude questions can be made. For purpose of testing it I will firstly describe a situation of some positive and negative experiences in a city where the subject has a contact person (friend). Then I will ask him/ her about his preferences for location to visit at the end of his visit (where he/ she must make a more careful choice). I will offer 3 answer options: go to a place which he/ she did not like, ask an advice or go to a place that he/ she enjoyed. Here I assume that strong recency bias will lead to choosing the place with the good experience, weaker one will require asking for an advice. I suspect that no one will choose the place where they made bad experience. My main idea is that it is not objective to judge about an unknown city using couple of days of experience and I expect the students to know that.
Next I will use again standard Bayesian probability calculation. I offer to the students a table with the results of the competition between two hockey teams. I present the results for the first 10 months of the year and the results for the last two moths are missing the numbers for the first 10 months sum up to 42 points for the both teams. In the first seven months the numbers are randomly distributed and in the next three the second team is winning convincingly. Therefore in accordance with recency I suggest that most people will point out the second team as a winner due to overweighting of the last 3 months results.
that are too low. The Biais definition of overconfidence is easier to test and is also consistent with the findings of Vega (2006) who implies that the origin of information is irrelevant for the stock price reaction. For the purpose of questioning the psychological phenomena we have to go more in depth in the two definitions. According Daniel et al. self-attribution adds short-lag autocorrelation (to which he refers as momentum), short run earnings drift and negative future earnings and long run past performance. Its expression is in underestimating of ones own errors. In addition confidence grows when ones judgement is confirmed by the public opinions but doesn’t fall when the public opinion differs. Self attribution is also characterised by attributing of ones success to intrinsic reasons and ones failure to external factors. Thus, keeping in mind the characteristics of self-attribution, I will question the students to attribute a bad exam result to a reason. In order to use this information in statistical manner I will further rework the answers, turning them in dummies. I will more specifically classify them as factors that lie in the person of the interviewee and such that lie in the environment. I suspect that the failure will be attributed to environmental factors in consistence with the self-attribution theory. Following the same logic I will ask the students to give reasons why are they good at the sport they chose. If they say they are not good I will ask why and thus a similar evaluation will be expected.
The testing of overreaction would be best to accomplish within a time frame while observing real time reaction of people to information. This research lacks the resource time, so question about existing event or fact has to be generated. In the spirit of Biais et al. (2008), the subjects can be asked to give a self-assessment of their skills in one of the main subjects of the Financial Management field. In addition, since Biais argues that more confident people would take more risk I will question for risk. Risk is not clearly related to the topic mispricing in the literature. I intend to view how they relate to the psychological constructs which we described above. In order to question risk propensity/ aversion, I will use the example of Rohrmann (2004)9, who identifies four different domains of risk which are to be tested: physical - accidents (e.g., climbing, fast-driving, fire-fighting); physical - illness (e.g., smoking, unprotected sex, working with x-rays); financial (e.g., stock market investments, gambling, horse-betting); social (e.g., standing for elections, dating, revealing homosexuality). For the purpose of my research the questions will be less in number and altered minimally. As a
9 I will use the example of Rohrmann together with his questionnaire form 1997 for risk orientation. The above
complementing test I will ask for gender and ethnicity since those issues are not clearly related to mispricing either.
Jager et al (2006) add normative conformity, the desire to comply with the positive expectations of others, and informational conformity, the tendency of one to accept information from others. In the paper of Jager at al. (2006) this is not explicitly linked to mispricing but rather summarized as a general phenomenon. Anyway, we can derive logically that normative and informational conformity both would influence investors behaviour and would lead large masses of people to react in the same direction (invest in the same stocks) triggered by forces, that might be other than the fundamental values of the assets (a case of mispricing). The authors argue that people might also invest in order to satisfy social needs. In order to question conformity I can not ask it directly - everyone tends to think about him/ herself as of a special person - thus people would tend to “play” the non-conformist. I will put the subjects in a situation where they will observe a third individual and will have to decide what will be the expectations of the others towards the person where he will be put under group pressure for his decision. I offer three options as an answer which will indicate if the people feel need to conform and the degree of conformation. Further I will ask the subjects if the discuss about difficult personal decisions with their closest people. Here I hope that the subjects’ attention will be districted form the main question (if they actually absorb information form their environment when taking decision - information conformity) and will be rather focused on the personal character of the issue. Thus the answers will not be biased by the question of competence of the discussion partner, status etc. In order to question if there are social needs involved in the decision making process of the subject I will put them in a situation where a close person disagrees with them. In this case I will question the reaction of the subject.
