Unethical behavior regarding groups or individuals : Is hurting an individual perceived as more severe than hurting a group, when the monetary costs are equal?

Unethical behavior regarding groups or individuals

Is hurting an individual perceived as more severe than hurting a

group, when the monetary costs are equal?

Erik Groenenberg

Master thesis MSc Business Economics Specialization Organization Economics 15 ECTS

First university supervisor:

Dr. G. Doğan Student number: 6178944

Second university supervisor Tel.: +31 (0)6 43 99 60 92

Thomas Buser

Abstract

Cases of fraud, cheating and other instances of unethical conduct are often in the media. An explanation according to some is that unethical behavior hurting large organizations or rich counterparts is perceived as less severe as hurting an individual while the same monetary costs are equal. In this research the effect of spreading the monetary consequences of unethical behavior over a group instead of one individual is researched through a behavioral experiment. The experiment consists of first an earnings stage and second a dictator game where earned wealth is redistributed. It is found that participants show unethical, so hurting, behavior in the group treatment as well as in the treatment concerning individuals. There is found to be no statistical significant difference between both treatments. There is also looked at gender differences in the experiment. A significant difference between behavior of men and women in the experiment, concerning the treatment with the monetary costs claimed from individuals, is found. Concerning the group treatment this difference between men and women is not significant.

Key words: Unethical behavior, Dishonest behavior, dictator game, differences in group and personal hurting

Acknowledgements

I would like to thank my thesis supervisor, Gönül Doğan, for thinking along about how to formulate my research question and for brainstorming with me about how to test this in an experiment. Without her support I would not have been able to truly test my research question in an experimental setting. Furthermore I thank Joep Sonnemans, who also had a large share in enabling me to actually execute the experiment, for helping me programming and performing the experiment. Lastly I would like to thank my family and friends for their ongoing support and trust in my during my whole student career.

Table of contents

1. Introduction ... 1

2. Literature review ... 3

2.1 Homo economicus ... 3

2.2 Deviations from homo economicus ... 3

2.3 Unethical behavior ... 5

2.4 Ethical dissonance gap and coping mechanisms ... 5

2.5 Group effects ... 7

2.6 Gender effects ... 7

2.7 Earned position and wealth effects ... 9

3. Methodology ... 12 3.1 Hypotheses ... 12 3.2 Experimental design ... 13 3.3 Sample ... 15 4. Results ... 16 4.1 Descriptive statistics ... 16 4.2 Unethical behavior ... 17

4.3 Differences in unethical behavior between treatments ... 18

4.4 Gender differences in unethical behavior ... 19

5. Discussion ... 21

6. Conclusion ... 24

References ... 25

Appendix ... 28

A. Matrices ... 28

B. Experiment instructions pair treatment ... 32

C. Experiment instructions group treatment ... 35

D. Shapiro-Wilk W test for normality of behavior ... 38

E. Sign test behavior ... 38

F. Shapiro-Wilk W test for normality of behavior group treatment ... 38

G. Shapiro-Wilk W test for normality of behavior pair treatment ... 38

H. Mann-Whitney U test unethical behavior, by treatment ... 38

I. Shapiro-Wilk W test for normality of behavior women... 39

J. Shapiro-Wilk W test for normality of behavior men ... 39

K. Mann-Whitney U test unethical behavior pair treatment, by gender ... 39

L. Mann-Whitney U test unethical behavior group treatment, by gender ... 39

M. Shapiro-Wilk W test for normality of behavior men in the group treatment ... 40

N. Mann-Whitney U test unethical behavior men, by treatment ... 40

O. Shapiro-Wilk W test for normality of behavior women in the group treatment ... 40

P. Shapiro-Wilk W test for normality of behavior women in the pair treatment ... 40

Q. Mann-Whitney U test unethical behavior women, by treatment ... 41

R. Mann-Whitney U test unethical behavior, by gender ... 41

S. T-test unethical behavior, by treatment ... 41

T. T-test unethical behavior pair treatment, by gender ... 42

U. T-test unethical behavior group treatment, by gender ... 42

V T-test unethical behavior men, by treatment ... 42

1. Introduction

Cases of fraud, cheating and other instances of unethical conduct by firms or individuals are often highlighted in the media. Examples of this are the fraud of Enron, Bernard Madoff, Barclay’s in the United States and recently the accountancy firm KPMG in the Netherlands. These are all large organizations and one of the possible reasons people might engage in these kinds of unethical conduct is explained by people being more accepting of fraudulent behavior hurting a large organizations or a rich counterpart than fraudulent behavior hurting only an individual. Despite of the monetary costs being identical, hurting an individual with the same amount is perceived as greater (Gneezy, 2005). The question how to stop these kinds of frauds, cheating and unethical conduct might raise. Before fraud can be brought to an end one first needs to better understand under which circumstance these kinds of behaviors are more likely to occur. To see if the former explanation of Gneezy (2005) holds, the following research question is formulated:

Is there a difference in unethical behavior when the negative cost of this unethical behavior is born by a group or by one individual?

In an attempt to answer the research question first some relevant literature is discussed. Second the research methodology is given. In that chapter hypotheses are formulated used for testing if there is a difference in unethical behavior . After that respectively the experimental design and sample are discussed. Third the results are discussed on the basis of the hypotheses formulated in the methodology after which the discussion follows and last the conclusion is given.

In the literature review at first a closer look is taken upon economic theory and the assumptions that say one could expect this kind of fraudulent behavior. Concerning the assumption that people act like homo economicus, unethical behavior will occur when this is beneficial for the individual’s wellbeing. Examples of theories based on the assumptions of homo economicus are, within game theory for example, Nash Equilibrium (1950); the lemon principle of George Akerlof (1970); or the well know contract theory from Kenneth Arrow (1962), all based upon the assumption of individuals acting like a homo economicus. Based upon homo economic thoughts, one should not expect a difference in unethical behavior. Also logic defines that if people are rational, they treat the same monetary outcomes the same and therefore do not show differences in unethical behavior. Also some social psychology literature and research is discussed. The latter literature suggests that individuals try to

maintain a private and a public positive self-image (Rosenberg, 1979) and that people want to feel good about themselves which likely includes seeing themselves as moral (Schlenker, 1982). This might suggest that there will be no unethical behavior at all. Subsequently literature that finds coping mechanisms to maintain a moral self-image while showing unethical behavior is discussed. One of these coping mechanisms concerns group effects and based on this literature one might suspect to see a difference in unethical behavior when the negative cost of this unethical behavior is born by a group or by one individual. Finally literature is discussed that might be relevant for the experimental setup, like gender effects, the effects of earned position and wealth effects.

2. Literature review

In this chapter the existing literature about the current knowledge of homo economicus is discussed. There is started with some known deviations and current research on unethical behavior, after which the gap between a moral self-image and unethical behavior is enlightened together with some coping mechanisms that let people exhibit unethical behavior without feeling to bad about themselves. One of these coping mechanisms concerns group effects and is elaborated on before there is looked upon possible gender effects and the effects of earned position and wealth effects.

