More is worse or more is better? : the impact of thinking too much on investment decision making in a stock market setting

(1)

1

More is Worse or More is Better?

The Impact of Thinking Too Much on

Investment Decision Making

in a Stock Market Setting.

Master’s Thesis Behavioural Economics and Game Theory

August 2014

Laurentiu Ghita

10599355

(2)

2

Abstract

It is a common preconception that investment decisions are difficult and should require a careful weighting of the attributes of each option beforehand, especially when investing in the stock market. An informed investment decision is usually the term that refers to a good investment decision. My study explores the impact of thinking too much by analyzing and correlating an extensive amount of information when making an investment decision in a stock market setting. Participants involved in the laboratory experiment conducted had to construct a portfolio of 3 out of 9 stocks made available using different amounts of data. The data consisted of 11 indicators that were calculated beforehand. The participants in the “thinking too much” condition had access to all the 11 indicators while the participants in the “not thinking too much” condition had access to only a limited amount of information and had to choose only 3 indicators.

Although the participants in the “not thinking too much” condition attained an average rate of return of 5.33% and the participants in the “thinking too much” only 3%, the results obtained do not empirically support that thinking an investment decision too much has detrimental implications. My study can only conclude that decisions taken by engaging in thinking too much are at least as good (but not always better) as investment decisions which are not thought that much. Moreover, there was no difference between the level of satisfaction and regret experienced by participants by engaging either in thinking their investment decision too much or not that much.

(3)

3

Part 1. Introduction

For more than half a century, researchers have been primarily concerned with studying one of the most interesting topics, which is the process of human decision making. During this time, they mainly focused their attention in two directions – thinking about a decision too much or too little. Considering the latter, some researchers have tried to underline the drawbacks of how people often guide their decisions using heuristics, because of limited capacity, time and knowledge (Kahneman and Tversky, 1974; Gigerenzer and Goldstein, 1996; Kahnemann, 2003) and how people could improve the outcome of the decision through more careful analysis (Raiffa, 1968). However, other researchers have argued that thinking too much has more negative implications then positive ones. The aim of this thesis is to try to prove that engaging in thinking about a decision too much, does not necessarily have a positive impact on the desired outcome. More specifically, my study tries to show that thinking an investment decision too much, could have adverse consequences. These proposed effects of thinking one’s investment decision too much include aspects like, a decrease in the outcome attained and in the motivation to commit to a choice (Baron and Kenny, 1986; Iynegar and Lepper, 2000; Gao and Simonson, 2008); a decrease in the level of satisfaction experienced by the participants (Iyinegar and Lepper, 2000; Sagi and Friedland, 2007; Schwartz et al., 2002; Wilson and Schooler, 1991; Wilson et al., 1993); and an increase in the level of regret and disappointment (Simonson, 1989; Shafir et al., 1993; Schwartz, 2000; Schwartz et al., 2002). The results obtained in the current study do not offer any empirical support to either of the claims previously mentioned. It was found that thinking an investment decision too much does not lower the satisfaction or increases the level of regret and also does not have an adverse impact on the outcome of participants.

Although there are multiple papers to explore the impacts of thinking too much and thinking too little on one’s decision outcome, still, it is not clear which of the two ways of making a decision is considered the best. For instance, upon gathering 63 published and unpublished studies on this topic, Scheibehenne et al. (2000) found a mean effect size of D=0.02.

Furthermore, none of the papers written so far, considered the implications of thinking too much on the investment decision process. Therefore, my study could be seen as the first to explore the positive or negative impacts of engaging in thinking an investment decision too much.

On one hand, proving the positive effects would not have the same impact magnitude as opposed to the negative ones. Nowadays, pension funds, brokers and investors are already engaging in thinking too much and acquiring as much information as possible to guide their investment decision. Obtaining a result that would indicate a positive effect of thinking too much would be just a confirmation of the current method of making investment decisions.

On the other hand, demonstrating the negative effects of thinking one’s investment decision too much would have important theoretical and practical implication. From a theoretical outlook, it

(6)

6

would challenge the classical economic assumption which states that “more is better”. More information and more effort employed in thinking about a decision should result in a higher quality of the decision.

But is this really true? As it was already previously established in the works of Allen, 1990; Arrow, 1996; Arrow, 1999; the information has certain features that could prove to be handy, such as easiness to be created and spread. However, the fact that it could also be hard to trust and to control, along with the elongated process of thinking that it involves, already can provide some insights into the negative implications that the large amount of information could carry with it. From a practical perspective, such investigations could lead to the construction of new models or new approaches to investing in the stock market, which might prove to be completely different from the ones that are currently being used. The implications of thinking too much have been explored with respect to consumer behavior, marketing strategies, advertising, job selection and many others, but none of the studies so far analyzed the effects, with respect to financial investments. The literature written so far on these topics, however, point out that assortment structure can trigger thinking too much, or more specifically, by having too many options or evaluating attributes, as it will be shown below.

1.1 Too many options

In a series of experiments (Iyengar and Lepper, 2000; and Sagi and Friedland, 2007), showed how having too many options to choose from can have multiple negative impacts on the motivation to choose and also on the level of satisfaction and regret experienced by the participants with respect to their decisions. Iyengar and Lepper asked people to choose from an array of 6 or 30 chocolates. The results obtained were in line with what Schwartz (2004) referred to as the paradox of choice. Although having multiple options to choose from would initially seem attractive, the participants who were exposed to 30 types of chocolates reported being significantly less satisfied and experienced more regret then the participants exposed to only 6 flavors of chocolate.

Moreover, the participants in the “6 types of chocolate” condition purchased more boxes of chocolate than the participants in the “30 types of chocolate” condition. It seems that thinking too much did not only make the participants’ decisions harder, but ultimately they did not choose at all. In a sequence of experiments, the research of Mogilnev et al. (2008) shows how categorization can have the exact opposite effect, while still having multiple options to choose from. They categorize a multitude of different types of magazines and different types of coffee and ask people to choose one from each category. They discovered a positive relationship between the number of categories and the level of satisfaction experienced by the participants with their final decision, even though the categories did not offer any information about the options in the assortment.