Almost funny and grotesque but surprisingly convincing is the finding of Alter and Oppenheimer (2006). The attribute some type of very short term stock price fluctuations to
processing fluency. In the same spirit as people who are in a good mood are more optimistic and invest more they claim that on the very short run people would rather invest in
(at least outside France) . Here since the fluctuation is explained as driven by the rules of fluency I will refer to it as a misjudgement since it is not driven by objective factors. In this sense processing fluency could lead to “overreaction” to more fluent names or data and under-reaction to less fluent ones. This is mainly when company releases shares for first time since until then investors don’t much information about it. For the purpose of testing the fluency theory I will ask subjects to select 5 company names out of 10 available. For the purpose I will take five from those that Alter and Oppenheimer classified as difficult and five from the easy ones. Thus I will put the subjects on the scale from one to five according to the degree to which they underlie the fluency bias.
ii. Literature related to information signalling
In the researched cases of both, representative heuristics and conservatism, an underlying assumption is that investors switch between two “regimes” while conducting investment decision - stocks are mean-reverting or trend in a direction e.g. once they start raising or falling they go on. Another valuable research insight of Barberis at al. (1998) is that the psychological biases are caused by information signals which can be characterised by their “strength” and “weight”11, where the investors frequently underestimate the weight which cause the misjudgement. Furthermore the authors attempt to summarize the types of qualitative information which cause under- or overreaction in combination with conservatism and representativeness implying that people tend to underweight some types of financial information and overweight others. Unfortunately no other significant attempt to systematize the information signals influencing investors has been documented. People tend to under-react on some types of information and overreact on others.
Hong and Stein (1998) approach the explanation of abnormal return in momentum strategies from another perspective. They assume the existence of two types of traders: “news watchers” who obtain information about future CFs but ignore past developments (also referred to as “informed investors”) and “technical traders” - such who trade on the basis of historic price data ignoring fundamental values. Clearly no individual can be classified as 100% news watcher or technical trader but rather we can argue that some people concentrate
10
Authors’ remark: note that the way the research is conducted implies that the easy named stock will outperform the other one after IPO or when both penetrate the market simultaneously.
11 The strength here is the salience or extremity etc. and weight is rather the statistical significance etc of the new
more on the historical financial analysis and others on the effect of new developments on the stock market. The momentum starts with the news watchers who partially incorporate the information in the stock prices. After that the information reaches the technical traders who boost the growth further and push the prices above the fundamental value of the asset. After the prices expand beyond the fundamental values they eventually revert back which could be seen as an explanation of some contrarian strategies.
While speaking of overreaction and the other causes of mispricing, we need to admit the undisputable role of information and perception as fundament of these phenomena. Comand and Heinemann (2006) believe that the public information signals are “focal point of higher order believes”, which makes the market actors react more than justified. They conclude that
public information should be provided with maximum precision and even not to all actors. In
regard to our case with overreaction we could translate that when information is available to many not professional traders and is imprecise over- or under-reaction could be caused. Of interest for this paper is to investigate how imprecise information exactly creates overreaction or which actors should not receive information. Almost on the opposite, Clark and Polborn (2004) state that in some cases information does need to be precise due to crowding externalities. The actors are better of if they are not fully and precisely informed. Thus there are types of information that are not needed, though when available they might cause mispricing12.
If we put together all those findings we can conclude that investors are influenced more than justified by the salience of the signals rather than their significance, they switch between trend following and mean reversion perception of stock returns and the degree and precision of information can influence the investment outcomes.