2.1 Homo economicus

A lot of economic theories are built on the assumption of homo economicus. Gneezy (2005) describes a homo economicus as a figure who acts selfishly and is unconcerned about the wellbeing of others. The assumption is made that each individual only cares about his or her own wellbeing. Unethical behavior will occur when this is beneficial for the individual’s wellbeing. Examples of theories based on the assumptions of homo economicus are, within game theory for example, Nash equilibrium (1950); also the lemon principle of George Akerlof (1970); or lastly the well know contract theory from Kenneth Arrow (1962), where contracts written in a world where information asymmetry is present cause principle agent problems, based upon the assumption of an agent acting like a homo economicus.

Homo economicus is first described by John Stuart Mill in 1874. Without giving it a name, he described it as an arbitrary definition of man, as a being who inevitably does that by which he may obtain the greatest amount of necessaries, conveniences, and luxuries, with the smallest quantity of labor and physical self-denial with which they can be obtained. The first appearance of the term ‘economic man’ was in the work of John Kells Ingram (1888), disparaging Mill’s theoretical abstraction (Persky, 1995). According to Persky (1995) the first use of the Latin term, homo economicus, was in Pareto's manual (1906).

2.2 Deviations from homo economicus

Kenneth Arrow, the economist who used the assumptions of homo economicus to develop the contract theory is also one of the first to question homo economicus. In Arrows work of 1972, gifts and exchanges, he reformulates the utility theory, also known as an individual’s welfare, as follows; individual welfare is dependent on one’s own satisfaction and the satisfaction of others, and depends not only on the utilities of himself and others but also on the contribution

of utility to others, as if there is a social contract such that individuals perform duties to enhance satisfaction of all.

Another economic model which includes deviations from homo economicus is given by Becker (1976). He formulated an altruistic preferences model. Altruism, a form of behavior well discussed in psychological research, presumes that some people in some circumstances are, to a certain degree, capable of benefiting someone else at the expense of their own time, energy or possessions with no expectation of any direct or indirect benefits in return (Batson, 2014).

George Akerlof, mentioned earlier for his Nobel Prize winning lemon principle, is also the economist behind the fair wage hypothesis (1982). In this hypothesis Arkerlof no longer assumes that individuals are acting like a homo economicus, but he suggests that wages are partly dependent on gift exchanges between the employer and employees and not entirely on the market exchange price for labor. For now it is assumed that these gift exchanges are indicating altruistic behavior. Arrow (1972), Becker (1976) and Akerlof (1982) all describe altruistic behavior occurring, where according to the theory of homo economicus altruistic behavior is not expected to be present. This contradiction is further discussed in the section below.

With pure altruism a person does not necessarily get any utility of helping someone else. This form of altruism is a point of debate in the existing literature, where different terms of altruism are introduced, like ‘warm glow altruism’ being a form of impure altruism (Andreoni, 1988). Warm glow altruism defines behavior of someone that enjoys the act of helping or giving. By doing so a person gets some kind of utility out of it, with an egoistic motivation as for example boosting self-esteem or receiving recognition from other people, instead of a pure altruistic motivation. It is not in the scope of this research to further discuss the different theories of altruism, but it should be considered that pure altruism is the opposite in terms of the behavior of a homo economicus. The question that then rises is: Why are some economic theories based upon homo economicus and others on altruistic thoughts? For this research it is not the intention to answer this particular question, but it is assumed that some individuals are altruistic and others will act like homo economicus. What this research will do is try to make an attempt to give an insight in under which circumstances people act like homo economicus, so for example possibly fraud, cheat or conduct unethical behavior, are enhanced. The scope will be on one particular part of behavioral economics, namely the part concerning if people show more or less of these kind of unethical behaviors in pair or group settings which will be further discussed later on in this research.

2.3 Unethical behavior

According to the assumption that individuals act like homo economicus people should only tell the truth if this is incentive-compatible, given the material outcomes and thus tell a lie instead if he or she could benefit from it (Hölmstrom, 1979). According to Allingham and Sandmo (1972) there are no costs involved within the act of lying. A growing body of empirical research supports the occurrence of telling a lie, for example DePaulo and Kashy (1998), who report one to two lies a day. In recent economic experiments, for example the one of Gneezy (2005), consistent results are shown on reporting some kind of lies. He finds two results, first, individuals are sensitive to their potential gain when given the chance and following deciding to lie. Second, he finds that the amount of lying also depends on the loses of the individuals affected with the lie, indicating that there though might be a cost involved with the act of lying, in contrary with earlier suggestions. Cheating, a form of lying is also found in other psychological and behavioral experiments from for example Mazar et al. (2008), Gino et al. (2009) and Gino et al. (2013). Where Mazar et al. find that participants cheat, although not in a large degree, when given the opportunity. Gino et al. (2009) find that social norms are of influence on the level of cheating, independently of what the social norms are the authors find cheating in both treatments that have the opportunity to cheat. Gino et al. (2013) find that in a competitive setting, disadvantaged groups will be more likely to avoid regulation and engage in unethical behavior. In the experiment of Gino et al. (2013) the bigger part of these cheating participants where women, but the authors allocate this mainly to the competitive setting and not to the effect of gender differences. The possible effects of gender differences in unethical behavior are later on discussed in this thesis.

These experimental outcomes are not what is expected within existing social psychology literature, that demonstrates individuals trying to maintain a private and public positive self-image (Rosenberg, 1979) and individuals that want to feel good about themselves which most likely includes seeing themselves as moral (Schlenker, 1982). According to Sanitioso et al. (1990) people value honesty and Aquino and Reed (2002) find that 84% of individuals claim to be moral and honest. The subsequent paragraph will further elaborate these strands of contradicting literature.

2.4 Ethical dissonance gap and coping mechanisms

The gap between reported experimental outcomes with unethical or dishonest behavior and, according to literature, so called striving to maintain a moral self-image, is called the ethical dissonance gap by Barkan et al. (2012). With the ethical dissonance gap Barkan et al. (2012)

try to capture the tension that arises in individuals from inconsistency between one’s actual unethical behavior and one’s ethical and moral values. According to Barkan et al. (2012) this tension calls for some kind of adjustment or coping mechanism in individuals. For instance one coping mechanism mentioned by Mazar et al. (2008) is that in many situations people show enough dishonest behavior to benefit from it, but restrict the severity of their dishonestly behavior in such a manner that allows them to keep a positive moral self-image. In an experiment of Ariely (2008) this is shown by the distribution of six packs with cans of Coca-Cola in publicly shared university refrigerators. In some of these refrigerators besides the cans six one dollar bills were placed and when after 72 hours all the Coca-Cola cans where gone, the one dollar bills were still all present in the refrigerator. Most likely the individuals that took the cans could keep a positive moral self-image thinking they only took a drink from someone, where actually taking money with the same value of a can was perceived as to dishonest. A similar result is reported by Mazar et al. (2008) with a laboratory experiment where participants inflated their performance more when it concerned taking tokens instead of taking money, although the tokens represented the same monetary value later on in the experiment. The authors of both experiments conclude that taking non-monetary goods creates a greater leeway to make self-serving interpretations of their behavior and so lower the level of unethical behavior and keep a positive moral self-image.