The negative implications of thinking too much, when the options are equally attractive was another approached topic which can be traced back to the problem referred to as “Buridan’s ass”, in which a starving donkey is put to choose between two piles of hay. Because the donkey is thinking

(7)

7

too much which pile of hay to select, it ultimately starves to death. The consequence experienced by the donkey is a little extreme, but it helps us grasp the magnitude of thinking too much when the options are equally attractive.

1.2 Atributes

Linville (1982) and later on, Wilson and Schooler (1991) provided evidence to suggest that people usually decompose a stimulus in different attributes when they evaluate it. Linville asked his participants to rate different types of cookies, while considering 2 or 6 attributes. The participants who had to consider 6 attributes erroneously gave weight to irrelevant attributes which resulted in more moderate evaluations when compared with the participants that had to consider only 2 attributes. It seems that there is a negative relationship between the number of attributes considered and the outcome of their evaluation when choosing among consumer goods.

Another interesting study that explores how different weights are applied when a stimulus is split in multiple attributes represents the paper of Weber et al. (1988). They asked graduate students from a German business school to evaluate future jobs. After analyzing the data, they found that the detailed parts of the job were weighted significantly higher by students than the less detailed parts. The explanation for this overweighting bias was due to an increase in perceived salience and availability of the attribute which was described in more detail.

Meantime, the decision theory in Economics and Psychology is formally being shaped by the notion that homo economicus is a rational and self-interested actor who has the ability to choose the best alternative available (von Neumann and Morgenstern, 1947). This would suggest that people have stable preferences and act according to them. Following the research of Ariely et al. (2003) and Ariely and Norton (2008) would seem to contradict the neo-classical view. In the later mentioned paper, they argue that actions do not only reveal preferences for attributes, but rather create them. This, in turn, has two important consequences. First, situational factors can influence people’s decision and second, when individuals asses their preferences with respect to their attributes, they also take into account the utility experienced with their previous decisions.

This is why when decision-makers are faced with the need to choose, they often seek reasons to justify their options to others and to themselves. In a series of experiments done by Simonson (1989) and Shafir et al. (1993) many decisions resulted from an attentive evaluation of attributes in which people tried to choose what they considered the best choice. In trying to choose the best option, they searched only for justifiable attributes that would explain why they chose one alternative over the other or tried to internalize the criteria employed by others who were expected to evaluate them. This in turn triggered thinking about their decision too much which had a negative impact on their satisfaction level, their preferences and ultimately on their decision making process (Nisbett, 1977 and later Wilson and Schooler, 1991; Wilson et al., 1993).

In the last study mentioned, Wilson et al. (1993), the participants had to choose among 5 posters. In one condition they had to provide reasons for their choice, while in the other they were asked only to choose. The subjects that were asked to provide reasons were less satisfied with their choice

(8)

8

than the participants in the “no reason” condition. This was because asking for reasons underlined attributes of the posters that were not important for their initial evaluation. They chose the posters without considering attributes at all, but when they were asked for reasons, this probably led to changes in their preferences and ultimately with being unsatisfied with the decision taken.

1.3 Decision making strategy

Another way in which thinking too much is triggered depends not only on assortment structure, but also on the strategy employed by people, a maximizing or a satisficing strategy. As the name of the strategy points out, a maximizing strategy is the one in which one is trying to choose the best possible option, while satisficing is a blend of sufficing and satisfying.

Although among these two strategies, maximizing seems to be a better strategy, it is also the one that might lead to thinking too much. Why is that? It encourages the gain of more information and afterwards, determines a shift through it all while making comparisons between the available options. But by doing this, the decision making process would prove to be much harder and ultimately a less satisfying experience (Schwartz et al., 2002).

There is a way of avoiding the pitfalls of thinking too much, while employing a maximizing strategy. One could eliminate the options not wanted. In this way, one would have a much smaller sample from which one could choose from and could ignore the rest of the insignificant options. The research of Gilbert and Mosteller (1966) and Hertwig (2010) provide evidence for this solution. Gilbert and Mosteller (1966) proved in their classic experiment that the best way to choose a secretary among 100 applicants is to first select 37 and then make your final choice.

Hertwig (2009) showed that small samples amplify the difference between the options, making the options more distinct and the decision making process much easier. Another solution for this maximizing strategy would be, of course, to adopt a satisficing strategy. In this way, obtaining more information or making comparisons between the options would be avoided, because one would choose an option that would satisfy a certain threshold. Once this option would be discovered, the search among the alternative options would stop.

Another strategy that leads to thinking about a decision too much is called introspection, in the form of deliberation or otherwise. There are quite a few papers that explore this phenomenon, such as Simonson (1989), Wilson and Schooler (1991), Shafir et al. (1993) and Wilson et al. (1993). The main findings of the researchers mentioned above are that introspection interferes with the participants’ preferences and ultimately makes them to experience more regret and less satisfaction than the participants who did not introspect. The reason for this is that verbalizing or just thinking of reasons on behalf of the decision taken can indicate features that were not central to the previous evaluation made or creates doubts with the final decision reached. The obvious solution for this strategy would be not to introspect, but to select using emotional responses or intuition, which has proved very useful in other cases. So why do not use it more in making a decision?

(9)

9

Moreover, thinking too much is triggered if the decisions that people make are relative compared to absolute, or if the decisions are final or not. Considering the relativity of the decisions, Gao and Simonson (2008) showed that purchasing likelihood depends on the order of decision made: to either select a certain product or to buy a certain item. If people first decide to purchase an item and then figure out which option to select, they will avoid the pitfalls of thinking too much, because they will practically have fewer options from which to select. Imagine a store with 5 T-shirts in which a customer enters. By first analyzing the options and then deciding which to buy he practically has 10 options (T-shirt 1 – to buy or not to buy, T-shirt 2 – to buy or not to buy), but by deciding first that he is going to buy a T-shirt and then select which one, he will have only 5 options (Buy either T-shirt 1, 2, 3, 4 or 5). If people do not follow this order of buying and then selecting, but they inverse it they will struggle with their decision making process and not only feel less satisfied, but also, most likely, will not be willing to buy at all.