Following Barberis et al. (1998) we can summarise some types of news which traders do not take seriously enough at first. Isolated quarterly earning announcements are assumed to have no significant explanatory power and are taken as regularly news updates. Earning announcements in general are also disrespected due to believe in mean reversion, thus stronger changes in those do not affect the future earning expectations enough. Stock repurchases dividend omission and seasoned equity offerings are also disrespected at first. An explanation
12As an example: Porsche decides not to disclose quarterly in order not to loose independence form the
of those phenomena is the general believe that earnings are more stationary than they are. Next, people tend to buy when prices fall and sell when prices go up.
Overreaction mostly follows series of good or bad news no matter of the weight of the information due to the fact that people overweight multi-period patterns. In addition Barberis et al. (1998) recognize that few earnings surprises (positive or negative) are able to make traders believe in a trend. Overreaction follows often salient and enthusiastic descriptions of products, management and companies.
III.
Findings and analysis
A.
Descriptive statistics
The sample taken consists of 21 of the approximately 50 Financial Management students which attended the course Advanced Financial Management in the winter semester 2009/2010 at the University of Groningen, the Netherlands. The constructs tested are to be found in below. Table 1 Concept Question Numbers Comment 1 Age - 2 Sex - 3 representative heuristics 1 to 3 4 conservatism 4 to 6
5 hot hand and recency 7 to 9
Those 2 constructs are tested together because the literature that suggests them doesn't offer clear differentiation.
6 self-attribution 10 & 11
Qualitative evaluation which will result in assigning a numeric value to the answers
7 overconfidence 12 to 15
8 risk propensity/ averseness 16
I will take an average of the scores of the sub questions
9 conformity and social needs 17 to 19
Conformity and social needs will be tested as one and an average of the 3 questions will be taken as a score for conformity needs.
10 processing fluency 20
11 under reaction to information 21 to 24
The degree to which people disregadr important information will be traced
12 overreaction to information 25 to 30
The degree to which people value less important information will be traced.
13 common questions 31 to 33
Valuable for the purpose of this study is the average score of the respondents to the variables and a Pearson cross-correlation between them in order to find out how they are related. The descriptive statistics is meant to provide the average values in order to make the variables comparable with each other and minimum, maximum values and standard deviation will show to which extent the answers of the different respondents are consistent with each other. The Pearson correlation is the most efficient and practical way to make assumptions about the relation of the variables to real life situation. A more advanced method would be to link the different variables in real cases to portfolio efficiency indicators through a regression but this approach is incomprehensively energy and time consuming which does not comply with the magnitude of this research.
Table 2
Descriptive Statistics
N Minimum Maximum Mean Std. Deviation
VAR00001 21 22.00 29.00 24.0952 2.14254 VAR00002 21 1.00 2.00 1.1429 .35857 VAR00003 21 .25 .75 .4814 .13320 VAR00004 21 .25 .67 .3971 .12386 VAR00005 21 .17 .75 .4543 .13441 VAR00006 21 .25 .75 .4381 .16744 VAR00007 21 .06 .88 .2576 .17754 VAR00008 21 .17 .69 .4176 .14460 VAR00009 21 .25 .67 .4676 .12255 VAR00010 21 .25 .75 .4405 .24881 VAR00011 21 .18 .89 .6600 .17950 VAR00012 21 .15 .87 .5695 .17548 Valid N (listwise) 21
the score, the more the respondent is biased. Thus we can assume that the respondents were biased but not heavily since the average is close to the middle score of 0.5. In the case of conservatism (variable number four), the slow adjustment of people’s beliefs to new information - the respondents are slightly more biased than in the case of representativeness (variable three), meaning that they are more exposed to the to not adjust adequately fast to new developments in their environment. The construct conservatism is relatively strong represented among the respondents - the test points out that the values for conservatism are higher than most of the values of the other variables.