Another coping mechanism is that individuals make unconscious errors in judging unethical behavior as ethical and so serving themselves with a positive moral self-image (Ayal and Gino, 2011). For example accountants, well paid for their auditing skills, that are not seeing mistakes in the books made by their clients. Bazerman et al. (2002) researched this phenomenon and conclude that participants were biased towards the interests of their own clients. Bazerman et al. (2002) gave undergraduate and business students a complex set of details about the upcoming sale of a fictitious company and divided the students in four groups; buyers, sellers and their auditors. The first result is not surprising, namely that the sellers valued the company higher than did the buyers. The second result was that the auditors of the sellers also came up with a higher price than did the auditors of the buyers. This difference persisted even when both of the auditors were asked to check the audit again and come up with a value as close as possible to the true value, determined by professionals. The students were promised a monetary reward, which was depending on how close their value was to the real value. Even then a difference of 30% remained between the buyer and the seller auditors. If these unconscious errors result in unethical or unethical behavior, this behavior should diminish with attending individuals to moral standards and so reducing

unethical or dishonest behavior. This is what Mazar et al. (2008) find, dishonest behavior reduces after attending participants on the Ten Commandments, a set of ethics originating from biblical literature. Participants with the option to cheat did not cheat after being reminded of the Ten Commandments, where others without the moral reminder did.

2.5 Group effects

Among others Diener et al. (1976) found more unethical behavior amongst infants when the children are part of a group instead of being alone. According to Zimbardo (1969) and Diener (1977) this could be explained by the presence of the group setting enabling individuals to distribute responsibility and guilt among group members. This could be seen as another coping mechanism to maintain a moral self-image. When individuals cannot distribute responsibility and guilt over the entire group, less unethical behavior is shown in the research of Diener et al. (1976). In the research of Diener et al. the influence of individuals being part of a group on unethical behavior is researched. In this thesis there is looked upon if this increase of unethical behavior is also present if the behavior of participants affect a group where individuals are not part of. As a control the amount of unethical behavior regarding groups is compared with the amount of unethical behavior against single individuals.

Another effect on unethical behavior, depending on being an individual or part of a group, is found by Gino et al. (2013) in a series of laboratory experiments. Gino et al. find that cheating behavior goes up when others benefit from it. The authors conclude that moral flexibility justifies the unethical behavior when such behavior benefits others in addition to the individual. Their findings suggest that when people's unethicaly benefits others, they are more likely to maintain a positive moral self-image and thus feel less unethical about benefiting from cheating and so coping with the dissonance gap. For this research the interest lays in, if this also works the other way around. With the experiment an attempt is made to test if this coping mechanism is also present when the negative effects of unethical behavior are spread over a group instead of over one individual. Does spreading the consequences of unethical behavior over a group lower the moral costs of unethical behavior enough, to cause more cheating in experiments with effects distributed over groups instead of individuals?

2.6 Gender effects

As mentioned before, Gino et al. (2013) found differences in unethical behavior between men and women. The authors allocated these differences to competitive disadvantages for the

women in their experiment and conclude that that is due to the influence on the behavior and not due to the gender difference. There do exist other behavioral experiments that indicate differences in behavior between sexes. Guth et al. (2007) performed a field experiment on a three way ultimatum game with one proposer, one responder and one dummy player. They find that females are more likely than males to offer all parties an equal part of total earnings. These results can be explained by on the one hand altruism or inequality aversion and on the other hand risk aversion. To see if it is caused by risk aversion, dictator game outcomes should be studied more carefully on gender effects. This is because in dictator games, risk aversion is ruled out.

Eckel and Grossman (1998) find that in an anonymous dictator game on average women give $1.60 and men give $0.82, this difference is statistically significant. In an experiment of Bolton and Katok (1995) on gender differences within a dictator game a smaller, not significant, difference is found. The authors find women giving on average $1.23 and men giving on average $1.13 . The difference between the experiments of Eckel and Grossman (1998) and Bolton and Katok (1995) is that the experiment of Eckel and Grossman is performed completely anonymous where the experiment of Bolton and Katok. It is interesting to see that changing the social conditions of the experiment, completely anonymous to less anonymous, makes women give $0.37 less and men $0.31 more. To control the social conditions of the experiment of this research as much as possible, the experiment is performed completely anonymous.

Amongst others Andreoni and Vesterlund (2001) found differences for women and men in standard dictator games. According to the authors women are more concerned with inequality and men are more concerned with maximizing efficiency. Dickinson and Tiefenthaler (2002) run a modified dictator game with a disinterested dictator and find the same results as Andreoni and Vesterlund. Selten and Ockenfelds (1998) perform an experiment with a different version of the dictator game, called the solidarity game. In that game three persons participated and each player could win ten Deutsche Mark with possibility two thirds. With a double blind condition players could give away parts of their earnings to one or both of the losers. The authors find education and false consensus effects, but also gender effects. The gender effects measured by Selten and Ockenfelds are that females are more generous than men.

Another modified dictator game is the group dictator game of Dufwenberg and Muren (2006), where groups of three divide money between themselves and a fourth individual. Dufwenberg and Muren (2006a) find that groups are more generous when woman are in the

majority. This is in line with the other literature, finding that women are more inequality adverse then men. In a second experiment of Dufwenberg and Muren (2006b), naming the receiving counterparts gender, male or female, while having almost no anonymity fails to find significant gender differences in giving. The authors address this failure to the non-anonymity of the experiment. This is partly explained by the results of the gender experiment of Ben-Ner

et al. (2004), in this completely anonymous dictator game, where sometimes gender information is given and in the other cases nothing. Giving gender information significantly affects giving in case of women, who give less to women than to men and persons of

unknown gender (Ben-Ner et al., 2004). One might think that this is not in line with literature

stating that women are more inequality averse, but there might be a reason for this behavior.

To support this Ben-Ner et al. (2004) performed a similar manipulation, where they did not

describe the gender of the receiver, but if the receiver lived in the state Minnesota or not. Within this experiment setup they found that women gave less to receivers that were not from Minnesota while by men there is no difference found in giving to from or not from Minnesota receivers. The latter results in more evidence that women are more sensitive to the social context of an experiment than men, in these cases where information about the gender and the

home state of the receiver is revealed (Croson and Gneezy, 2009). This is in line with the

previously discussed literature of the difference between the experimental outcomes of Eckel

and Grossman (1998) and Bolton and Katok (1995). This is why in this research there will also be looked upon the differences in behavior between men and women.