And last, with respect to the decision being final or reversed, there is a study made by Gilbert and Ebert (2002) in which they asked students who followed a photography course to print 2 photos. They had to keep one and hand in the other one. Some of the students were told that their decision could be reversed after a few days, while others were told that their decision will be final. After a few days they measured the satisfaction levels of the participants and the potential attractiveness of having their decision being reversed. Although most of the participants reported preferring to have a reversible decision than having an irreversible one, none of the participants actually changed their photo. It seems that wanting their decision to be reversible acts more like a moral comfort in case they change their mind in the following days. Even though this reversible characteristic of the decision acts more like insurance without actually exchanging the photo, the participants ultimately experienced more satisfaction in the “irreversible decision” condition, than in the “reversible decision” condition.

Considering the impressive amount of research already done with respect to the impact of thinking too much in other areas then investments, negative effects were expected to be found regarding investment decisions and engaging in thinking too much. More specifically, it was expected that thinking too much will have a negative impact on the outcome of participants and also that it will lower the satisfaction and increase the level of regret experienced.

The current study does not find any support to the idea that thinking too much has detrimental effects on the outcome of an investment decision. Indeed, it showed that thinking too much is not necessarily the best way to make an investment decision, but if employed you would still attain a positive outcome. Moreover, thinking an investment decision too much will also not have a negative impact regarding the level of satisfaction or regret experienced, although much of the previous research taken into consideration indicates differently.

(10)

10

Part 2. Methodology

2.1 Overview

The purpose of this experiment is to determine whether thinking too much can have a detrimental effect on a person’s investment strategy, in a stock market setting. To achieve this task, a sample of participants was selected. The participants involved in the trial have been exposed to different types of information. In order to examine the process of reaching a certain decision, the participants have been split into two groups, depending on the level of information supposed to be given prior to the investment decision required to be taken. The control group received an abundance of

information, to trigger thinking too much their decision; while the treatment group was given less information, so that they will not think their investment decision too much. After completion of the

task, they had to fill out a specific questionnaire.

2.2 Participants

Each of the two groups (N = 15 and N = 15) comprised students enrolled in a postgraduate program, in Economics and Business at the University of Amsterdam. The participants were recruited randomly in the University library and were approximately the same number of males and females (17females and 13males). Each participant was informed in advance that after the filling out the questionnaire and making the conclusion of the task, each of them will receive an amount of money, depending on the results achieved.

2.3 Materials

The experiment was conducted using a between-subjects design, in individual sessions, and implied a stock market game. The participants had to choose 3 out of 9 stocks to invest in, by using a certain level of information given to each of them. The participants were randomly assigned to one of the two sample groups.

The total number of stocks was chosen from the New York and NASDAQ stock exchange. To avoid bias and external influences, such as the names of the companies and the informed players’ advantage, the chosen stocks’ have been renamed. Each company was randomly associated with a color name, as described below in Table 1. Moreover, using the experiment’s instruction sheet, the participants were also informed of the arbitrarily assignment of names.

(11)

11 Table 1. The selected stocks color associations.

The stocks were picked based on the dynamics of the price in the period of just one year from May, 2013 until May, 2014. In order not to have a dominant performing stock, the companies were chosen to satisfy the criterion of having their price changed in a specific closed interval as shown in the Table 2, below.

Table 2. The price change intervals used to choose the stocks.

Interval of percentage changed in price used

in stock picking Stock Symbol Stock Name Color Assigned

Percentage change in the period from May 2013 to May 2014

30% -> 40% BA The Boeing Black 34.41%

20% -> 30% DIS

The Walt Disney

Company Red 29.98%

10% ->20% JNJ Johnson & Johnson Yellow 17.18%

0% -> 10% MCD McDonald's Corp. Pink 4.40%

0% IP

International Paper

Company White 0.04%

-10% -> 0% C CitiGroup Orange -9.06%

-20% -> -10% Dk Delek US Holdings Indigo -14.44%

-30% -> -20% NL NL Industries Inc. Green -26.22%

-40% -> -30% MGI

Moneygram

International Inc Violet -32.32%

Capital Stok Name Capital Stock

Symbol Color Assigned

Stock Market

Traded Industry

The Boeing BA Black NYSE Aerospace Defense

NL Industries Inc. NL Green NYSE Industrial Distribution

Delek US Holdings Dk Indigo NYSE Energy and Infrastructure

CitiGroup C Orange NYSE Financial Services

McDonald's Corp. MCD Pink NYSE Restaurants

The Walt Disney

Company DIS Red NYSE Mass Media

Moneygram

International Inc MGI Violet NASDAQ Financial Industries

International Paper

Company IP White NYSE Pulp and paper

Johnson & Johnson JNJ Yellow NYSE

Medical Equipment and Pharmaceutical

(12)

12

After the stocks were chosen, 11 indicators were calculated for each one. The indicators were computed using data from two brokerage firms and the companies’ websites. From the two brokerage firms various prices and the dynamics of the price were collected and from the companies’ websites the information found in the balance sheets, income statements and cash flows was gathered. To be able to form an investment strategy, participants needed information on the evolution of the companies for at least the previous 4 years. Therefore, the indicators were calculated for a historical period starting from May, 2009 until May 2013.

Next, it will be shown how the indicators were calculated, their interpretation, what data was used and where it was collected from.

Yearly prices and dividends paid.

The participants were provided with various prices and dividends paid in the specific historical period. The prices provided contained opening and closing prices, the variation of the price in that day (only the highest and the lowest quotes), the volume of stocks traded and also the adjusted closing price. The prices were given for specific dates, four historical years , namely 1st_{of June,}

2010; 31st_{of May, 2011; 30}th_{of May, 2012 and 30}th_{of May, 2013;. These indicators were given in}

order to form an idea about the evolution of the prices and the market sentiment over time for a given security.

Along with the yearly prices participants received information on the total amount of dividends in one year until the specified date. This indicator could be used to assess if the company is secure and/or stable. Companies usually offer dividends to their stockholders to make up for a small move in the price. High-growth companies rarely offer dividends because they reinvest all of their profits to help sustain a higher than average growth rate.