In the case of Hot Hand and Recency (variable number five in the table) the respondents gave almost average answers, slightly tending to be biased. Hot hand is the phenomenon which causes traders to rely on the latest past record of the firm’s stock price in order to forecast what will follow. This phenomenon was investigated together with Recency - the relying on recent information rather than on all information available. After the research was conducted I could argue that the subjects are in general not exposed to the risk of overreacting to recent information significantly. Yet interesting is that the value for Minimum is very low compared to other variables implying that different and extreme reactions from investor to investor are possible.
Self attribution - the attribution of success to the own skill and effort and the attribution of failure to other people or external factors - is slightly more represented. The average is definitely higher than that of the most other variables but the minimum and maximum values are less extreme - thus the public is more homogeneous when considering self attribution. On the opposite, considering the bias of overconfidence (variable seven) - a stronger reaction to information than justified, due to higher trust in the own judgement - I could argue that the subjects are extremely exposed to misjudgement since the average indicates the strongest exposure. In this case not only the average is very significant but also the minimum and maximum values have the highest deviation, which indicates that people are capable of very various reactions in different situations. Important is to mention again that for the purpose of this study we recognize that it is enough for a small group of investors to be convinced in the correctness of an investment decision in order to start a momentum on the market. Therefore not only the average values but also the extreme values are meaningful for the analysis.
assume that the respondents are generally risk lovers whereas the average is interplay of the answers of extreme risk lovers and extremely risk averse students. The ninth variable - conformity & social needs - expresses the degree to which the respondents are dependent on their environment for the choices they make. Here the average is close to the middle but still slightly underneath which implies small dependency on the social environment.
The phenomenon of Fluency (variable number 10) is, opposing to the expectations and that what was proven by Alter and Oppenheimer, not so explicitly represented by the subjects, nevertheless it remains a matter. The mean, even though pointing towards biased respondents, is not far from the average, neutral level. The next two variables (number eleven and twelv) measure respectively the tendency of people to, either disregard and not appreciate enough some essential for the company’s profit news feeds or information, or overvalue other, less essential information. When evaluating the findings I will consider not only the average values scored from the respondents but also the minimum and maximum scores in order to be able to make conclusion about the level to which people tend to go into extreme in their preferences. Even though the first variable questioned the importance of valuable according most scholars information and the second variable less valuable news signals, the average difference between the responses on the two variables is very small compared to the range of the answers. Indeed the minimum and maximum values of the respondents stand very far apart form each other in both variable eleven and twelve. Also interesting is that the minimum value of variable eleven is way below the average of variable twelve, however variable eleven has still overall slightly higher values than twelve. Thus the respondents reacted in a right way (on average) but the difference is not too large and the diversity of the answers points out that there are a certain number of people who are by far exposed to misjudging financial information.
significant for the stock price development information the most but definitely appreciate the salient news signals. Thus they are exposed to the risk to focus on unessential information while making their investment decision. The respondents are pretty homogeneous in their answers to conservatism, fluency and social needs behaviour. Furthermore, the strongest represented psychological bias - overconfidence yields the highest diversity in answers. My focus on the diversity is reasoned by the fact that actually the diversity of the interests and the perceptions of the market players is the reason why the stock market is difficult to gain insight into and to predict.
B.
Correlation
Keeping in mind the mentioned considerations I will take the next step in the analysis of the variables and I will investigate how the psychological phenomena relate to each other and therefore a Pearson cross-correlation follows.