2.7 Earned position and wealth effects

Except for gender differences there are many factors in experiments that can be of influence on the experimental outcome, like the aspects of the decision environment and the context of interactions. One example that could be of influence on this research is found by Hoffman and Spitzer (1985) and Hoffman et al. (1994). They find that when individuals earn their roles in an experiment, like first mover or seller, the payoffs are different then when roles are randomly assigned. Among others Cherry et al. (2002) find that payoff distributions in dictator games are depending on the source of wealth. This holds that participants of experiments with rightful claims to wealth, having earned these wealth, receive more than when the wealth is distributed randomly by the experimenter, and so being unearned. Cherry et al. (2002) performed their research on undergraduate students who were unfamiliar with games like dictator or ultimatum games. Participants are divided in participants A and B, where the participants A first entered an earning stage where they could earn $40 with

answering ten out of seventeen Graduate Management Admission Test (GMAT) questions correct. With less than ten questions correct participants A earned $10. After the earning stage the participants A were divided into two rooms with A1 containing participants A who earned $40 and in A2 containing participants A who earned $10. Participants in A1 and A2 are now randomly assigned to participants B, who did not have earnings stages, after which a dictator game is performed. In this dictator game participants in A1 and A2 are the dictator and the participants B are the receivers. This holds that participants A are dividing their own earned wealth. In the control experiment there is no earning stage and the experimenter randomly assigns participants A to A1 and A2 and distributes $40 and $10 wealth to them accordingly and participants A are the dictator again and participants B the receivers. Therefore the wealth of participants A is unearned in the control experiment. Comparing the control and the treatment experiment Cherry et al. (2002) find zero offers in 79% and 70% in respectively the high and low earnings treatment and only 15% and 19% of zero offers in the unearned high and low control experiment.

As an addition to the research of Cherry et al. (2002), Oxoby and Spraggon (2008) repeat the research of Cherry et al. (2002) and add one treatment. In this added treatment the receiver of the dictator game earns the wealth instead of the dictator, after which the dictator allocates the wealth between himself and the receiver. Oxoby and Spraggon (2008) thus have one control treatment, with distributed wealth, one dictator earned treatment and one receiver earned treatment. Another differences is that Oxoby and Spraggon use three levels of wealth; $10 for zero to eight correct answers, $20 for nine to fourteen correct answers and $40 for fifteen or more correct answers of the twenty GMAT questions. In the receiver earned wealth treatment the dictator does not learn the correct amount of answers but only the amount of money earned by the receiver, afterward which the dictator divides the receiver earned money between himself and the receiver. Oxoby and Spraggon find 35%, 26% and 11% zero offers in the $10, $20 and $40 control treatment, what is in line with other dictator treatments. Consistent with Cherry et al., Oxoby and Spraggon find 100% zero offers in the dictator earned treatment. Further strengthening the importance of adding property rights to experiments is the outcome of the receiver earned treatment, where some dictators offer 100% of the earning to the receiver and in the $40 receiver earned treatment there are no dictators that offer zero to the receiver. This is, as mentioned before, contrary to the control treatment. In the $20 and $10 receiver earned treatment, respectively 3% and 38% of the dictators offered zero to the receivers. Comparing these experiment results, the $20 and $40 show statistical significant differences between the receiver earned, dictator earned and the control

treatment. Comparing the three $10 experiments results the receiver earned, dictator earned and the control treatment no significant difference are found.

Because decisions made in real life almost always concern earned positions and earned wealth, in this research there is chosen to perform the experiment with an earnings stage before the decision stage. This is for the reason that if the results of this research are to be translated to real world examples, it is as close to the real world as a lab experiment can possibly get and with the highest external validity possible.

3. Methodology

In this chapter the hypotheses used for testing if there is a significant difference in unethical behavior will be formulated. After that respectively the experimental design and sample will be discussed.

3.1 Hypotheses

In the previous chapter studies reviewing unethical behavior were discussed. Also contradictory social psychology literature that demonstrates individuals trying to maintain a private and public positive self-image and people that want to feel good about themselves which likely includes seeing themselves as moral was considered in the previous chapter. Despite of this social psychology research and based on recent research that has found that when individuals have the opportunity to cheat when the probability of being caught and reputational costs are minimized, most people do cheat, but not as much as they could (Ayal and Gino, 2011 and Gino et al., 2009). They cheat enough to benefit financially, but not to the extent that they feel obligated to negatively revise their self-image (Mazar and Ariely, 2006). Taking all of this into account the first hypothesis of this research is:

Hypothesis 1: Participants show unethical behavior in the experiment.

If this first hypothesis cannot be rejected it is possible to test the second hypothesis of this research in an attempt to answer the research question: ‘Is there a difference in unethical behavior when the negative cost of this unethical behavior is born by a group or by one individual?’ Do people use the distribution of negative effects of unethical behavior over multiple group members as a coping mechanism to close the ethical dissonance gap and so succeeding in maintaining a positive self-image despite of their unethical behavior? To test this, the following hypothesis is formulated:

Hypothesis 2: Participants show equal amounts of unethical behavior whether the negative effects concern a group or one individual.

For this research there is chosen to formulate the second hypothesis as the amount of unethical behavior being equal in both treatments. This is chosen based on the assumption that when people act like homo economicus or think rational, there should be no difference in taking from a group or from one individual. The latter is because if everybody thinks stealing

the same amount from a group is better, everybody would steal from the group and group members end up with zero, what in fact is the same as stealing everything from one individual at once. If the second hypothesis is rejected, people do not think rational. Before any conclusions can be made about possible coping mechanisms and homo economic or altruistic motivations, it should be researched if this difference is prevalent among both men and women. This is because prior research of amongst others Eckel and Grossman (1998) and Bolton and Katok (1995) has shown women being more sensitive for the social context of an experiment and so the third hypothesis is the following:

Hypothesis 3: Men and women show equal levels of unethical behavior in the experiment.

The third hypothesis is formulated as behavior of men and women being equal. This is because based upon previous research it is not known if women exhibit more or less unethical behavior then men in this particular experiment, but only that women are more sensitive for the social context of an experiment. All three hypotheses concern unethical behavior, how this unethical behavior is measured in the experiment will be discussed in the subsequent paragraph about the experimental design.

3.2 Experimental design

To answer the hypotheses stated above an experiment was run at the CREED laboratory for experimental economics, at the University of Amsterdam. The experiments are all run on computers in the CREED laboratory. In total two sessions were run, with 48 students from different disciplines. In both sessions there were two treatments tested simultaneously. Because it regarded a between groups experimental design, participants only participated in one treatment. The decision to run two treatments in one session was made because participants learned the median performance during the experiment, before they had to make a decision. If each session would test one treatment, participants might learn different median performances due to population differences and so possibly behave different, if decisions are based upon median performance. With this design in each session, participants in both treatments observe the same median.

In the experiment participants first enter an earning stage, to earn wealth for the second part of the experiment. In the earning stage of the experiment participants are given twenty matrices with twelve numbers inside all depicted in Appendix A. The matrices were randomly put together by a spreadsheet application. In each matrix only two of the twelve

numbers sum up to ten, the participants were asked to check the two boxes that exactly sum up to ten. The earning stage of the experiments took seven minutes and participants had to correctly solve as many matrices as possible, each correct solved matrix possibly earning them €1.25.

After part one, the earnings stage, participants were divided in participants A and participants B. In part two of the experiment, the decision stage, Participants A became dictators and participants B became the receivers. The dividing of the participants in A and B players is based upon performance, the participants on the low side of the median performance of the earnings stage became participants A and participants on the high side of the median became participants B.