Both yearly prices and the dividends paid on the respective dates were expressed as numbers and were collected from marketwatch.com and rechecked using Google finance.

The dynamics of the price.

From bloomberg.com graphs with the evolution of the daily prices for the historical period of the 9 chosen stocks were collected. These were used mainly to give an idea about the volatility of the stock price over time. If a participant would invest in a highly fluctuating stock, it would mean that he/she would chose a risky investment strategy which could in return provide a high profit or loss. While if the price was constant over time, investing in that particular stock would imply a more safe investment, but also a moderate profit or loss.

Dividend Yield Ratio.

From the companies’ websites quarterly data concerning the dividends paid were collected. By combining it with the adjusted closing prices the dividend yield ratio was calculated using the formula:

.

(13)

13

This indicator was expressed as a percentage for each year of the historical period, starting from 2010 until 2013. In this way, the participant could capture the evolution of the indicator over time. This proxy shows the distribution of the company’s earnings to a class of its shareholders. This distribution is decided by the board of directors and is not guaranteed. A company can decide to eliminate or reduce its dividend in times of financial hardship. We can use the dividend yield as a tool to measure how much cash flow an investor is receiving for each dollar invested in a dividend-paying stock.

Book Value per Common Share.

With the data obtained from the quarterly balance sheets that were available on the companies’ websites the book value per common share indicator was determined using the formula:

.

This proxy was expressed as a number for each year of the historical period that started from 2010 until 2013.

This indicator can be used as a tool for showing the level of safety associated with each individual share after all debts are paid. In other words, it determines the dollar value remaining for common shareholders after all assets are liquidated and all debtors are paid.

Price to Book Value Ratio (P/B ratio).

Using the adjusted price of the capital stock together with data taken from the quarterly balance sheets of the companies, the Price to Book Ratio was determined using the formula:

.

The historical period chosen is from May, 2009 until May, 2013. The indicator was expressed as a number for each year of the historical period. This is a very important indicator because it shows whether a stock is underpriced or overpriced. It also compares the cost of a stock to the value of the company if it was broken up and sold today.

Earnings per Share (EPS).

By combining the data from the quarterly income statements and balance sheets of each company, the earnings per share indicator was calculated using the formula:

.

This indicator is expressed as a number only for the first year of the historical period, namely from May, 2009 until May, 2010. Afterwards it is described as a percentage change with the reference point being changed to each previous year until May, 2013. I chose this way of describing this proxy to help participants to see the evolution over time.

(14)

14

The EPS indicates the company’s profitability and is used to present the portion of the company’s profit allocated to each outstanding share of common stock.

Price to Earnings per Share Ratio (P/E ratio).

Using the adjusted closing prices together with the Earnings per Share previously described, the price to earnings per share indicator was computed using the formula:

.

Only in the first year of the historical period the indicator was expressed as a number namely, the P/E of the period from May, 2009 until May, 2010. The rest of the historical years were presented as a percentage change with the previous year taken as a reference point. I chose this way of representing to show the evolution over time.

The P/E ratio is also referred to as the “multiple” because it shows how much investors are willing to pay per dollar of earnings. In other words, it shows if the company is overvalued or undervalued. Generally, a high P/E ratio means that investors are anticipating higher growth in the future, while a negative P/E ratio shows that the company is actually losing money.

Earnings growth.

By accessing each company’s website and looking at the quarterly income statements it was possible to calculate the earnings growth indicator. I took from the statements the after-tax net income of the companies for the historical period needed. This indicator was presented to the participants both as a number and as a percentage change. Only the first year of the historical period was expressed as a number, the rest of the years being described as a percentage change with the reference point being the previous year.

The earnings of a company can indicate whether the business will be successful and profitable in the long run and constitutes the main determinants of its share price.

Earnings Before Interest, Taxes, Depreciation and Amortization (EBITDA).

Using the quarterly income statements of each company, I computed the EBITDA indicator with the help of the formula:

EBITDA is a financial performance indicator that can be used to analyze and compare the profitability between companies and industries as a result of eliminating the effects of financing and accounting decisions. This indicator is a good proxy to evaluate profitability, but not cash flow. It is usually used together with other performance indicators because EBITDA can be utilized as an accounting trick to try and hide something, for example cash required to fund the replacement of old equipment which can be large in some situations.

(15)

15

The indicator was presented to the participants both as a number and as a percentage change. Only for the first year of the historical period, EBITDA was expressed as a number. For the consequent years following 2011 until 2013, it was described as a percentage change with the previous year taken as a reference point.

Market Capitalization.

The Market Capitalization indicator represents the total market value of a company’s outstanding shares. Investors use this proxy to determine the company’s size which in turn helps them to estimate the risk of their investment. Generally speaking, a company that has a large market capitalization is considered to have a potentially low risk and low return. While a company with a small market capitalization is considered to have, on average, a high return, but also high risk. This indicator was described to subjects as a number, only for the first year namely, May, 2009 until May, 2010. For the next consequent years until the last year of the historical period, it was expressed as a percentage change with the previous year being the reference point. I chose to represent it this way to help participants see more easily the evolution of the market capitalization over time.

I calculated the indicator using the adjusted closing price collected from Bloomberg and the total common shares outstanding from each company’s quarterly balance sheets. The formula used was the following:

. Leverage Ratio.

A leverage ratio is any of several financial measurements that look at how much capital comes in the form of loans. It can be used to assess a company’s ability to meet financial obligations. In my study, I chose one of the most popular leverage ratios, the debt to equity ratio. The formula used to calculate is straightforward:

Generally speaking, if all else is equal between two companies, the one with the lower leverage ratio is considered to be a safer investment. In addition, a business that has a leverage ratio above 2 is viewed as a risky investment.

The data to calculate the indicator was collected from the quarterly balance sheets available on each of the chosen companies. The debt-to-equity leverage ratio was presented for the entire historical period only as a number. This way of describing was preferred so that participants will have a better understanding of the evolution of the company across time.