Pearson Correlation -.060 -.096 -.151 -.149 .224 .007 1 .062 .122 .215 -6.770E-1 -6.908E-1 Sig. (2-tailed) .797 .677 .512 .519 .330 .976 .789 .598 .350 .001 .001 VAR0000 7 N 21 21 21 21 21 21 21 21 21 21 21 21 Pearson Correlation -.027 .499* .498* .437* .580** .404 .062 1 -.330 -.355 -.222 .000 Sig. (2-tailed) .909 .021 .021 .047 .006 .070 .789 .144 .114 .334 .998 VAR0000 8 N 21 21 21 21 21 21 21 21 21 21 21 21 Pearson Correlation .149 .111 -.240 -.067 .090 -.025 .122 -.330 1 .344 .085 -.206 Sig. (2-tailed) .518 .633 .296 .774 .697 .914 .598 .144 .127 .715 .371 VAR0000 9 N 21 21 21 21 21 21 21 21 21 21 21 21 Pearson Correlation .105 -.320 -.499* -.079 -.085 .141 .215 -.355 .344 1 -.174 -.353 Sig. (2-tailed) .651 .157 .021 .734 .713 .542 .350 .114 .127 .452 .117 VAR0001 0 N 21 21 21 21 21 21 21 21 21 21 21 21 Pearson Correlation .231 .085 .001 -.050 -.339 -.079 -6.770E-1 -.222 .085 -.174 1 .696** Sig. (2-tailed) .313 .713 .995 .829 .133 .735 .001 .334 .715 .452 .000 VAR0001 1 N 21 21 21 21 21 21 21 21 21 21 21 21 Pearson Correlation .168 .208 .210 -.153 -.073 .016 -6.908E-1 .000 -.206 -.353 .696** 1 Sig. (2-tailed) .467 .366 .360 .508 .752 .946 .001 .998 .371 .117 .000 VAR0001 2 N 21 21 21 21 21 21 21 21 21 21 21 21
**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).
Before going deeper in the relations between the variables I will explicitly link the variable numbers from the table above to their descriptions. Variable 1 is age, 2 is gender, 3 is Representative Heuristics, 4 is Conservatism, 5 is Hot Hand hypothesis and Recency, 6 is Self Attribution, 7 - Overconfidence, 8 - Risk Propensity, 9 Social Needs and Conformity and number 10 is Fluency. Variable 11 measures the degree to which people appreciate significant information and variable 12 measures the degree to which people value salient information which is not necessarily of high importance. I will analyse the variables row by row going upside down, skipping repetitive entries and will discuss every significant relation between the variables or the lack of a relation, since that can also have some explanatory power.
representativeness. The gender variable was constructed so that the higher number corresponds to female participants. Positive relationship mans that women are less biased by the construct representativeness. Thus men tend to rather identify uncertain events with similar ones from their experience and women are more open-minded and less biased. Thus men are generally stronger exposed to the risk of misjudgement by this criterion. Similar relations with gender have two other concepts - Hot Hand and Recency. Men tend to bet more on a “winning stream” or with other words men tend underestimate the random nature of the environment. Surprisingly gender also relates to risk propensity. The female respondents showed slightly higher preference to risking. Therefore they were slightly more biased by fluency which implies that women would rather prefer easily named stocks at a first glance.
Hot Hand and Recency are related to Representativeness, which could imply that the both phenomena, even though researched by different scholars and having different foundations, might have commonalities. After analysing the definitions I can argue that Hot Hand and Recency are a kind of Representativeness. Since unknown events or samples are associated with the degree to which they resemble known ones in the case of representativeness, and Hot Hand and Recency are the identifying of upcoming outcomes with previous ones, one can see that the term representativeness can be viewed as broader and more general term, whereas the other two terms concern specifically the time frame of events. I will conclude that in the different situations the market participants can have the incentive to apply representativeness in one of its forms and considering the definition of the term I can suggest that representativeness occurs in times of high uncertainty and stress on the market - thus when people cannot handle the information available and have doubts about the outcome of their actions they would prefer to “play safe” and make use of experience they made in a similar situation.
summarized as someone who has a more general view on things, thus generalize and move fast forward - risk is associated with certain high dynamics in lifestyle, therefore risk loving and generalizing (through representative heuristics) seem to belong together which would explain the relation that we observe in the statistical data set.
Even more difficult is to argue why does Representativeness have such a strong negative relation to Fluency (Hot Hand has also negative relation but it is considerably smaller so it will not be analysed any further). The more people are biased by Representativeness the less they are biased by Fluency. Explanation could be found again in the nature of the both heuristics. Representativeness is a way of summarizing and generalizing information in order to gather it quickly whereas Fluency is a way of avoiding the gathering of information and rather acting out of emotion of liking. Thus Fluency is more similar to a low attention involving purchase and Representativeness is a way of making difficult decisions easy.