In the decision stage, participants A learned their own performance and the median performance of all players of the session. With their own performance participants A learned how much they had earned in part one of the experiment. Participants A were told that they could choose their own earnings and that everything they chose to earn above their earned amount of part one is subtracted from the earnings of a group of five or from one individual, depending on which treatment they were in. From now on the treatment concerning taking from a group will be referred to as the group treatment and the treatment concerning taking from an individual will be referred to as the pair treatment. To ensure no negative earnings, participants A can only steal the median earnings and because they steal from above median participants B, participants B cannot earn below zero. See Appendix B for the pair treatment instructions and Appendix C for the group treatment instructions.

In this research unethical behavior is specified as follows: If the actual earnings (E) at the end of the experiment of a participant A, a dictator in the second part of the experiment, is above the earned wealth (W) of the first part of the experiment, then this amount of earnings is subtracted from other participants and is interpreted as unethically earned and following this expression, these participants are assumed to exhibit unethical behavior. If participants A in part two claim their earnings to be the same as the wealth they earned in part one, the behavior is interpreted as ethical and if they claim lower earnings then earlier earned wealth the behavior is assumed to be altruistic. The latter is depicted below in Table 1. The last type of behavior is called altruistic because the second part of the experiment is an anonymous one shot game and so participants could not have reputation concerns or reciprocity intentions when under-claiming their earnings.

As discussed in the previous chapter, the level of anonymity could influence decisions, so therefore it is chosen to perform the experiment completely anonymous, participants are randomly seated and participants will not learn with whom they are matched with during or after the experiment.

Table 1: Possible experiment behavior a. E(part2) - W(part1) > 0 Unethical behavior b. E(part2) - W(part1) = 0 Ethical behavior c. E(part2) - W(part1) < 0 Altruistic behavior

3.3 Sample

As mentioned before, the experiment consisted of two sessions, with in the first and second session 24 participants. In the first session 7 female and 17 male students attended, and in the second session 10 female and 24 male students attended. Ages form participants ranged from 19 being the youngest to the oldest being 58 years old. Subjects were selected on being able to read and speak Dutch. The 48 students had different educational backgrounds, which are classified in eight fields of study. Besides these eight fields of study there is one category added, unidentified, in which one student answered ‘different’ and three others answered an unidentifiable answer. Because the second part of the experiment is a kind of a dictator game, participants where asked if they have had courses in game theory. In the experiment 24 participants followed courses with game theory and 24 did not. For the fields of study, game theory courses and numbers of students, see Table 2 below.

Table 2: Fields of study participants

Dentistry 1

Economics and Business 23

Humanities 5

Law 3

Medicine 1

Psychology 3

Science 5

Social and Behavioral science 3

Unidentified 4

4. Results

In this chapter the main findings of the research will be discussed. In the first paragraph some descriptive statistics will be given. In paragraph two there is discussed if participants show unethical behavior Subsequently, in paragraph three the results indicating if participants show equal amounts of unethical behavior in both treatments are presented and in paragraph four is discussed if men and women exhibit different levels of unethical behavior in the experiment.

4.1 Descriptive statistics

In table 3, 4, 5 and 6 below, the descriptive statistics of the experiment are given. The median number of the total sample is 12 and thus participants scoring below 12 become participants A in the second part of the experiment. In this research unethical behavior is specified as follows: If the actual earnings (E) at the end of the experiment of a participant A, a dictator in the second part of the experiment, is above the earned wealth (W) of the first part of the experiment, then this amount of earnings is subtracted from other participants and is interpreted as unethically earned. See table 1 in the previous chapter for mathematical explanations. Comparing the means of these statistics about unethical behavior, it looks like there is seen more unethical behavior in the group treatment, with a mean of 9.73 in comparison with the pair treatment which had a mean of 6.53, when we look at the total sample. This difference is not shown when there is looked upon the differences of the means of unethical behavior in the pair and group treatment of men, with means of 13.75 for the pair treatment and 10.83 for the group treatment. This could possibly be because of the low number of two observations for men in the pair treatment. The difference between the pair and the group treatment present in the total sample then must be explained by a difference in behavior by women. This is in fact what is seen when the mean of unethical behavior of the pair and group treatment of women is looked upon, for women the mean unethical behavior of the pair treatment is 4.46 and for the group treatment it is 5.33. In the next paragraphs of this chapter the three different hypotheses, formulated to answer the research question, will be tested and doing so there is looked upon if the above found means are statistically significant different from each other.

Table 3: Descriptive statistics correctly solved matrices

Correctly solved matrices n Mean (s) Min Max Median

Total sample 48 13.13 (3.80) 2 19 12 Men 31 13.61 (3.83) 2 19 13 Women 17 12.24 (3.70) 5 19 12

Table 4: Descriptive statistics total sample

n Mean (s) Min Max

Correctly solved matrices 48 13.13

(3.80) 2 19

Unethical behavior

pair treatment 9

6.53

(6.25) 0 17.50

Unethical behavior group

treatment 15

9.73

(7.13) -1.5 17.5

Table 5: Descriptive statistics men

n Mean (s) Min Max

Correctly solved matrices 31 13.61

(3.83) 2 19

Unethical behavior

pair treatment 2

13.75

(5.30) 10 17.50

Unethical behavior group

treatment 12

10.83

(6.69) -1.25 17.50

Table 6: Descriptive statistics women

n Mean (s) Min Max

Correctly solved matrices 17 12.24

(3.70) 5 19

Unethical behavior

pair treatment 7

4.46

(5.00) 0 15

Unethical behavior group

treatment 3

5.33

(8.61) -1.5 15

4.2 Unethical behavior

In this paragraph the first hypothesis (Participants show unethical behavior in the experiment) will be tested. To test if participants show unethical behavior there is looked upon the amount of earnings of participants A, the dictators of part two of the experiment. In theory the amount earned by participants A at the end of the experiment must be the same or less as what they earned in part one of the experiment, to possibly reject the first hypothesis. If the actual earnings of part one of the experiment, the earnings stage, are subtracted from the earnings of the participants A and there is a positive amount left, unethical behavior is expected to be present. To test this, the mean of the actual earnings of participants A at the end of the experiment minus the earnings in part one of the experiment of participants A is tested with a

one-sample location test. There is tested if the location is above zero, for equal to and below zero representing no unethical behavior and above zero being unethical. With a sample size larger than thirty this one-sample location test is performed with a one sample t-test, but when the sample size is smaller than thirty the data must be normally distributed to correctly use the t-test. Because in the experiment of this thesis the sample size is 24, we perform a Shapiro-Wilk test to test for normality, see Appendix D for the detailed test outcome. Because the zero hypothesis of the data having a normal distribution is rejected, it is assumed that the data is not normally distributed. Therefore to test the first hypothesis using a one sample t-test is not recommended. To correctly test the first hypothesis, we use a sign test. Because 20 of the 24 participants show unethical behavior, two divide accordingly to the amount of matrices solved and two even give more to the receiver than actually earned, it is expected that the first hypothesis (Participants show unethical behavior in the experiment) cannot be rejected. The latter is in fact the case, for the detailed test results of the sign test performed see Appendix E. The results indicate that the mean of unethical behavior for the participants of this experiment is statistically significantly larger than zero with a significance level of 1% and a probability level of 0.01%. The Hypothesis 1: Participants show unethical behavior in the experiment is not rejected.