After the indicators were calculated I needed to acquire data that would reflect the outcome of the investment decision made by participants. To do this I collected the dynamics of the price for a period of 1 year. The starting date was from the last day of the historical period namely, May, 2013

(16)

16

until May, 2014. I chose this way of presenting the results instead of just the final price so that participants would have a much clearer image of how the price fluctuated during that year for the 3 chosen capital stocks. The graph did not show only the movement of the price, but also the percentage change in that year so it made it easier to calculate the payoff.

The participants have been provided with all the materials necessary to complete the task, a pen, color markers, a scientific calculator and extra blank sheets of paper. They were given all the time needed to complete their tasks. They were allowed to make further calculations and any marks or signs on any of the sheets of paper provided to them at the beginning of the experiment. These papers were afterwards collected for further analysis.

2.4 Experimental design

A between subjects experimental design was preferred instead of a within subjects design to avoid carry-over effects. If I would have exposed the participants to both condition, the influence of the first treatment would have most definitely carry over when the second treatment would have been applied. Therefore, I would not have obtained reliable results. The experimental design opted for implies dividing the participants in two distinctive groups: a control group and a treatment group. Each group was comprised of 15 randomly chosen students. After finishing the experiment, I compared the results obtained by the control group with the ones of the treatment group. In this way, I managed to observe the effects of thinking too much on participants’ decisions.

Decision difficulty was manipulated through the number of indicators available to them. The aim with the first sample, the control group, was to observe the decision-making process of the participants when a complete level of information was provided to them. However, the participants in the second sample group, the treatment group, received only a limited amount of information to make their decision.

The participants had to construct a portfolio of 3 distinctive stocks (see Appendix 1 and Appendix 2 for instructions). They had to choose among the 9 stocks selected from the New York or NASDAQ stock market. Each participant received a list with the interpretation of the 11 indicators (see Appendix 3 and Appendix 4 for indicators list). The amount of information received on those 11 indicators differed, depending in which group the participant was assigned to. Subjects in the control group had access to all the data on the 11 calculated indicators for the historical period chosen while participants in the treatment group had access to only 3 out of the 11 indicators (see Appendix 5 for data on the calculated indicators). After receiving the data, the participants had to form an investing strategy to construct their portfolio.

The experiment was not conducted in a one-day session. I considered that I could obtain more reliable data by performing the experiment in individual sessions. This way of performing the experiment had two major advantages. First, I could observe more closely the behavior of each participant while performing the task. These observations proved to be extremely important for the purpose of my experiment. And second, I could ask additional questions about the task and about the behavioral observations made during the experiment.

(17)

17

These behavioral observations together with the additional question asked provided more data to be taken into account that gave insight in the decision making process of each participant.

The experiment investigates whether thinking too much can have a detrimental or a positive effect on the decision making process when investing in the stock market. In other words, thinking too much leads participants to a positive or a negative outcome. Participants were randomly assigned to either one of the two groups by blindly selecting a slip of paper from a small bag. The slips of paper contained either a line or were blank. The ones with a line on them assigned the participant to the treatment group, while the blank ones to the control group. Both groups had to perform a task after which they had to fill in a questionnaire. The task was the same for both groups with slight modifications.

The order of the list of explained indicators together with the order in which the 9 stocks were presented was randomly chosen. I did this to avoid any order effect that might arise. Some participants might have correlated the position of the indicator in the list with its importance or the position of the stock with its performance. Correlation that would result in biased results. To avoid that, I mentioned in the instruction the fact that both the indicators list and the stocks were randomly ordered.

After the task was completed, all participants received a questionnaire to fill in. The questionnaire had either 34 or 36 questions, depending in which condition the subject was randomly assign to (see Appendix 6 and Appendix 7 for the questionnaires). The questions were related to the task, to the expected outcome and their investment strategy compared with others in the same condition, to counterfactual thinking and decision making process. The questionnaire also assessed the level of satisfaction or regret experienced by the participant. The subjects had to fill out the questionnaire with one of the 3 appropriate types of answer. The participant had to circle the answer, give an open answer or both.

If the question required a “Circle the answer” type of response, the participants had to either select between 2 options, “YES” and “NO”, or use a scale. The scale was from 1 to 5 and had different interpretation depending on the question. For example 1 could represent the lowest limit in one question, while in other could represent the highest limit. I choose a scale to represent the answers of the participants, so I could have a more accurate response that would help me in analyzing the data.

Other questions required an “Open Answer”. Usually this type of response was used to describe the decision making process of the participant, to evaluate predictions and to assess counterfactual thinking. The participants had to write in detail their investment strategy or just to write a number, depending on the question.

The last and final type of answer combined both types of responses mentioned above. Participants had to both “Circle the Answer”, using the scale or choosing between “YES” and “NO”, and also to give and “Open Answer”. The appropriate type of answer was specified after each question to ensure the validity of the response.

(18)

18

While each participant was solving the assignment their behavioral reactions had been recorded by the experimenter. The participants were not made aware of this observational stage of the experiment, in order to avoid the Hawthorn effect, seeing as the information might have had an impact on the participants’ performances during the trials. As a result, the participants were seated in a chair approximately 2 m from the experimenter, while the experimenter recorded their behavioral reactions during the trial in a hidden from their view notebook.

At the end of each experiment, a discussion with each participant took place. During this time, additional questions were asked related to the task and the decision making process. Afterwards, the results and the payoff were provided to each participant, individually (see Appendix 8 for data that was used to reflect the outcome of participants).

The participants’ payoff was dependent on the decisions made in the task. The amount of money received was calculated using the investment return of the 3-stock portfolio over a period of one year. Because some of the stocks suffered a decrease in their prices in the interval May 2013 - May 2014 and also because there was a possibility that the overall return of the portfolio could have been negative, the initial investment of each participant was considered to start from 8 Euros, regardless of the prices of the stocks chosen. This way, participants could have obtained a negative return on their portfolio.