Interesting is the lack of any relation between Conservatism and Hot Hand which means that the occurrence of the both phenomena is absolutely independent. Thus they can occur at any time when provoked simultaneously or one by one. Conservatism is strongly related to Risk propensity. Thus the more conservative the respondents the more they are risk lovers. Since Conservatism is actually the relying on already available “old” information for making decisions, it can be that those subjects feel safe in their information environment and act overly assertive.
In the case of Self Attribution impressive is the lack of relation to Overconfidence apparently the both phenomena do not occur in the same persons or under the same circumstances. Self Attribution is positively related to Risk Propensity - it is obvious that when people would always believe that their success is due to reasons in them they would be more brave and willing towards taking risks. Important to note here is that even though the correlation is not strikingly strong for the purpose of the research we are more interested in the directions rather the strength of the initial thrust. The reason is that mispricing actually results form the cumulative action of many agents, thus a slightest notion in one direction could lead to over- or under pricing.
Risk Propensity is slightly negatively related to Social Needs and Conformity and Fluency. It does not make sense to examine close the relation between Fluency and Risk because the explanation could be way too personal in order to be generalized. That what the relation between Social Needs and Risk tells us is that people that are strongly dependent on their environment avoid to risk.
There is a very strong correlation between variable eleven - the variable measuring the degree to which people react to significant information and twelve - the degree to which people react to salient information. People value information differently and to a different degree, nevertheless they take both significant and salient information signals with a respectively similar degree of attention: those that valued very high significant signals valued also very high salient ones and the opposite.
C.
Open questions
Out of the 21 respondents only few filled in the open questions which were respectively: investment choice criterions, if those alter in case of crisis and investment preferences (long-, short-, or medium term investments). Considered were mainly respondents who prefer medium and short term investments since the investment preferences of those who prefer long term investments are not so strongly related to the momentum strategies - the starting point of the research.
Interesting is that most of the investors address rather subjective issues and have less technical considerations which is sound with the idea of mispricing. It is valuable to follow through some of the responses sine they are also supporting some of the findings mentioned in the analysis of the relations of the questioned variables. Image of the company, image of the products, image of the ownership and management and especially image of the board of directors, are the first and also some of the most frequently used answers on the question about the criterions for choosing a stock. At this point one can observe that those criterions can not be precisely and quantitatively evaluated. We also notice that the concept addressed by the eleventh variable is confirmed by the findings - indeed many of our respondents pay a lot of attention to information signals which are qualified by Barberis et al. (1998) as of secondary importance.
company is into. Addressed was also the corporate governance of the company but these was mentioned by a respondent who was interested in medium to long term investments. In addition many of the respondents would look at what others invest in or at what “experienced investors” invest in. That comes to strengthen the mentioned above. Risk and return considerations are also represented but not that frequently. Some of the respondents would also examine the investment rating of the company they are considering. Some would also analyse the relevant profitability and liquidity ratios in order to conclude about the “financial health” of the company, the market prospects and relevant investment in R&D were also addressed - thus the study didn’t lack on respondents that were stronger fundamentally oriented. Interesting is that in a situation of turbulent economy or difficult market conditions half of the respondents decided to be “more careful”, “more risk averse” and to invest in “stable and safe” companies. Here important is that determination to drastically change ones behaviour by change of the relevant stock environment and rely on “other’s opinion” in case of uncertainty. Those facts explain why and also support the “how” mispricing happens.
D.
Analysis summary and added value
Our respondents are typically overconfident, conservative risk lovers who are also slightly biased by self attribution. They become less conservative with the age. Females are more biased by Representativeness, Hot Hand and are more less risk averse. People biased by Representativeness are also biased by the Hot Hand phenomenon; they are risk lovers and tend to be less biased by Fluency. I also suggested that the Hot Hand is a form of Representativeness which particularly deals with the time frame of events. My conservative respondents, as also those, stronger biased by the Hot Hand were also less risk averse. The people biased by Overconfidence were rather risk averse unlike those biased by Self Attribution. Generally the risk lover were less biased by the Social Needs bias and Fluency where Social Needs and Fluency are positively correlated. People do sometimes value information that is salient more than information that is rather significant. Some people value information to a higher degree than others - they are attributing more importance to all news feeds and would tend to rather undertake something when the situation changes, whereas others tend to take news as less influential and thus tend to not react on them.