4.3 Differences in unethical behavior between treatments

In this paragraph the second hypothesis, Participants show equal amounts of unethical behavior whether the negative effects concern a group or one individual, will be tested. To test if the amount of unethical behavior differs, we look if the difference between the earnings of part one and the actual earnings at the end of the experiment of participants A, differs between the participants in the pair treatment and the participants in the group treatment. To test if there is any difference in unethical behavior between both treatments we must compare both samples. Because n = 9 in the pair treatment and n = 15 in the group treatment, first we test if both treatments separately are normally distributed. This is again tested with the Shapiro-Wilk test, for detailed test results see Appendix F and G. Because of the small sample size and because of the data of the group and the pair treatment being not normally distributed, with respectively a significance level of 5% and 10%, it is not recommended to compare both means with an independent t-test. This is why we use the Mann-Whitney U test (also known as the Wilcoxon rank-sum test), for the detailed test results see Appendix H. Looking at the results the hypothesis stating that participants show equal amounts of unethical behavior, hence the two groups do not significantly differ from one another, whether the

negative effects concern a group or one individual cannot be rejected at a significance level of 10%. The probability of both samples having the same rank-sum and thus being equal to each other is 38%.

4.4 Gender differences in unethical behavior

In this paragraph the third hypothesis, men and women show equal levels of unethical behavior in the experiment, will be tested. Because this difference in unethical behavior in the total sample is not significant, according to the former paragraph, it might be that the differences between men and women offset the difference in unethical behavior between treatments in the total sample. As discussed in the first paragraph of this chapter the difference seen in the means of unethical behavior between treatments in the total sample, being more unethical in the group treatment, is coming from women showing on average more unethical behavior in the group treatment in comparison with the pair treatment, because men on average show more unethical behavior in the pair treatment then in the group treatment.. This is what we look at in this paragraph, is the unethical behavior exhibited by men and women is equal in both treatments? To answer this question we first test both gender’s unethical behaviors separately on normality with the Shapiro-Wilk test, for the detailed test results see Appendix I and J. The null hypothesis of being normally distributed is rejected for women at a significance level of 5%, with the probability level of being normally distributed being 2.0%. For men this is not the case, the null hypothesis of being normally distributed can only be rejected with a significance level of 10%, with the probability level of being normally distributed being 6.10%. Because the unethical behavior of women could not be assumed being normally distributed again we use the Mann-Whitney U test to test if the rank-sum of unethical behavior of both genders per treatment is equal. For the detailed test results see Appendix K and L. Looking at the results the hypothesis that men and women exhibit equal levels of unethical behavior in the experiment is rejected for the pair treatment at a 10% significance level. The probability level of having the same rank-sum, and thus being equal, is 7.70%. The hypothesis that men and women exhibit equal levels of unethical behavior in the experiment could not be rejected for the group treatment at a 10% significance level. The probability level of having the same rank-sum, and thus being equal, is 18.55%.

Because there is a possible difference between men and women in the pair treatment, we decided to test if there is a treatment effect for men and women separately. Because there are only two results for men in the pair treatment, the Shapiro-Wilk test for normality could

not be performed on the pair treatment test results. Besides this, the null hypothesis of being normally distributed is rejected for the test results of unethical behavior in the group treatment of men with a significance level of 5%, for the detailed test results see Appendix M. Because of this we test the treatment effect among men with a Mann-Whitney U test, for the detailed test results see Appendix N. The Mann-Whitney U test tests if the rank-sum of the pair and group treatment of men are equal. This could not be rejected at a significance level of 10%. The probability level of having the same rank-sum, and thus being equal, is 45.87%. This holds that no statistically significant treatment effect is found within the sample that contains only men.

First we separately test the test results of both treatments of women for normality with the Shapiro-Wilk test, for the detailed test results see Appendix O and P. The null hypothesis of being normally distributed could not be rejected for the group treatment at a significance level of 10%, it is rejected for the pair treatment with a significance level of 5% and a probability level of being normally distributed of 2.00%. Because not both test results of unethical behavior of women can be assumed normally distributed, we test the treatment effect of women with the Mann-Whitney U test, for the detailed test results see Appendix Q. The Mann-Whitney U test tests if the rank-sum of the pair and group treatment outcomes of women are equal. This could not be rejected at a significance level of 10%. The probability level of having the same rank-sum, and thus being equal, is 73.01%. This means that also no statistically significant treatment effect is found within the sample that contains only women.

5. Discussion

Based upon literature from Zimbardo (1969) Diener et al. (1976) Diener (1977), Gneezy (2005) and Gino et al. (2013) a difference in unethical behavior is expected when the negative effects are distributed over a group or over one individual. However, within this lab experiment the latter effect is found not to be statistically significant. There could be two different explanations. First, people can be rationale and realize that if everybody distributes the hurt over the group that that has the same effect as hurting one individual. Second, because of the small sample size and relatively large variances the found differences in mean unethical behavior are not significant in this sample, but might be in a larger sample.

The sample size of this experiment is 48 participants who together deliver 24 observations. The ideal situation would have been if these 24 observations where evenly divided over both treatments with the same amount of men and women in both treatments. In the experiment 17 females and 31 males participated. Therefore the dividing of male dictators (twelve) and female dictators (twelve) over both treatments could have been equal. This in fact was not the case, because subjects earned their roles in part one of the experiment and were of course randomly assigned to the treatments, so the dividing of both genders over both treatments was not so equally as preferred for statistical analyses. If there was to be no difference in behavior between men and women this would not have been a problem, but based on the test results of the statistically significant behavior of men and women in the pair treatment and based on the existing literature of amongst others Eckel and Grossman (1998) and Bolton and Katok (1995), this unequal distribution is a problem. To further enlighten this problem we perform an extra Mann-Whitney U test, to test the difference in unethical behavior between men and women regardless of the treatment they are in, for the detailed test results see Appendix R. The test results show that the rank-sum of unethical behavior of men and women being equal is rejected at a significance level of 5%. The probability level of having the same rank-sum, and thus being equal, is 2.12%. Because of this difference between gender and the unequal dividing of males and females over the treatments, it is not statistically accurate to test the treatment effects in the total sample.

The next problem arises when there is looked at the treatment effects at gender level, the dividing of females and males over both treatments is also not in proportion, with seven females in the pair treatment and only three in the group treatment. This problem is even larger with males, where there are only two observations of males in the pair treatment and twelve in the group treatment. Because of these small samples, there are large variances.

Large variances cause statistical tests failing to reject incorrect null hypothesis of means being equal. Another reason for this failing to reject incorrect null hypothesis is the restricted range of unethical behavior, ranging between just below zero and the median performance of 17.50 (Because of low occurrence, altruistic behavior and thus values up to 17.50 below zero are not considered constructing the possible range of unethical behavior).