The actual prices of the stocks were used only to calculate the percentage change of their initial investment. Using the formula:

, m, n and p represent the percentage change in price of the 3 stocks chosen,

All the materials needed to conduct the experiment were printed on just one side; this way of presenting the information ensured better observational recordings and helped the participants in making their decision. I mainly wanted to see if the way people hold the sheets of paper with the information about the calculated indicators might have an impact in their decision. They could have hold them either altogether or scattered across the table. In my point of view, the dispersed way of looking at the calculated indicators for the historical period would make the comparison between stocks more easy because they would have multiple indicators visually displayed when comparing. When designing the experiment, both external and internal validity interferences with the experiment were taken into account.

For avoidance of external validity issues, measurement biases were taken into account. Minimization of the systematic errors was considered by reducing various factors that might influence it. As a result, changes in environment, the calibration of the measurement instruments, locations, time effects and, the Hawthorne effect are some of the aspects that the experiment was concerned with.

First, the environment in which the experiment was conducted did not change across time. Participants were approached randomly in the university library and asked if they would like to take part in an experiment at the end of which they would be financially rewarded. After accepting,

(19)

19

I took the participant into one of the university’s individual study room, to avoid external factors that might interfere in the process, including the use of mobile phones.

Second, before and after each participant, careful attention was paid to the stop-watch used for the time variable factor related to the experiment. The timer started after the instructions were handed in and stopped after they wrote the final selection of the stocks in which they wanted to invest in. By doing this I was able to perfectly calibrate the measurement instrument.

Third, the location in which the experiment took place varied. Although it might have been regarded as creating problems with external validity issues, it was not the case. Time constrains factors for the amount of time possible to use the study rooms, meant switching to other rooms. However, all the rooms used were identical in all aspects, thus not tainting the experiment.

Forth, opting for individual experimental sessions, instead of a one-day experimental session meant that I had to take into account time effects. These effects might have arisen because the experiment took approximately one week to complete. It has been a priori considered that varying my behavior might influence the participants and obtain time effects biases in the participants’ responses. That is why, during this week, each potential participant was approached with a neutral tone and a minimum level of enthusiasm. After they accepted, the participant was showed to the room where the experiment would take place. During the experiment, a professional and a reserved attitude were maintained towards the subject. The participants had been informed that they could ask any questions, during the experiment and an answer will be provided to them as long as the question does not require a leading answer.

Lastly, the Hawthorne effect was taken into account in trying to minimize the systematic error. The participant’s behavior could change due to the attention received from the researcher during the experiment rather than because of any manipulation of independent variables. As a result, the experimenter was sitting at a fair distance from the subject and the observation notebook was hidden in a book that he was pretending to read.

In addition, aspects concerning representativeness of the sample were considered. To achieve this task, sampling errors and biases have been aimed at reducing. Sampling error refers to the deviation of the selected sample from the true characteristics, qualities or behaviors of the entire population. The only way in which they could be eliminated altogether is by testing the entire population that holds a certain degree of knowledge in the Business and Economics fields. Since this is impossible to perform, uses of the unbiased probability samplings and by using a large sample size were engaged.

The proper and unbiased sampling method was ensured by approaching students randomly. With the use of randomization a representative sample was established. However, random sampling is just one method to diminish the sampling error. The other method is by increasing the sample size. Originally, the experiment was design as having 10 participants in each group, instead of the one actually done, of 15 participants. The reason each group was increased by 50% was not only to reduce the sampling error, but also to provide a more relevant perspective of the experiment’s findings.

(20)

20

For all the above reasons, without minimizing the systematic and sampling error the whole experiment could have posed serious external validity issues.

Moreover, the experiment had to also address concerns about the internal validity aspects that might have jeopardized the experiment. To avoid multiple internal validity issues testing effects, instrumentation, selection-maturation interaction, diffusion and, experimental bias had to be considered.

First, testing effects usually arise when subjects are tested multiple times and learn the results or they predict what they are tested for. Therefore, the results of the participant would increase as a direct result of repeated testing of the same task. In this case, the candidates that just took part in the experiment could have talked to others after completion of the task. As a result, volunteers for the experiment without the prior asking to participate on the part of the experimenter were turned down. Second, instrumentation is another factor that could have created problems. To ensure that all the conditions are the same for all subjects, identical instrument materials were used. For example, I used the same book to hide the observation notebook which had the covers wrapped in white paper; I dressed in the same way, used the same type of pens and provided the participants with the same color markers to make any signs or marks on the given sheets of paper.

Third, in timing the candidates the same stop-watch for all 30 participants was used. The timer was started after participants received the instructions and began reading. It was stopped after the participants wrote their final 3 stocks chosen and said that they are done with the task. The timer was stopped twice in the treatment condition; more specifically, after the participants selected the 3 indicators for which they would chose to receive data to help them in their investment strategy and after the task was completed.

Forth, the selection-maturation interaction is another factor that could have threatened the internal validity of the study. By choosing to approach students in the library, it was ensured that the participants had approximately the same age. This would ensure that the participants were more or less equal in the degree of knowledge and insight level. Fifth, a factor that could have posed a serious danger to the study was diffusion. Sometimes the treatment and control group participants are able to communicate and exchange information with each other; for example, methods, materials and points of view. By conducting individual experiment sessions this cross-contamination between the groups was avoided.

However, other factors might have led to imitation of treatments such as, great homogeneity between the participants that may result in them interacting outside of the research process or sharing similar social groups. For this reason, it was not accepted more than one participant from a group of students approached in searching for potential candidates. The last and final factor that was taken into account, in order to avoid internal validity problems was experimental bias. This should not be confused with the Hawthorne effect. Experimental bias refers to the individual who conducts an experiment and inadvertently affects the outcome by non-consciously behaving in different ways to members of treatment and control groups. During all 30 individual experimental sessions, a professional attitude towards the participants was maintained, a neutral tone and a limited conversation to the task performed. Even so, some of the participants have tried to find out

(21)

21

some extra information that was not related to the experiment or asked questions that required a leading answer. Such questions remained unanswered until the end of the experiment.

Part 3. Results

As already mentioned, the participants had to choose 3 out of 9 stocks available. They had to construct a 3-stock portfolio using different amounts of information depending in which group they were in. Participants in the control group had access to all 11 indicators, while those in the treatment group had access to only 3. The percentage change of the subjects’ portfolio represented their outcome.