February 2010 British Airways announces “surprising third quarter profit ”. If one pays more attention to the news feed it becomes obvious that this profit is less due to an increased profitability but rather a consequence of a heavy cost cut predated by a very long period of losses. Thus this profit can not be a reason for a serious stock price rise. The investors on the market start buying and the stock price rises throughout the days after the announcement. When the price reaches a certain level, above the starting point from before the announcement, it falls back down (please find below).
Table 4
The psychological biases identified by the respondents are not so easy to find on the market. As a market participant, one lives in the market environment; then it is easy to observe
13
which spirit is dominating the environment, the group opinion and the press. Thus one can make conclusions on what he/ she should beware of. Is the common opinion “we can achieve everything”? Then one can think of people raising way to confident. Is the common mood more in favour of preserving the traditional values, one can expect that people will underestimate new coming information; in addition one could think that in this case people would rather prefer taking risks.
There turns to be no other way to predict the short term market developments but through communicating with the market participants and critically and objectively observing the market environment. The phenomena discussed in this paper were researched separately before, but the uniqueness of this work is to consider them together and in interplay, analysing the relationships between them.
IV.
Conclusion and implications for further research
and are represented on a personal level in every individual to a different extent. Thus we can argue that those phenomena must occur in larger masses of people simultaneously in order to trigger momentum, from which follows that they can be traced in the population of market participants.
The first most distinctive finding of the research is the level to which the future investors are exposed to overreaction based on the psychological biases discussed. Especially interesting is to discover the level of exposure to the different phenomena in comparison. This is essential in order to assigning the weight of the separate constructs for the determining the direction of mispricing in a real life situation.
As mentioned, the psychological biases were never researched together - this led to the second distinctive finding - the relation of the different phenomena to each other is described and thus a dynamic picture of the behavioural aspects of mispricing was created. One can expect to rather observe the correlating phenomena to co-exist in one and the same situation on the stock market. Thus the environment of mispricing can be described and this description can be universally applied to real life case analysis.
The research answered some questions but also opened new ones. We cannot argue about the explanatory power of the psychological phenomena because they affect the markets in a different way and to a different extent every time. In order to see how far and how much the different phenomena affect the market, in depth analysis of the information signals on the market should be made, using both official and private sources. In the frame of this research I did not find any suitable classification of financial information in the literature and I think that this would be a major key for gaining a better understanding of the market. This classification will have to throw some light on the issue how do the different information signals under different economic circumstances precisely affect the market participants. This promises to be useful but also extensive research, diving into group psychology and financial science, the foundations for which are laid by my research on mispricing.
The questions left to be answered in further research are related to information classification and availability. This paper was limited by the fact that no extensive and in depth classification of financial information is attempted before. If one would have a systematic summary of all possible kinds of information the exploration of the different biases could be combined with the information signals that cause their occurrence on the market and thus a better and in depth understanding of the market could be gained.
A second direction of further research is imposed by the idea that different market actors react differently to market information. Private investors are different than institutional investors, speculators or noise traders - different from fundamentalists, etc. Since we know how mispricing and overreaction function it is necessary to know also who is on the market in order to predict the explanatory power of the biases or the level to which they would influence the prices. That can be done through determining the volumes traded by less experienced or noise traders
Both, knowing who is participating on the market and on which types of information signals do they react, would on a long run decrease the gain from speculation and increase market efficiency thus enhancing long run sustainable investment.
V.