Another possible explanation for the differences in unethical behavior might be found in the performance of part one of the experiment. Participants A, the dictators of part two, in that part two learn the median behavior. They also learn they have scored under the median performance. Because of this their decision in part two might be altered based upon how much they score under the median performance instead. To test this the correlation coefficient of performance in round one and the amount of unethical behavior in part two is calculated. The overall correlation coefficient of performance and unethical behavior is -0.05. Meaning that in the total experiment sample there can assumed to be no correlation. The correlation of performance and unethical behavior for males and females separately are respectively -0.24 and 0.19. Taking into account the small sample size, these last two correlations are assumed to be non-existent as well. Because the dividing in roles of participants is based upon performance, and so there is no data collection of behavior in part two for all participants, there is a selection bias that rules out further testing performance relationships.

One other problem of the small sample size is the power of the Mann-Whitney U test, the statistic test that is mainly used in this thesis. The Mann-Whitney U test has power advantages over the t-test regarding distributions that are not normally distributed, mainly regarding distributions that have great skewness. Besides the latter, the Mann-Whitney U test does also have power advantages when the sample size is relatively small, however these power advantages disappear when sample sizes gets to small (Blair and Higgins, 1980; Bridge and Sawilowsky, 1999). When we test the observations of unethical behavior on skewness, no skewness is found with a significance level of 5%. This is why we repeated the Mann-Whitney U tests, performed to test the hypotheses, with independent t-tests to see if any significant differences in unethical behavior are found. With these t-tests no significant differences other than with the Mann-Whitney U test are found, for the detailed test results see Appendix S,T,U,V and W.

Because of these problems with sample size, for further research it is suggested to enlarge the sample size for better statistical analyzes afterwards. With the experimental design no major problems are found. Something questionable that has to be noted is that for practical experiment payout reasons, like the inability to score below zero, in this experiment is chosen

to make the below median performers of the first part of the experiment the dictators in the second part and the above median performers the receivers. It is not known if there is any effect on the outcome of the experiment because only the decisions of participants with lower abilities are taken into account. Also it is recommended that for further research men and women are tested separately, based on existing literature and measured differences in test outcomes. In previous discussed literature about earned wealth the importance of earning wealth in dictator games is discussed, to increase the external validity of these kind of experiments, it is suggested to test with different kinds of earning tasks to compare if there is any difference in unethical behavior after the wealth is earned with a fun, tedious, hard or easy task.

6. Conclusion

In an attempt to better understand and to give an insight in the public question how to stop all the kinds of frauds, cheating and unethical behavior of by for example Enron, Bernard Madoff, Barclay’s and KPMG, this research looks upon an explanation for this behavior given by Gneezy (2005). Gneezy states people being more accepting of fraudulent behavior hurting a large organizations or a rich counterpart than fraudulent behavior hurting only an individual, where despite of the monetary costs being identical, according to Gneezy hurting an individual with the same monetary amount is perceived as greater. To see if the former explanation of Gneezy holds, an attempt is made to answer the following research question: Is there a difference in unethical behavior when the negative cost of this unethical behavior is born by a group or by one individual? This is done by performing a two treatment experiment which measures the amount of unethical behavior exhibited by a dictator. In one treatment the dictators decisions affected one person and in the other treatment a group of five. Although no statistical different behavior is found between both treatments, there is found a difference between behavior of men and women in one treatment. This is why there is looked upon a treatment effect within sexes, but this research failed to find a significance treatment effect at gender level in this sample. It is presumably because of the sample size, and it is not sure if the effect is not there. The gender differences found in literature and in this research, call for further testing of differences in unethical behavior but then for men and women separately.

References

Akerlof, G. A. (1982). Labor contracts as partial gift exchange. The Quarterly Journal of Economics, 97(4), 543-569.

Allingham, M. G., & Sandmo, A. (1972). Income tax evasion: A theoretical analysis. Journal of public economics, 1(3), 323-338.

Andreoni, J. (1988). Privately provided public goods in a large economy: the limits of altruism. Journal of Public Economics, 35(1), 57-73.

Andreoni, J., & Vesterlund, L. (2001). Which is the fair sex? Gender differences in altruism. Quarterly Journal of Economics, 293-312.

Aquino, K., & Reed II, A. (2002). The self-importance of moral identity. Journal of personality and social psychology, 83(6), 1423.

Ariely, D. (2008). Predictably irrational: the hidden forces that shape our decisions New York: HarperCollins.

Arrow, K. J. (1962). Economic welfare and the allocation of resources for invention. The National Bureau of Economic Research, 609-625.

Arrow, K. J. (1972). Gifts and exchanges. Philosophy and Public Affairs, 1(4), 343-362. Ayal, S., & Gino, F. (2011). Honest rationales for dishonest behavior. The Social Psychology

of Morality: Exploring the Causes of Good and Evil. Washington, DC: American Psychological Association.

Barkan, R., Ayal, S., Gino, F., & Ariely, D. (2012). The pot calling the kettle black: Distancing response to ethical dissonance. Journal of Experimental Psychology: General, 141(4), 757.

Batson, C. D. (2014). The altruism question: Toward a social-psychological answer. New York: Psychology Press.

Bazerman, M. H., Loewenstein, G., & Moore, D. A. (2002). Why good accountants do bad audits. Harvard business review, 80(11), 96-103.

Becker, G. S. (1976). Altruism, Egoism, and Genetic Fitness: Economics and Sociobiology. Journal of Economic Literature, 14(3), 817-826.

Ben-Ner, A., Kong, F., & Putterman, L. (2004). Share and share alike? Gender-pairing, personality, and cognitive ability as determinants of giving. Journal of Economic Psychology, 25(5), 581-589.

Blair, R. C., & Higgins, J. J. (1980). A Comparison of the Power of Wilcoxon's Rank-Sum Statistic to that of Student's t Statistic Under Various Non normal

distributions. Journal of Educational and Behavioral Statistics, 5(4), 309-335.

Bolton, G. E., & Katok, E. (1995). An experimental test for gender differences in beneficent behavior. Economics Letters, 48(3), 287-292.

Bridge, P. D., & Sawilowsky, S. S. (1999). Increasing physicians’ awareness of the impact of statistics on research outcomes: comparative power of the t-test and Wilcoxon rank- sum test in small samples applied research. Journal of clinical epidemiology, 52(3), 229-235.

Cherry, T. L., Frykblom, P., & Shogren, J. F. (2002). Hardnose the dictator.American Economic Review, 92(4) 1218-1221.

Croson, R., & Gneezy, U. (2009). Gender differences in preferences. Journal of Economic literature, 448-474.

DePaulo, B. M., & Kashy, D. A. (1998). Everyday lies in close and casual relationships. Journal of personality and social psychology, 74(1), 63.

Dickinson, D. L., & Tiefenthaler, J. (2002). What is fair? Experimental evidence. Southern Economic Journal, 414-428.

Diener, E. (1977). Deindividuation: Causes and consequences. Social Behavior and Personality: an international journal, 5(1), 143-155.

Diener, E., Fraser, S. C., Beaman, A. L., & Kelem, R. T. (1976). Effects of deindividuation variables on stealing among Halloween trick-or-treaters. Journal of Personality and Social Psychology, 33(2), 178.

Dufwenberg, M., & Muren, A. (2006a). Gender composition in teams. Journal of Economic

Behavior & Organization, 61(1), 50-54.

Dufwenberg, M., & Muren, A. (2006b). Generosity, anonymity, gender. Journal of Economic

Behavior & Organization, 61(1), 42-49.

Eckel, C. C., & Grossman, P. J. (1998). Are women less selfish than men?: Evidence from dictator experiments. The economic journal, 108(448), 726-735.

Gino, F., Ayal, S., & Ariely, D. (2009). Contagion and differentiation in unethical behavior the effect of one bad apple on the barrel. Psychological Science, 20(3), 393-398. Gino, F., Ayal, S., & Ariely, D. (2013). Self-serving altruism? The lure of unethical actions

that benefit others. Journal of economic behavior & organization, 93, 285-292. Gino, F., Krupka, E. L., & Weber, R. A. (2013). License to cheat: Voluntary regulation and

ethical behavior. Management Science, 59(10), 2187-2203.

Gneezy, U. (2005). Deception: The role of consequences. The American Economic Review, 95(1), 384-394.

Güth, W., Schmidt, C., & Sutter, M. (2007). Bargaining outside the lab–a newspaper experiment of a three person ultimatum game. The Economic Journal, 117(518), 449-469.

Hoffman, E., McCabe, K., Shachat, K., & Smith, V. (1994). Preferences, property rights, and anonymity in bargaining games. Games and Economic Behavior, 7(3), 346-380. Hoffman, E., & Spitzer, M. L. (1985). Entitlements, rights, and fairness: An experimental

examination of subjects' concepts of distributive justice. The Journal of Legal Studies, 259-297.

Hölmstrom, B. (1979). Moral hazard and observability. The Bell Journal of Economics, 10(1), 74-91.

Ingram, J. K. (1888). A History of Political Economy. Reprint, New York: Augustus M. Kelley, 1967.

Mazar, N., Amir, O., & Ariely, D. (2008). The dishonesty of honest people: A theory of self concept maintenance. Journal of marketing research, 45(6), 633-644.

Mill, J. S. (1874). Essays on Some Unsettled Questions of Political Economy. London: Longmans, Green, Reader and Dyer (second edition).

Nash, J. F. (1950). Equilibrium Points in n-Person Games. Proceedings of the National Academy of Sciences of the United States of America, 36(1), 48-49.

Oxoby, R. J., & Spraggon, J. (2008). Mine and yours: Property rights in dictator games. Journal of Economic Behavior & Organization, 65(3), 703-713.

Persky, J. (1995). Retrisoectives: The Ethology of Homo Economicus. The Journal of Economic Perspectives, 9(2), 221-231.

Rosenberg, M. (1986). Conceiving the self.Publisher, Basic Books, 1979. Original from, the University of Michigan.

Sanitioso, R., Kunda, Z., & Fong, G. T. (1990). Motivated recruitment of autobiographical memories. Journal of Personality and Social Psychology, 59(2), 229.

Selten, R., & Ockenfels, A. (1998). An experimental solidarity game. Journal of economic behavior & organization, 34(4), 517-539.

Schlenker, B. R. (1982). Translating actions into attitudes: An identity-analytic approach to the explanation of social conduct. Advances in experimental social psychology, 15, 193-247.

Zimbardo, P. G. (1969). The human choice: Individuation, reason, and order versus deindividuation, impulse, and chaos. In Nebraska symposium on motivation. University of Nebraska Press.

Appendix A. Matrices 1 4.147 8.679 2.210 8.283 3.356 9.130 1.725 0.444 0.282 9.718 1.648 3.118 2 3 4 5 6.630 0.592 9.408 4.151 1.034 0.222 7.586 0.593 6.206 0.227 6.677 1.485 6 6.179 3.111 6.363 1.798 1.752 0.149 4.537 9.961 2.756 7.141 8.044 3.821 0.397 4.996 8.768 7.064 8.112 1.329 8.647 2.504 3.742 6.637 6.258 8.142 8.007 5.882 3.957 3.440 1.759 6.560 6.893 4.391 3.764 9.168 6.771 2.259 28

3.831 6.525 4.517 3.557 3.935 4.103 9.366 5.647 2.164 1.340 2.695 5.897 7 3.730 0.689 6.270 4.009 0.585 5.499 2.792 8.351 9.621 0.963 1.406 0.406 8 8.009 4.912 9.820 3.153 1.535 4.830 5.918 7.170 1.814 5.054 4.946 3.088 9 3.587 9.606 4.996 4.580 1.941 0.394 0.068 7.590 0.024 8.692 0.559 0.998 10 1.790 3.381 0.637 4.489 6.539 8.210 2.016 8.178 5.200 5.079 7.772 0.171 29

11 2.608 5.343 7.392 5.896 0.301 5.846 1.219 1.831 9.187 7.723 7.787 3.417 12 1.079 1.351 1.357 8.873 2.334 6.020 7.531 1.352 7.417 8.643 0.281 3.229 13 2.606 5.026 4.974 6.833 0.858 7.528 5.361 4.464 9.408 9.339 3.797 6.107 14 4.738 9.442 7.119 9.707 5.656 8.452 0.972 6.300 0.305 0.293 4.983 0.624 15 5.174 1.305 7.041 8.904 8.184 5.005 1.290 7.433 1.486 8.695 6.151 2.400 30

16 6.485 0.348 5.829 1.425 0.016 0.898 6.054 2.322 3.946 7.462 4.983 1.656 17 7.503 4.065 5.610 3.700 0.406 3.712 9.092 4.758 1.950 4.761 5.044 9.594 18 3.794 7.874 2.194 6.349 1.228 2.970 1.158 3.176 7.706 8.772 2.915 6.186 19 4.933 2.449 4.870 2.877 6.775 4.938 6.776 4.373 2.164 4.716 0.847 5.284 20 9.280 3.046 8.791 3.888 0.809 2.165 4.725 2.136 3.002 2.172 2.968 6.112 31

B. Experiment instructions pair treatment

Instructions

Introduction

Welcome to this experiment. Talking is strictly forbidden during the experiment. If you have a question, please raise your hand and I will come to your table to answer your question. The experiment will last about 15 minutes. The experiment consists of two parts. Below you find the instructions for the first part. After the first part is completed, you will receive instructions for the second part.

Part I

In part one of the experiment there are twenty tasks. Each task is as follows: You are given a matrix with 12 numbers. In each matrix, (only) two numbers sum up to 10. You are asked to find these two numbers that sum up to 10. Each correct answer is worth 1.25 Euros.

You can give your answer by checking the boxes next to the numbers. When two boxed are checked, you can proceed to the next task. Your answer will count only if you press the submit button. If you check more or less than two boxes, you will get a warning message. For the first part of the experiment you have seven minutes. When the seven minutes are over, you will proceed to the second part of the experiment.

Example:

Below is an example matrix. Find the correct answer.