3.1 Average return

A priori, it was expected that the average percentage change of the 3-stock portfolio of participants in the ”thinking too much” condition (also known as control group) to be smaller than the average return of participants in the “not thinking too much” condition (also known as the treatment group) (H1).

After analyzing the results, the prediction was not confirmed. Even though the average return on investment for participants in the control condition was just 3.14%, while the average return for subjects in the treatment condition was 5.33% the result was not confirmed after conducting a Mann-Whitney U-test (p-value is 0.77182 which is bigger than 0.05). Table 3 also describes why the result was not statistically significant.

(22)

22

Table 3. The ranking table of outcomes obtained by participants in both groups while performing the Mann-Whitney U-test.

The smallest U, between U1 and U2 (in this case U1) is not smaller than the critical value (CV) found in the Mann-Whitney critical value table. Therefore, the result is not statistically significant which means that there is no difference between the mean ranks of outcome based on the amount of thinking. Thus, it cannot be concluded that thinking too much has detrimental effects on one’s decision making. The decisions taken by thinking too much were at least as good as those taken by thinking less, leading to profit, even if they were not the best (average return obtained by the control group was 3.14%, which is good, bit less than the average return of 5.33% obtained by the treatment group)

3.1.1 Stocks chosen

The number of times each stock was chosen to include in their final portfolio that contributed to the difference between the average outcomes obtained by each group of participants is illustrated in Figure 1, below.

Thinking too much

condition (n=15) Rank

Not thinking too much condition (n=15) Rank -22.53 2 -24.32 1 -18.04 4.5 -18.6 3 -18.04 4.5 -8.03 7 -10.29 6 -3.56 8 2.1 10 -1.54 9 2.38 11 4.17 12.5 5.19 14 4.17 12.5 7.2 16.5 6.64 15 7.2 16.5 10.9 18 11.47 19.5 11.47 19.5 11.74 21 17.18 24.5 15.73 22 17.18 24.5 17.18 24.5 18.43 27 17.18 24.5 18.66 28.5 18.66 28.5 27.19 30 R1 225 R2 240 U1 105 U2 120 CV 64

(23)

23

Figure 1. The graph describes the number of times each stock was chosen by the participants.

Looking on the left side of the graph, at the stocks that had negative changes in price, namely, Violet, Green, Indigo and Orange, it can be observed that they were chosen in approximately the same proportion by each group. 35.56% of participants in the control group chose the negative performing stocks, which is exactly the same percentage as in the case of the participants in the treatment group; 13.34% of candidates in the “thinking too much” condition picked the White stock, which is the neutral performing stock (it had a percentage price change of 0.04%), while the participants in the “not thinking too much” condition chose this particular stock only 2.22% of the times. Looking on the other side of the graph, at the stocks with positive changes in price, namely, Pink, Yellow, Red and Black, it can be observed that the participants in the control condition chose only 51.1% of the times, while the subjects in the treatment condition picked the specified stocks about 62.22% of the times.

Moreover, the highest and lowest return obtained by participants in either one of the two groups is worth mentioning. On one hand, looking only at the subjects in the control group, the highest average return acquired was 18.66%, while the lowest return was at a -22.53% level. On the other hand, turning now to the treatment group, 27.19% was the highest return on investment, while the lowest return was at a -24.32% level.

3.1.2 Indicators used in making the stock picking

The 3-stock portfolio was constructed using either all the 11 calculated indicators or using only 3, depending in which condition the participant was in. Figure 2 illustrates which indicators were

(24)

24

preferred by the participants in the treatment group which had to choose from the multitude of indicators made available.

Figure 2. Indicators chosen by participants in the treatment group.

Apparently the main indicator chosen was earnings growth and the dynamics of the graph. Each being selected 20% of the times. The second most popular indicator chosen was earnings per share, selected 13.5% of the times, while the rest of the indicators were picked less than 9% of the times. These participants chose the specific indicators more because of the popularity and the apparently simple and straightforward interpretation that they offered.

3.2 Satisfaction and regret level

The second main hypothesis (H2) regarding the level of satisfaction and regret experienced by the participants in the two groups was also not confirmed. It was predicted that subjects in the “thinking too much” condition will feel less satisfied and experience more regret than subjects in the treatment condition.

In the questionnaire, participants had to answer a question related to their level of satisfaction or regret experienced after they finished the task. The question required the subjects to provide an answer using a 5 point Likert-type scale. The lowest extremity of the scale (1) meant that the participant did not experience any regret at all, thus he/she was completely satisfied; while the highest extremity (5) would represent the candidate who experienced no satisfaction at all. The results obtained are depicted in the Figure 3, below.

0 1 2 3 4 5 6 7 8 9 10 N u m b e r o f tim e s ch o sen Indicators

(25)

25

Figure 3. The average level of satisfaction and regret experienced by participants.

Candidates in the treatment group, with access to only a limited amount of information, had been more satisfied and experienced less regret with their investment decision than participants who had access to all information of all the 11 indicators.

As it is clearly shown in the graph above, there is not a big difference between the two groups. The average level of satisfaction experienced by the participants in the treatment group was higher than the one experienced by the subjects in the control group. The participants in the “thinking too much” condition reported on average a satisfaction level of only 2.33, while those in the “not thinking so much” condition reached a level of 2.53, on the satisfaction scale.

When considering the average level of regret, we observe that participants in the control group reported a higher level than subjects in the treatment group. While participants in the “thinking too much” condition expressed, on average, a level of regret of 2.67, subjects in the “not thinking too much” condition stated a level of only 2.47.

The differences between the level of regret and the satisfaction experienced by the two groups are rather small to be able to conclude that the hypothesis was confirmed. Both groups were in fact fairly satisfied, experienced moderate regret and had a positive return rate. The decisions taken by both groups were good, even if putting more effort and thinking too much did not automatically lead to the best decision, but only to a good one. From their perspective, an acceptable decision, a profitable and at least as good as the one taken by the other group which did not think their decision too much. Consequently, it cannot be concluded that thinking too much is detrimental considering the satisfaction and regret level experienced, as the results are close to those obtained by not thinking too much.

(26)

26

3.3 Difficulty of the task

Depending on the amount of thinking employed in making the investment decision imposed by the condition in which the participant was in, he/she might have perceived the difficulty of the task differently. I predicted that the participants in the “thinking too much” condition will experience the task as being harder than the participants in the “not thinking too much” condition (H3). However, the data did not confirm this prediction either. Indeed, the participants in the control group perceived the task as being slightly more difficult, the difference was too small to be accounted for. Therefore, overall, the participants felt that the task was approximately the same when they considered its level of difficulty.

A difficulty Likert scale that ranged from 1 to 5 was used to assess the level of difficulty experienced by the participants. The starting point of the scale was associated with an easy task, while the ending point would represent a difficult task. Figure 4 summarizes the results obtained.

Figure 4. The difficulty of the task perceived by participants in the control and treatment group.

Looking at the most common response provided by the participants in each group the prediction should have been confirmed. The participants in the control group felt like the task was difficult, most of them marking the point 5 on the difficulty scale, while most of the participants in the treatment group assessed the task as being only slightly difficult, most of them answering with 4, on the difficulty scale.

Thus, both groups found the task slightly difficult, but participants in the treatment group considered it less difficult than the subjects in the control group. But when considering the average response of both treatment and control groups it can be observed only a slight difference. In the control group, the average response indicates a 4.13 level and the range of responses are distributed only in the upper part of the scale. While in the treatment group, it can be seen that the answers are distributed across the whole scale and have a slightly smaller average than the one of the control group, reaching a level 4 on the difficulty scale. Therefore, the participants in the

(27)

27

treatment group regarded the task as being slightly easier than participants in the control group, but the difference was too small to be considered relevant.

3.3.1 Decision time

It was predicted that the time spent by the participants in the “thinking too much” condition will be relatively close to the total time of subjects in the “not thinking too much” condition. Total time, refers to the time spent in choosing the indicators, plus the time spent in completing the task after the information requested was provided. The results are depicted in Figure 5 below.

Figure 5. The total time spent in solving the task by participants in the control and treatment group.

As it can be seen, aggregate subject responses were consistent with the forecast assumed. Looking at the shortest time in both groups, the graph shows that the fastest participant was the one in the treatment group with a time of 7 minutes and 12 seconds. The shortest time in the control group was only 8 seconds more, thus the timer indicated 7 minutes and 20 seconds.

The only big difference between the control and the treatment group is by looking at the longest time. In the “thinking too much” condition, the longest time spent by a participant in solving the task was 42 minutes and 51 seconds, while in the “not thinking too much” condition the longest time was just 31 minutes and 5 seconds.

Looking at the average time spent by participants in each group we can notice just a small difference. The average time spent by participants in the treatment group was 18 minutes and 50 seconds to completing the task, while subjects in the control group followed their time closely. Their average time was 20 minutes and 50 seconds, with just 2 minutes behind the other group. Overall, it can be concluded that the average time required by the participants in the control group to complete the task was almost the same as the average total time spent by subjects in the treatment group.

(28)

28

3.4 More information or less

The 4th_{and last main hypothesis assumed that the subjects in the control group would be}

overwhelmed by the excessive amount of information and would rather have access to a more limited amount. While subjects in the treatment group, because they were asked to choose only a limited number of indicators, would have liked more information (H4).

To find out whether the participants considered that they had either too much information or too little, a base line had to be established. Therefore, the participants were asked if they prefer more or less information, in a general investment situation. To answer this question, participants had to circle the answer using a scale that was gradually rising from 1 (less information) to 5 (more information). Figure 6 sums up the participants’ responses.

Figure 6. The graph describes the level of information needed by participants to make a decision in a general investment situation.

It can be observed that 40% of the total number of participants usually prefers slightly more information while making an investment decision. However, the average amount of information needed in the control group reached a level 3 on the scale. This can be interpreted as requiring a balanced amount of information. While the average amount of information needed in the treatment group attained a level 4 on the scale can be interpreted as requiring slightly more information in a general situation. After obtaining a base line there was a question that reflected the opinion of participants regarding having too much or too little information about the stocks to be able to make a good decision. The participants had to use a Likert-type scale from 1 to 5, here 1 would represent the answer “no” and 5 would represent the answer “yes”. The graph below illustrates the results and also describes the intermediate points in the scale.

0 1 2 3 4 5 6 7

Control Group Treatment Group

N u m b e r o f r e sp o n ses Scale

More or less information, in general

1 (less) 2 (slightly less)

3 (neither less nor more) 4 (slightly more)

5 (more) Average

(29)

29

Figure 7. Participants’ opinion about having too much information to make a good investment decision.

Looking at the most common response provided by the participants in each group, the tendency would be to assume that H4 has been confirmed. Indeed, in the control group, the most frequent answer provided by the participants was “4 (slightly yes)”. It could be interpreted that most of the participants considered that they had too much information in order to make a good decision. This would imply that they would have liked to be subjected to less data so that they would not be confused. However, 7 out of 15 subjects in the treatment group answered the same question with “1 (no)”, meaning that they obviously considered that they did not have too much information to make a good decision. But turning now at the average response within each group, we could see why H4 is not confirmed. In the control group, the average response is 2.8, which is really close to the neutral point of the scale. This means that the participants considered they had just the right amount of information, neither too much nor too little. In the treatment group, the average answer was 2, meaning that the subjects in this group thought that they could have used slightly more information, but the amount of information provided was enough to be able to make a good decision.

Part 4. Discussion

The aim of the study was to observe if thinking too much can have a detrimental or favorable effect on a person’s investment strategy. The experiment was designed in a way that the participants

0 1 2 3 4 5 6 7 8

Control Group Treatment Group

N u m b e r o f p ar tici p an ts

Answers using Scale

Too much information to make a good investment decision?

1 (no) 2 (slightly no) 3 (neither no nor yes) 4 (slightly yes) 5 (yes)

More is worse or more is better? : the impact of thinking too much on investment decision making in a stock market setting