List of references
Articles:
Jegadeesh and Titman; Returns to Buying Winners and Selling Looser: Implications for Stock Market Efficiency; The Journal of Finance; Vol. XLVIII, No. 1 March 1993
Jegadeesh and Titman; Profitability of Momentum Strategies: An Evaluation of Alternative Explanations; May 17, 1999
Hoffmann, A. O. I., Delre, S. A., Von Eije, J. H., & Jager, W.; Artificial Multi-Agent Stock Markets: Simple Strategies, Complex Outcomes. In C.Bruun (Ed.), Advances in Artificial Economics: The Economy as a Complex Dynamic System (pp. 167-176). Heidelberg: Springer Verlag. 2006
Massey, C. & Wu, G.; Detecting Regime Shifts: The Causes of Under- And Over-Reaction; 4 June 2004
Goldstein, K. & Sheerer, M.; Abstract and Concrete Behaviour - An Experimental Study with Special Tests; Psychological Monographs; Vol. 53, No. 2 Whole No. 239; 1941
Jager W., A.O.I. Hoffmann and J.H. von Eije; Individual Investors’ Needs and Conformity Behavior; University of Groningen; 28 Nov 2006
Eisenhardt, K. M.; Building Theories form Case Study Research; Academy of Management Review; 1989; Vol. 14, Nr. 4, 532-550
Bondt, W. F. M. de and Thaler, R,; Does the Stock Market Overreact?; The Journal of Finance; Vol. XL, No. 3, July 1985
Jegadeesh, N.; Evidence of predictable behaviour of security returns; The Journal of Finance; 1990
Lehmann, B. N.; Fads, martingales and market efficiency; Quarterly Journal of Economics 60, 1-28; 1990
Li, H. and Rosser, B.; Market dynamics and stock price volatility; European Physical Journal; B, 39, 409-413; 2004
Edwards, W.; ``Conservatism in human information processing,’’ In: Kleimutz, B. (ed.), Representation of Human Judgement, John Wiley and Sons, New York, 17-52; 1962
Barberis, N., Shleifer, A. and Vishny, R.; A model of investor sentiment; Journal of Financial Economics 49(3), 307-343; 1998
Hong, H. and Stein, J.; A unified theory of underreaction, momentum trading and overreaction in asset markets; Working paper; MIT Sloan School of Management; 1998
Offerman, T. and Sonnemans, J.; What’s Causing Overreaction? An Experimental Investigation of Recency and the Hot Hand Effect; Forthcoming Scandinavian Journal of Economics; June, 2004
Biais, B., Nosic, A. and Weber. M.; Overreaction and Investment Choices: An Experimental Analysis; July 21, 2009
Clark, C. R. and Polborn, M. K.; Information and crowding externalities; Economic Theory 27, 565-581; Springer-Verlag; 2006
Alter, A. L. and D. M. Oppenheimer; Predicting short-term stock fluctuations by using processing fluency; Department of Psychology; Princeton University; May 2006
Kahneman, D. and Tversky, A.; On the psychology of prediction; Psychological Review; Vol 80(4), 237-251, Jul 1973
Petty, R. E.; Krosnick, J. A.; Lawrence Attitude Strength, Antecedents and Consequences; Volume on Attitudes and Persuasion; The Fourth Ohio State University; Erlbaum Associates; 1995
Maltby, J.; The concurrent validity of short measure of conservatism among English students; School of Health and Community Studies; Sheffield Hallam University, England; March 1997
Daniel, K., Hirshleifer, D. and Subrahmanyam, A.; Investor Psychology and security market under- and overreactions, Journal of Finance 53, 1839- 1886; 1998
Vega, C.; Stock price reaction to public and private information; William E. Simon School of Business Administration; University of Rochester, Rochester, NY 14627, USA; April 2006
Rohrmann, B., Risk Attitude Scales: Concepts and Questionnaires, Project Report, University of Melbourne, Australia, August 2004
Online sources:
Wikipedia (question about representativeness involving simple probability calculation): http://en.wikipedia.org/wiki/Representativeness_heuristic
Rohrmann, B., homepage: www.rohrmannresearch.net
Online Financial Times; News Feed - British Airways: http://www.ft.com/cms/s/0/597529c2-1225-11df-8d73-00144feab49a.html?nclick_check=1 - on the 5.02.2010 British Airways surprisingly declares surprising third quarter profit.
Books:
