Does everyone have a piece of Scrooge McDuck hiding in them? : the influence of money on social economic behavior : a study focused on how people their social economic behavior may be influenced by their relative amount

(1)

Does everyone have a piece of Scrooge McDuck

hiding in them? The influence of money on social

economic behavior.

A study focused on how people their social economic behavior may be

influenced by their relative amount of economic resources.

Amsterdam, September 6th 2014

Derek T. Westra, 6067751

Thesis MSc Economics, track Behavioural Economics and Game Theory Universiteit van Amsterdam

(2)

2

In everyday life people make several decisions a day, while during that day there are many factors that, direct or indirect, have an influence on the outcome of these decisions. One of these factors that is present for everyone, every second of the day, is someone’s socioeconomic status (SES). Someone’s SES is a measure of his economic and social position in relation to others, based on his economic resources, education, and occupation (Adler et al., 2002). The more economic resources, the better your education and the better your occupation, the higher your SES and vice versa. Because your SES is always present in some way, it could be the case that your SES has a direct or indirect influence on each decision you

make. Figure 1

In psychology this phenomenon has already been studied manifold. One of the earliest and most famous studies is that of Maslow (1943) and his hierarchy of needs (figure 1). This hierarchy of needs consists of five stages and one can “grow” to a higher stage if a lower stage of needs is satisfied. The people with the highest SES (also called the upper class

individuals) stand on top of the hierarchy and have primarily self-actualization needs – realizing personal potential, while people with a low SES (also called the lower class individuals) prioritize on other needs, primarily social needs, needs for safety, or even physiological needs. Thus people make decisions in accordance with their level of needs. This suggests that someone’s SES can influence his behavior in everyday life.

Based on Maslow his theory, more studies were conducted that aimed at how SES could be measured and what relationships exist between people their SES and their behavior. Although many of these studies reached the similar conclusion(s), it is important to note that the way of measuring SES changed and developed over time. In 2000 Adler et al. (2000) published the theory that the measure of people their (objective) SES is determined by three factors: their financial abilities (economic resources), their education, and their job.

In the years after the publication of Maslow his theory several studies were published showing relationships between SES and factors like health (e.g. Winkleby et al., 1992; Adler et al. 1994), child development (McLoyd, 1998), intelligence and education (Sewel & Shaw, 1967; White, 1982), and obesity (McLaren, 2007).

(4)

4

The last couple of years the aim of research has changed from studying differences in indirect behavior as a result of someone’s SES to differences in direct behavior. There are two different theories that could intuitively explain what differences there could exist between upper class individuals and lower class individuals. One of these theories is that people from a lower class have less economic resources, more difficulty to be independent, and more trouble in finding a job (Oaks & Rossi, 2003) and therefore have a harder time to achieve their goals or to satisfy their needs as stated in Maslow’s hierarchy of needs. This could result in individuals thinking more about themselves than about others, because for these people you would expect it to be more important to get their own priorities straight before thinking of others. This could mean that lower class people will behave more selfish than upper class people who have already satisfied several of their basic needs.

However, another theory has been proposed that states the opposite (e.g. Piff et al., 2012; Piff et al. 2010; Kraus & Keltner, 2009): because people from the lower class have a harder time to satisfy their needs on their own, lower class individuals are more dependent on others than those from the upper class. To satisfy their own needs or to become more independent they first need help of others to accomplish this. Because of this, lower class individuals are more used to cooperating with each other and therefore they become more capable of communicating with others. As a result of this, lower class individuals communicate in a different way (Kraus & Keltner, 2009). Another factor that supports this theory is that because higher class individuals are more used to an “easier” life, it is more difficult for them to empathize with those who face problems they have never faced. Lower class individuals generally face more problems and therefore have an easier time empathizing with other, both upper class and lower class, individuals. This theory therefore suggests that lower class individuals show more prosocial behavior than upper class individuals. Many studies (e.g. Piff et al., 2012; Piff et al., 2010; Kraus & Keltner, 2009) have shown results that corresponds with this second theory. Piff et al. (2010) also show that lower class individuals are not only more inclined to empathize with others, but also show more trust behavior and compassion in stress situations towards others compared to upper class individuals.

On top of this Piff et al. (2010) show that people already behave different if they momentarily get manipulated in thinking they have a particular SES. People their perceptions of their SES got nudged towards a particular direction so that they would rank their own SES higher or lower on a socioeconomic ladder compared to the average SES. People who were nudged to rank themselves lower indeed showed more prosocial behavior than average and the people who were nudged to rank themselves higher showed less prosocial behavior. Those who

(5)

5

ranked themselves relatively low were willing to donate more money to charity than those who were nudged to rank themselves relatively high. This shows that when people their perception of their SES is only momentarily influenced, it already leads to differences in social behavior. If people their perception of their SES is so easily influenced, people can experience different perceptions of their SES during the week, or even during the day. However, when asking subjects about their SES, they become aware of their own SES, while in everyday life people are not consciously aware of the state of their SES. Therefore it is interesting to know if people are also influenced in their perception of their SES if only one of the three determinants (economic resources, education, occupation) changes relative to the same determinant of others. With this, no questions about someone’s SES have to be asked, and the comparison with the influence on everyday life would be more realistic.

Paul Piff has performed a study on this topic, which he talked about during TEDxMarin (Piff, 2013). In this study, he never actively influenced participants to be aware of their own SES. He only let them play a game of Monopoly, with one participant having a huge advantage over his opponent. Because of this, the advantaged players did much better during the game, which led to having a lot more economic resources as the game progressed. With economic resources being one of the three determinants of someone’s SES, these advantaged players perceived their SES much higher than the disadvantaged player. As the game progressed, the economic resources of the advantaged players grew, which resulted in their behavior

becoming less prosocial towards their opponent, corresponding to the existing theory. With this, Piff (2013) showed that even when subjects are not consciously thinking about their SES, they are still easily rigged in their perception of it by only influencing their relative economic resources, which, in turn, influences their social behavior.

However, with Monopoly, an advantage in economic resources gives the advantaged player the possibility to buy and own a lot of the resources (if not all) of the game, while the

disadvantaged player will be left with (almost) nothing. So having an advantage in economic resources with Monopoly could not only lead to a higher perception of your own SES, but could also lead to a relative gain in power. Keltner et al. (2003) define power as “An

individual’s relative capacity to modify others’ states by providing or withholding resources or administering punishment”. Also the feeling of having more power, with SES being a

subcategory of power, could lead to less prosocial behavior (Keltner et al., 2003). So it is still unclear if it was the relative increase in power that induced the antisocial behavior or the relative increase in economic resources. Therefore it remains interesting to study if people their social behavior will also change if only the relative economic resources are influenced

(6)

6

(primed), without effecting the relative power. So will people who only feel richer than others already show less prosocial behavior? Based on the before mentioned theories the following research question is stated:

“Does “only” earning relatively more (less) money in a short period of time already lead to people showing less (more) prosocial behavior?”

Another interesting phenomenon to study comes forward from the study of Kraus & Keltner (2009) that showed that lower class individuals are more inclined to empathize with others. Also, Piff et al. (2010) showed that lower class individuals show more compassion to those who need help than upper class individuals do. So, when only looking at the economic resources, it is interesting to study if people who have a disadvantage in earning money (economic resources), show more empathy towards others who also have this disadvantage than to those who have an advantage in earning money. This is also interesting to study for upper class individuals. However, because of the lack of compassion (compared to lower class individuals) and their trouble with empathizing, upper class individuals are not expected to make any distinction in their social behavior between socioeconomic classes. Therefore it is interesting to study if people who have an advantage in earning money indeed will not make a distinction in their social behavior between those who also have an advantage and those who have a disadvantage in earning money. This leads to the following two research questions:

“Do people who have a disadvantage in earning economic resources behave more prosocial towards others that also have a disadvantage in earning economic resources than to those who have an advantage in earning economic resources?”

“And do people who have an advantage in earning economic resources behave as prosocial to others that also have an advantage in earning economic resources as to those who have a disadvantage in earning economic resources?”

Most of the studies performed regarding the phenomenon of SES focused on the effects of it on people their psychological behavior (e.g, Maslow, 1943; Adler et al, 1994), on the indirect effects of it on people (e.g. Kraus & Keltner, 2009), or the effect on particular situations in their lives (e.g. Sewel & Shaw, 1967; White, 1982). Only a few studies, for example Piff et al. (2010) and Piff et al. (2012), studied it from an economic perspective. However, these studies

(7)

7

primed people their perception of their SES in a subjective way, making people think about their own SES while trying to manipulate it. To answer these research questions people will not need to consciously be aware of their own SES. They only have to be aware of one of the three determinants (economic resources) of their own and/or someone else’s SES.

Also when Piff et al. (2010) primed their subjects, they only asked hypothetical questions afterwards like “what would you do if…” instead of looking at their actual behavior and how this behavior was influenced by the priming. The only studies that focused on people their actual behavior as a result of their SES focused on people their objective SES (Piff et al. 2010) or focused on unethical behavior like taking candy that was supposed to be handed out to children (Piff et al., 2011). Therefore it is interesting to study these three research

questions, through which it will become clear if people are indeed influenced in their social behavior (towards others) by only influencing one of the determinants of their SES: their (relative) economic resources.

The afore mentioned research questions will be studied by performing an experiment. This experiment will be divided into four parts. The first part of the experiment consists of a version of the game “Rock, paper, scissors”. The game is luck based and the participants can earn tokens, which represent a value of 10 eurocents per token. The rock, paper, scissors game (from now the RPS game) used in this paper consists of five rounds which are all played against the same opponent. The RPS game is included in the experiment in order to study what kind of effect the amount of earned tokens has on the social behavior of the participants during the next parts of the experiment. The RPS game was designed in such a way that in each pair of participants one will always earn more tokens than the other in order to see if this absolute difference has any influence on their social behavior during the following parts of the experiment. Furthermore, during the RPS game, half of the participants will have an

advantage and half of the participants a disadvantage in the chance of earning tokens in each round of the game. Because the measure of someone’s SES is determined by their job,

education and economic resources, it is expected that people with a lower SES will face more difficulty in earning money than those with a higher SES. The (dis)advantage included in the game is implemented to mimic the afore mentioned effect as to research whether participants differentiate between the other players in the degree of shown prosocial behavior.

Parts 2, 3 and 4 of the experiment are designed to test the participants their social behavior after the RPS game. Parts 2 to 4 consist of different versions of the dictator game. In part 2, participants are asked to divide a set of 100 tokens between themselves and a charity of choice and in part 3 and 4 participants are asked to divide tokens between themselves and a

(8)

8

co-participant. In part 3 and 4 the participants are informed about the type of player (“A” or “B”, where players A have an advantage during part 1 and players B have a disadvantage during part 1) to whom they can give a chosen amount of their tokens. Parts 2, 3 and 4 are designed to test whether part 1 indeed had an influence on people their prosocial behavior. The fact that participants receive information in part 3 and 4 is added to test whether participants differentiate in to whom their prosocial behavior is directed.

The results of the experiment show that the participants who earned more during part 1 show less prosocial behavior towards others afterwards. Participants who earned less during part 1 sent on average 35.561 tokens to co-participants during part 3 and 4, while participants who earned more during part 1 sent on average 52 tokens during part 3 and 4. Furthermore participants with a disadvantage during part 1 sent on average 17.18 tokens more to

participants who also had a disadvantage during part 1 than they sent to participants that had an advantage. Moreover the results show a negative relation between the tokens earned during part 1 and the average amount of tokens sent during part 3 and 4. Each extra token that a participant earned in part 1 resulted in sending 0.22 tokens less during part 3 and 4. All these results are significant.

This thesis is divided into 8 different sections. In section 2 existing literature on the subject of SES and relevant factors is discussed. Section 2 explains how this thesis tries to add new insight on the topic discussed. In section 3 the experimental design will be discussed more elaborate including the reasoning for the chosen design. Furthermore the hypotheses are introduced and the data collection methods are explained. In section 4 the results of the experiment and the tests of the hypotheses are discussed. Section 5 discusses the results of the experiment, the interpretation of the results, and the limitations of the thesis. Section 6

consists of the explanation of the application of the results of this thesis and section 7 highlights several possibilities for future studies based on the research performed for this thesis. Last but not least, in section 8 the answer to the research question is given and conclusions related to the research are stated.

2. Literature

2.1 Measurement of SES

As stated in the introduction, measures of SES in relation to people their welfare or (social) behavior have been studied for years. However, the measurement of SES changed over time.

1_{This is the average amount of tokens sent in part 3 by all players (of the particular type) plus the average amount of tokens sent in part 4 by} all players (of the particular type) divided by two.

(9)

9

Maslow published his Hierarchy of needs in 1943, showing the less needs someone could satisfy, the lower the person was standing on this pyramid, which was one of the first “measures” of someone’s welfare and (social) status. In the years that passed, more measures of SES were proposed to measure factors that influence someone’s SES. For example Lucas (1977) and Thaler & Rosen (1976) made use of people’s census geocode characteristics (which explains the geographical location of the individual) in order to predict the characteristics of the person’s job, as well as the person’s health. Other studies show that it was not a person’s geocode that had an influence on his health, but that the race of this person had a dominant influence on his health (e.g. Wise et al., 1985, Gould and LeRoy, 1988). As a result of this, Geronimus et al. (1996) state it is a problem to use someone’s geographical location as a determinant for socioeconomic factors, because with this, sociodemographic factors (like race), will also have an effect. This automatically indicates that also race is not a good determinant for socioeconomic factors. They show in their paper that indeed there is no high correlation between someone’s SES and their geographical location.

While other researchers seemed to agree (e.g. Pickett & Pearl, 2001; Krieger et al., 1997), with Geronimus et al. (1996), it took a while until an alternative measure was presented. In their paper of 2000, Adler et al. (2000) presented a new measure of SES, which so far is considered as the best measurement of SES and currently it is the most applied measurement. As stated in the introduction, they show that the most important determinants of someone’s SES are their financial abilities, their education, and their job. The measurement that Adler et al. (2000) use is called the MacArthur Scale of (Subjective) Social Status which can be divided into a subjective and an objective SES measurement. For the subjective SES measurement participants were shown a drawing of a ladder with the following description:

“Think of this ladder as representing where people stand in our society. At the top of the ladder are the people who are best off, those who have the most money, most education, and the best jobs. At the bottom are the people who are the worst off, those who have the least money, least education, and the worst jobs or no job at all.” After reading this, participants

had to place themselves on one of the ten rungs of the ladder where they thought they belonged. For the objective SES, subjects were asked about their actual education, household income, and occupation, each divided in several categories. Since adolescents do not have a fulltime job or an income to be financially independent, Goodman et al. (2001) presented a variant of this measurement for adolescents. Instead of using only the one ladder used by Adler et al. (2000), one ladder is added which represents the people going to the same school as where the subject is enrolled. The subjects is told that on the top of the ladder the

(10)

10

schoolmates are placed that have the highest marks, the most respect and the highest standing and at the bottom logically the ones with the lowest marks, who are not respected and no one wants to hang around with. Through these measures the relation between specific behavior and both the subjective and the objective SES measurement can be studied.

2.2 SES in relation to psychological behavior

The aim of research into the influences of SES has changed from indirect effects like people their bad health and education as a result of a lower SES (Winkleby et al., 1992; Adler et al. 1994; Sewel & Shaw, 1967), to more direct effects like people their social behavior. For instance Kraus & Keltner (2009) studied the relationship between SES and nonverbal behavior towards others. Based on the theory that lower class individuals are more social in conversations they let two different individuals have a conversation for about 60 seconds. The subjects had to tell some personal things about themselves to the other subject. While they were talking to each other the researchers taped the conversation to analyze what kind of nonverbal behavior the participants showed. The results show that the higher class individuals wriggled more with nearby objects than the lower class individuals, were more likely to self-groom, and doodled more. The lower class individuals looked more at their partner, laughed more and were more likely to nod their heads or raise their eyebrows. These results demonstrate that subjects with a higher SES show their independence in the form of disengagement cues, whereas lower class individuals show their relative dependence in the form of social engagement cues, supporting the existing theory.

Anderson et al. (2006) did research to the fact if people indeed rank their SES position accurately by comparing what a group thinks about an individual’s SES to what the individual himself thinks of his own SES. Through this, they studied what effect it has if someone ranks himself higher or lower compared to how an individual is ranked by the group. They show that individuals who had an overly positive perception of their status in face-to-face groups were less liked by the group than individuals who had an accurate or lower perception of their status. Because of this, to be accepted by the group, individuals should perceive their own status within the group accurately. Anderson et al. (2006) show that this is the case, proving that (without priming) individuals have an accurate perception of their SES.

2.3 SES in relation to behavior in economic situations

In the last couple of years the relationship between SES and economic situations has been studied more frequently. Piff et al. (2010) show in their study different economic situations in

(11)

11

which the SES of someone is a determining factor for his behavior. In their paper they present their findings from several studies into different economic situations. One of these studies examines the effect of SES on playing the dictator game, where the participants individually were asked to divide an initial endowment of tokens or money between themselves and a matched receiver. The paper shows significant results that people with a relative low objective SES were willing to allocate a larger portion of their endowment to their partner in comparison with individuals with a relative high objective SES.

Another study from the same paper shows the effect of priming. Priming is a non-conscious form of human memory in which exposure to one stimulus has an influence on the response of another stimulus (Tulving and Schacter, 1990). Piff et al. (2010) try to manipulate the subjects’ relative perception of their own SES in a negative (or positive) way by asking the same question as with the MacArthur Scale of Subjective Social Status, with the only difference that they had to compare themselves with people standing at the very top (or bottom) of the social ladder. Through this, they primed the subjects in such a way that the participants who were asked to compare themselves with the top, placed themselves lower on the ladder than average (they had a lower perception of their own SES) and those who had to compare themselves with the bottom, placed themselves higher on the ladder than average (they had a higher perception of their own SES). After the priming, all the subjects were given the same (hypothetical) main question, asking how much of their salary should go to charity. The results show that people who have a relative high perception of their social rank, are willing to donate less than people who have a relative low perception of their social class. Also the difference between individuals with a low objective SES and individuals with a high objective SES was clearly visible by the fact that the lower class individuals were willing to donate more to charity than the upper class individuals. This study shows that if people momentarily get a relative high (low) perception of their SES, this immediately leads to less (more) prosocial behavior.

Paul Piff also talked at TEDxMarin (Piff, 2013) about another study, which is already briefly discussed during the introduction, he performed regarding the relation between SES and social behavior. For this study he did not ask the participants for a subjective ranking of their own SES in order to prime them, but he asked them to play a classic game of Monopoly against each other. However, the Monopoly game was manipulated in order to try to affect (prime) the subjects’ assessment of their own SES. One of the two subjects (the game was played with two subjects per game), was randomly chosen to get some crucial advantages over the other subject: the advantaged subject was allowed to roll two dices instead of one,

(12)

12

received three times as much money at the beginning of the game, and double the amount of money when he passed start. Thus, to make a link with the factors that determine someone’s SES, the advantaged player had more financial possibilities, more access to resources and moved around the board twice as fast. Because of this, the researchers found some major differences in social behavior between the two players. The subjects with the advantage showed more antisocial behavior as they became louder, made mean remarks towards the disadvantaged player, moved their game piece louder on the board, and even ate a larger portion of the pretzels that were supplied for both players.

2.4 SES and power

However, this study does not only show that people who perceive themselves as an upper class individual show other behavior than those who perceive themselves as lower class, but also suggests that the difference in power affects the differences in the subjects’ behavior. Keltner et al. (2003) define power as “An individual’s relative capacity to modify others’

states by providing or withholding resources or administering punishments”, and Kraus et al.

(2011) specify this by stating: “In essence, elevated power allows individuals to have elevated

control over others’ outcomes and increased freedom to make decisions according to their own goals and motivations.”

Because the advantaged player has more money, and moves around the board a lot faster, he has the opportunity to purchase a large portion of properties and develop those properties with houses and hotels. The more houses a player owns, the larger the punishment for the other player when he unluckily lands on one of those properties, which indicates that the advantaged player has several possibilities to punish the disadvantage player. The advantages in money and moving around also give a player the possibility to withhold resources from the other player. The available resources in a game of Monopoly are money, properties and real estate in the form of houses and hotels. But with the advantaged player having the possibility to buy a lot more properties, houses, and hotels before giving the disadvantaged player a chance of buying them, it gives the advantaged player a way to withhold these resources. Therefore it could be that through this manipulation of the Monopoly game, not only the subjects’ perception of his SES has been influenced, but also a difference in power was created between the subjects which grew ever larger as the game progressed. This power aspect could have had an effect on the results of the study.

In fact, there are various studies showing a relation between someone’s (assessment of their) power and his behavior (e.g. Brass & Burkhardt, 1993; French, 1956, Côte et al., 2011). Some

(13)

13

of these studies show similar outcomes as the studies into SES. For example, the results Kraus and Keltner (2009) noted about the relation between SES and nonverbal behavior were also noted by LaFrance and Banaji (1992) about the relation between power and nonverbal behavior, showing that females (with the assumption that they are less powerful than men) show more face expressions. The study Snodgrass et al. (1998) show similar results, whereas the less powerful also show more face expressions. Also the study to the relation between power and behavior of Anderson & Berdahl (2002) shows results comparable with the results of the study of Piff (2013). In this study is shown that the participants with high power were, as well as the participants from Piff (2013) with a relative high perception of their SES, more likely to show particular social behavior: they did not bother to express their true attitudes (while participants with low power were more inhibited in expressing this), were more focused on earning rewards (while low power participants were focused on evading threats), and were more likely to express their positive emotions.

Although SES and power are two different phenomena, it is not a surprise that studies to these two phenomena are similar and show comparable results. There is a lot of overlap between the two, because of the fact that status, together with authority and dominance, is one of the main determinants of someone’s power (Keltner et al., 2003). This, however, does not necessarily mean that a high status gives power in any situation (e.g., a bank director still has to wait in line for the checkout at the supermarket) or that power gives a high status (e.g., police officers in regions where they are not respected).

Kipnis (1972) states that when someone’s power increases the temptation to influence other people their behavior grows. As a results of this, the assumption arises that the behavior of others is controlled by the power holder, so not self-controlled, and therefore the power holder creates the assumption that when subordinates show good performance, this is attributed to the power holder’s own control. As a result, the power holder increases psychological distance from the less powerful. This may explain the derogatory remarks during the Monopoly game of Piff (2013) made by the advantaged player towards the disadvantaged player, to increase psychological distance. This suggests that the advantaged player indeed felt a difference in power.

Thus, in the study of Piff (2013), the only study to SES where people were primed with actual behavior and without naming SES, it is not clear which factor had a dominant influence on the social behavior of the participants. It could have been the increase in power, or the increase in the perception of their SES, or maybe both.

(14)

14

Also, the study only shows that the advantaged players show more antisocial behavior, but it does not show that the ones with the disadvantages show more prosocial behavior. So it is interesting to know if people also show less (more) prosocial behavior, when only their relative earnings increase (decrease), without gaining (losing) any power.

3. Experimental Design

First of all, the experiment was divided in three different sessions on three different days, with each session having 12 unique participants. Each session was exactly the same, except for one session where the order of the different versions of the dictator game2 was changed. The experiment was totally anonymous, meaning that the participants did not know with who they were playing each part. All pairings of participants were randomly assigned. For part 1 this was done in advance by the experimenter, but for the other parts the random matching was only performed when the particular part was selected to be paid out. The participants had to draw a card, which was downward faced, in order to know which seat number (and with this, the participation number) was assigned to them. The participants were not allowed to talk to each other. When a participant had a question he first had to raise his hand so that the experimenter could come to him and answer the question in private.

The experimental design consisted of four different parts, with part 2, 3 and 4 being different versions of the same game. The Rock-Paper-Scissors game (from now on the RPS game) was played in the first part, during the second part a version of the dictator game was played where participants had to divide tokens between themselves and a charity of choice (five different options), and the third and fourth part were versions of the dictator game where they could divide tokens between themselves and co-participants. The participants were randomly divided into two groups. One group that had an advantage (group A) and one group that had a disadvantage (group B) in part one, during the RPS game.

3.1 The Rock Paper Scissors game

The RPS game is a game in which two participants play against each other. They both have three options to choose from: “rock”, “paper”, or “scissors”. Their payoffs depend on the choice they make and the choice their opponent makes. If you choose rock, you win from scissors, scissors wins from paper, and paper wins from rock.

This RPS game consisted of five rounds that were played with the same matched pairs. A participants of group A (player A) always played against a randomly chosen participant of

(15)

15

group B (player B), and hence a player B always played against a randomly chosen player A. When a player chose the winning option, he received 50 tokens, with each token being worth 10 eurocents if they were eventually paid out. As said, the players A were given an advantage. The advantage they had was that when both players chose the same option (so a tie), player A would win. The other way around players B were given a disadvantage as they would lose in the case of a tie. This means that, on average, a player A would win two out of three times and a player B only one out of three. This gave an expected value for part 1 for a player A of 166⅔ tokens3_{, which is twice as large as the expected value of 83⅓ tokens}4

for a player B. The payoffs for the RPS game looked as follows:

RPS Game Player B

Rock Paper Scissors

Player A Rock 50,0 0, 50 50, 0

Paper 50, 0 50, 0 0, 50

Scissors 0, 50 50, 0 50, 0

For part 1, six subjects participated as player A, and six participated as player B. Each participant were given the same specific instructions (see Appendix C) for this part. They were told that at the end of their specific instructions was stated to which group (A or B) they were assigned. Participants had two minutes to make their decision and after each round they received feedback on the choice of their opponent for that round and how many tokens they earned as a result of this. Also feedback on the cumulative earnings were given after each round.

When all five rounds were finished the participants were shown their total earnings on their last feedback form and were informed that they should keep this form to know how much they could earn if part 1 was chosen to be paid out. After this, they were informed that part 1 of the experiment was finished and part 2 would begin.

Part 1 of the experiment was designed in order to let the subjects earn an amount of tokens (money), to study what effect this would have on the parts that were played afterwards. Because of the distribution in payoffs, for each pair one of the participants earned in total more (less) than the other, and the participant were aware of this. This part was also designed to give half of the participants a (dis)advantage in earning economic resources in order to test if this would have an effect on their social behavior in the parts that were played afterwards.

3_{Expected value = ⅔ (chance to win) * 50 (tokens earned for the winner) * 5 (number of rounds) = 166⅔.} 4

(16)

16

3.2 The different versions of the Dictator game

For the second part of the experiment each participant had to play a version of the one shot dictator game where they were given the option to divide 100 tokens between themselves and a charity of choice (there were five choices of general charities). They could choose to send none, a part of – or all of their tokens (in tenth) to the charity of their choice.

This part of the experiment was designed to test whether participants who earned more (less) money, indeed show less (more) prosocial behavior than those who earned less (more) than their matched participant in part 1. The findings of this part of the study should provide the answer to the question:

“Does “only” earning relatively more (less) money in a short period of time already lead to people showing less (more) prosocial behavior?”

The third part was another version of the one shot dictator game. Each participant was told that everyone would play as a sender and would be randomly matched with a receiver from a particular group (A or B, which was the same for everyone in the particular part). The fourth part was designed exactly the same as the third part with as only difference the group from which the receivers were picked. So each participant played as a sender (and had a chance to get paid out as a sender), but the participants that were the player type from the chosen group of receivers also had a chance to get paid out as a receiver. Figure 2 gives more insight in this. So in case the senders were matched in part 3 to receivers from group A, both the players A and B played as -, and could be paid out as a sender in this part, while the players A also could be paid out as a receiver (and in part 4 this would be the other way round). However, they could only get paid out as one of these roles and the participants were aware of that. This was decided by a dice, but will be explained in more detail at the end of this section. Also, the participants were told that the receiver they played with, was not the sender to them and not the same participant they were matched with during part 1. On top of this, they were (again) told that the game was totally anonymous so that no one would discover what choices they made during these parts.

In both part 3 as part 4 the participants received information about their matched receiver. The only information that was given, was what type of player (A or B) their matched receiver was, and it was made sure that they understood that this participant also was this type of player during all the other parts.

(17)

17 Figure 2

So in two sessions part 3 was played with the matched receiver being a player B for every sender and in one of the sessions part 3 was played with the matched receiver being a player A for every sender. When the matched receiver was a player B in part 3, the same game would be played in part 4 with the matched receiver being a player A (session 1 and 3) and when the matched receiver was a player A in part 3 the same game would be played in part 4 with the matched receiver being a player B (session 2). These different orders in parts were implemented to avoid an order effect.

As stated, all the participants were given an initial endowment of 100 tokens, with each token being worth 10 eurocents. When they had seen the information what type of player their matched receiver was, they had to make a choice on how many, if any, tokens (a tenth number) they wanted to send to their matched receiver. No one was given any feedback after this.

(18)

18

The participants were made clear that the matching would again be totally random. When part 3 was finished, as told, the participants did not get any feedback, and part 4 started. When all the parts were finished, the participants were given a small questionnaire to fill in. While the participants were filling in the questionnaire, the experimenter decided with a dice which part was going to be paid out. If a one was rolled, part 1 would have been paid out, if a two was rolled, part 2 would have been paid out, if a three was rolled part 3 would have been paid out, if a four was rolled part 4 would have been paid out, and if a five or a six was rolled, the dice would have been rolled again until a one to four was rolled.

Also, after part 1, 3 and 4 the participants were given control questions to fill in. For almost all of the participants it was the first time they participated in this kind of an experiment. Therefore for part 1 control questions were asked to make sure the participants fully understood the game and the rules of payment. For part 3 and 4 it was important that it was clear to the participants that they could not be matched with themselves, not with the same participants as they were matched to in part 1, and not with a receiver of another type than stated in the instructions for that part. Because this could be difficult for some participants, control questions were asked to make sure that everyone understood this.

If part 1 would have been selected for payment, one pair of players would have randomly been chosen to receive their payoffs in euros equal to the accumulated amount of tokens earned after the fifth round, with each token being worth 10 eurocents. Which pair, would have been decided by drawing one ticket from all the participants’ participation numbers. The participant corresponding to the participation number of the drawn ticket and his matched opponent would then each have been paid out the accumulated amount of tokens individually earned during part 1 in euros. In none of the sessions part 1 was selected by the dice for payment.

If part 2 would have been selected for payment, two players would have randomly been chosen to receive their payoffs in euros. All the participants’ participation numbers would have been put in a hat and from that hat, two tickets would have been drawn. The two participants corresponding to the participation numbers of the two drawn tickets would then each have been paid out the amount of tokens, with each token being worth 10 eurocents, they decided to keep (100 tokens – the amount sent to charity). The amount they decided to donate to charity would then have been donated to charity by the experimenter. To make it credible, the chosen participants would have been sent an e-mail afterwards proving that the amount was truly paid to the chosen charity. In none of the sessions part 2 was selected by the dice for payment.

(19)

19

If part 3 or 4 would have been selected for payment, one pair of players would have been chosen to receive their payoffs in euros. All the participants’ participation numbers would have been put in a hat and from that hat, one ticket would have been drawn to choose the sender. The participant corresponding to the participation number of the drawn ticket would then have been paid out the amount of tokens he decided to keep (100 tokens – the amount sent to the matched receiver), with each token being worth 10 eurocents. Next, a receiver would have been chosen between all the players of the player type that was applicable to this part. This could not have been the same person that had already been chosen for payment not his matched participant from part 1 and not another player type than applicable for the particular part. The participation numbers of all the remaining players of the player type applicable for the particular part would have been put in a hat and from that hat, one ticket would have been drawn to choose the matched receiver. The participant corresponding to the participation number of the drawn ticket would then have been paid out the amount of tokens, with each token being worth 10 eurocents, the chosen sender had sent. Part 3 was selected by the dice in all three sessions to be paid out, thus during none of the sessions part 4 was selected.

The participants were told that when it was decided who would get paid out, everyone would receive an envelope. Only the envelopes of the participants that were chosen to receive their payoffs in euros would contain money and all the other envelopes would remain empty. This was done to make sure that everyone’s decision remained anonymous and therefore the participants were also told to not tell anyone the content of their envelope.

Note that in every case two participants would have been chosen to get paid out and each participant was aware of this. This was done because of the fact that participants their choices could be influenced if the different parts of the experiment would lead to a different number of people that could get paid out. For part 1 these two players would have been of a different type, while for part 2, 3 and 4 it would have been possible that two players of the same type would have been paid out. Also for part 1, 2, and 4 it would have been a pair of players that would get paid out, while for part 2 this would have been totally random. This was explained to all participant for each specific part.

3.3 Hypothesis

As already explained in the previous section, this experiment is designed to test whether people already show more (less) prosocial behavior as a result of their earnings in playing a game based on luck, and where half of the participants get a clear advantage with respect to

(20)

20

the other half. Because of this advantage, the advantaged participants will have a higher chance to win and with this to earn more money. The expected cumulative payoff for part 1 for someone of group A is 166.67 tokens, while the expected cumulative payoff for someone in group B is 83.33 tokens.

As stated by Adler et al. (2000) someone’s economic resources is one of the three main determinants of their SES. With the RPS game, participants from group B should on average win twice as less economic resources than participants from group A. The fact that the players B have this disadvantage could affect their behavior in such a way that they are more inclined to empathize, or feel compassion towards others. And if they are indeed affected in this way, they will send more money to charity during the version of the dictator game than the players who have an advantage. This leads to the first hypothesis:

Hypothesis 1a: Participants who receive an ex-ante economic advantage (disadvantage) in

chances with respect to their opponent during part 1 of the experiment, will show less (more) prosocial behavior afterwards in the form of donating money to charity.

However, the advantage the players have during this game will not necessarily lead to higher earnings. With some luck, some of the disadvantaged players could earn more than their matched and advantaged participants and therefore not have the feeling to be disadvantaged at all. Because of this, it is interesting to study the difference between the behavior of players who earned more and those who earned less during part 1. As stated, when people earn more, it could already lead to showing less prosocial behavior, which leads to the next hypothesis:

Hypothesis 1b: Participants who earn more (less) tokens during part 1 of the experiment

compared to their opponent, will show less (more) prosocial behavior afterwards in the form of donating money to charity.

It is also interesting to look at whether participants with a(n) (dis)advantage show less (more) prosocial behavior towards co-participants of the experiment. Therefore in part 3 and 4, participants played a version of the dictator game in which they were able to divide 100 tokens between themselves and another player. The expectations are that the advantaged players will show less prosocial behavior towards their co-participants than the disadvantaged players and that the participants who earned more during part 1 will show less prosocial behavior towards their co-participants than the players who earned less during part 1. Based on the afore stated expectations the following hypotheses 1c and 1d are formulated:

Hypothesis 1c: Participants who receive an ex-ante economic advantage (disadvantage) in

chances with respect to their opponent during part 1 of the experiment, will show less (more) prosocial behavior afterwards in the form of sending money to co-participants.

(21)

21

Hypothesis 1d: Participants who earn more (less) tokens during part 1 of the experiment

compared to their opponent, will show less (more) prosocial behavior afterwards in the form of sending money to co-participants.

These hypotheses will be tested during the entire experiment. In part 1 the participants will be primed in having a(n) (dis)advantage and with this earn less or more money. The effect of the afore mentioned on their social behavior will be examined by studying the amount of money they are willing to donate to charity in part 2 and the amount of money they are willing to send to another player in part 3 and 4.

According to Piff et al. (2010) and Kraus & Keltner (2009) lower class individuals have an easier time empathizing with others than higher class individuals. Piff et al. (2010) also show that lower class individuals show more compassion towards people that need help than upper class individuals. This implies that disadvantaged individuals are more inclined to show prosocial behavior towards other disadvantaged players than to advantaged players. Because the advantaged players are less inclined to empathize with others and are less likely to show compassion to those in need, an advantaged players is expected to not show any difference in social behavior towards another advantaged player or towards a disadvantaged player. In the third and fourth part the participants can divide their tokens between themselves and a matched opponent. In both parts, the only information they receive about their matched opponent is whether he is part of the advantaged group or of the disadvantaged group. The second hypothesis is formulated based on the afore mentioned:

Hypothesis 2: The participants with an ex-ante economic disadvantage during the first part

of the experiment show more prosocial behavior towards others with the same disadvantage than to those with an ex-ante economic advantage during the first part, while participants with an ex-ante advantage during the first part do not differentiate in their behavior towards others with the same advantage and those with the afore mentioned disadvantage.

This hypothesis will be tested during part 3 and 4. In each part the participants have to choose how much tokens they want to send to a particular player type. This way can be tested whether there is a significant difference in tokens sent per type of player between the two different types of players.

(22)

22

3.4 Data collection

The experiment was conducted in the living room of the experimenter. The living room was arranged in such a way that it largely corresponded to the CREED lab, used by the University of Amsterdam. For the experiment 36 participants were selected with approximately the same objective SES. The participants consisted of friends and family. Because of this, most of the participants were familiar with each other. This could have led to more prosocial behavior during the experiment than would have occurred in a group of unfamiliar people. However, this should not influence the results as the hypotheses is aimed at identifying differences in prosocial behavior and not at identifying prosocial behavior in general. The experiment was, because of organizational reasons, conducted on three separate occasions with twelve participants per experiment (six players B and six players A). In total 36 unique participants participated in the experiment. The participants were incentivized by real money. Each session two of the twelve participants were chosen to receive a payment based on the outcomes of one of the played parts, which gives a corresponding chance of 16.67% that they would actually be paid.

The instructions were in Dutch in order to make sure that the participants understood the experiment. As most participants had never participated in an experiment before plenty of time was reserved for the instructions to make sure that everybody would understand what was expected of them. The participants were randomly matched in advance by the experimenter for part 1, based on the participation numbers. For part 3 and 4, there would only exist matches when these were selected for payment. The participants were never informed about who their matched participant was, most importantly because of the fact that most of them were friends or at least acquaintances, so that this could have had effect on their behavior.

Furthermore, during the experiment the participants could earn tokens instead of real money. Each token represented a value of ten eurocents which was clearly communicated to the participants. All the instructions, forms and questionnaires that were handed out to the participants are included in Dutch in Appendix C.

4. Results

4.1 Describable results

As stated, 36 people participated in the experiment divided over three different sessions. 24 of them are male (66.67 percent) and 12 of them female (33.33 percent). Most of them (26) are 28 years or younger and the rest (10) are 48 or older. The average age of the participant is 31

(23)

23

years. 66.67 percent of them are currently undergraduate students either in HBO or WO. The other 33.33 percent is already graduated and has a job. 92 percent of the participants was already familiar with the game of rock paper scissors, and 25 percent was familiar with (a version of) the dictator game.

Because of the small sample size (N=36), the results will be tested with non-parametric test. Comparisons between players for parts 2,3 and 4 are conducted with independent samples Mann-Whitney U tests, while comparisons of the same player between different treatments are conducted with Wilcoxon Signed Rank Tests.

To control for an order effect, an independent samples Mann-Whitney U test is conducted between the sessions where the part with a player B as matched receiver was played as third part and the session with a player B as matched receiver was played as fourth part. For players B (p-value: 0.18) as for player A (p-value: 0.44) no order effect is found for tokens sent in part 3 and 4. Also for players B (p-value: 0.31) there is no order effect found for tokens sent to other players B in both parts, as for players A (p-value: 0.28) no order effect is found for tokens sent to other players A in both parts.

4.2 Main results

The first main results are provided in the figures below. These figures give a clear insight in how part 1 had an influence on the participants their behavior during the parts 2 through 4. First in figure 3 the difference between the average amount of tokens sent during parts 2 to 4 by players A and B is represented. The first column shows the average amount of tokens sent per type of player in part 2. The second column shows the average of the cumulative amount of tokens sent during part 3 and 4 per type of player5. There is made no distinction between part 3 and 4, because these parts differed during the different sessions. The third column shows the average amount of tokens sent to players A per type of player and the fourth column shows the average amount of tokens sent to players B per type of player.

5_{This is the average amount of tokens sent in part 3 by all players (of the particular type) plus the average amount of tokens sent in part 4 by}

(24)

24 0 10 20 30 40 50 60 70 80

Part 2 Avg Part 3, 4 Player A Player B

Av er ag e am ou nt o f t ok en s s en t

Part in which tokens were sent / Player type where tokens were sent to Difference in average amount of tokens sent between those who

earned more and those who earned less than their opponent during part 1

Earned More Earned Less

Figure 3

This figure shows that players B sent on average more tokens during part 3 and 4, more to players A and more to players B compared to players A. Only during part 2 players A sent more tokens than players B. However, the only difference that is significant is the difference in the amount tokens sent to players B. These statistics will be discussed in more detail later in this section.

As stated, being a player B, so having a disadvantage does not necessarily mean that you earn less. Therefore figure 4 shows the average amount of tokens sent during the same situations as in figure 3, only than for those who earned less (100 tokens or less) and those who earned more (150 tokens or more) than their opponent in part 1.

Figure 4 0 10 20 30 40 50 60 70 80

Part 2 Avg Part 3,4 Player A Player B

Av er ag e am ou nt o f t ok en s s en t

Part in which tokens were sent / Player type where tokens were sent to Difference in average amount of tokens sent between player types

Player A Player B

(25)

25 0 20 40 60 80 100 120 140 160 180 0 50 100 150 200 250 (C um ul at iv e) a ver ag e am ou nt o f t ok en s s en t du rin g pa rt 3 a nd 4

Cumulative amount of tokens earned during part 1

Amount of tokens sent during part 3 and 4 dependent on

cumulative earnings in part 1

Part 4 Part 3

This figure shows that the participants who earned less in part 1 sent on average more tokens during part 2, part 3 and 4, to players A, and to players B. So in each situation those who earned less during part 1 seem to be more generous than those who earned more during part 1. Again not all results are significant. Only the difference in the average of tokens sent in part 3 and 4 and the difference in tokens sent to players B are significant. Again the details of these statistics will be discussed later in this section.

Furthermore it is interesting to study if it the cumulative number of tokens earned in part 1 has any influence on the number of tokens sent during the other parts. Figure 5 shows the relation between the total number of tokens earned in part 1 and the average amount of tokens sent during 3 plus the average amount of tokens sent during part 4. The first column shows the average amount of tokens sent during part 3 plus the average of amount tokens sent during part 4 by the participants who earned 0 tokens in total during part 1. The second column shows the average amount of tokens sent during part 3 plus the average amount of tokens sent during part 4 by the participants who earned 50 tokens in total during part 1, and so on.

Figure 5

There were a total of two participants who earned 0 tokens during part 1 (so their two matched opponents earned 250), four participants that earned 50 tokens (so their four matched opponents earned 200), and twelve participants that earned 100 tokens (so their twelve matched opponents earned 150 tokens). Figure 5 clearly shows that on average the more a participant earned during part 1, the less the participant was willing to send to co-participants. As example, the two players that earned 0 tokens during the first part, sent on average a

(26)

26

cumulative amount of 170 tokens during part 3 and 4 to their co-participants. As opposed to these players, the participants who earned 250 tokens during part 1 sent on average only a cumulative amount of 45 tokens during part 3 and 4 to their co-participants. The negative relation between the average amount sent during part 3 and 4 and tokens earned during part 1 is significant and will be discussed in more detail later in this section.

4.2.1 Hypothesis test

To test if these findings are significant or not, and to answer each hypothesis, all the hypotheses will be tested with the use of SPSS.

Hypothesis 1a: Participants who receive an ex-ante economic advantage (disadvantage) in

chances with respect to their opponent during part 1 of the experiment, will show less (more) prosocial behavior afterwards in the form of donating money to charity.

To test this hypothesis a distinction is made between players A (who had an advantage during part 1) and players B (who had a disadvantage during part 1), to see if there is a difference in average tokens sent to charity during part 2. A one-sided Independent Sample Mann-Whitney U test was conducted to test for a significant difference. As already seen in figure 3 there was almost no difference between players A and players B in tokens sent. Players A sent on average 73.33 tokens to charity and players B sent on average 63.33 tokens to charity. So it seems already clear that this hypothesis is not supported. With a p-value of 0.7595 the null hypothesis is retained, so it cannot be concluded that the disadvantaged players donate more tokens to charity than the advantaged players.

The reason why no effect is found can be explained by the fact that some of the participants explained after the experiment that because it was “free money” they did not have any reason to keep some of it, so most of the participants sent all of their tokens to charity. Also the fact that the participants are friends and family of the experimenter made it more difficult for them to accept money from him, so they would rather donate it to charity, whether they had an advantage or not. Because of this, because of the small N, or because of another (non-observed) reason, no significant difference was found.

Hypothesis 1b: Participants who earn more (less) tokens during part 1 of the experiment

compared to their opponent, will show less (more) prosocial behavior afterwards in the form of donating money to charity.

To test this hypothesis a distinction is made between the participants who earned less during part 1 than their matched opponent (100 tokens or less) and those who earned more than their matched opponent (150 or more). Figure 4 shows that the difference is very small, with the

(27)

27

players who earned more during part 1 donating 66.1 tokens on average and the players who earned less donating 70.6 on average. A one-sided Independent Sample Mann-Whitney U test was conducted and with a p-value of 0.384 there was no significant difference found. The null hypothesis is retained, so it cannot be concluded that participants who earn less during part 1 donate more to charity afterwards than those who earn more.

The same explanation for this may apply as for the fact that there was no significant difference found for hypothesis 1.a.

Hypothesis 1c: Participants who receive an ex-ante economic advantage (disadvantage) in

chances with respect to their opponent during part 1 of the experiment, will show less (more) prosocial behavior afterwards in the form of sending money to co-participants.

For hypothesis 1c the difference in tokens6 sent between the players A and players B will be tested for part 3 and 4. For each player the average is taken of the accumulated amount they decided to send during part 3 and 4. Players A sent on average 34.72 tokens during parts 3 and 4, while players B sent on average 52.78 tokens during parts 3 and 4. A one-sided Independent Sample Mann-Whitney U test was conducted to test if this difference is significant. With a p-value of 0.085 this difference is not significant.

Thus, based on this one-sided Independent Sample Mann-Whitney U test, it cannot be concluded that with a significance level of 5%, that players with an economic disadvantage in part 1 (players B) send more tokens to co-participants during part 3 and 4 than the players with an economic advantage during part 1 (players A). So it cannot be said that participants with an ex-ante economic disadvantage are more generous than participants with an ex-ante economic advantage. This is against the expectations. One of the reasons why the hypothesis is not supported could be the small N. Maybe with a larger N a significant difference would have been found. Also the fact that some players B, four in total, earned more during part 1 than their opponent could have had an influence. Because these players B earned more than their opponent, they may not have experienced the disadvantage as a true disadvantage, because it did not affect them negatively in their earnings compared to their advantaged opponent.

Hypothesis 1d: Participants who earn more (less) tokens during part 1 of the experiment

compared to their opponent, will show less (more) prosocial behavior afterwards in the form of sending money to co-participants.

(28)

28

For hypothesis 1d the difference in tokens sent between those who earned more in part 1 and those who earned less in part 1 will be tested for the average of the accumulated amount sent in part 3 and 4. Again, a one-sided Independent Sample Mann-Whitney U test was conducted. The participants who earned more than their matched opponent sent on average 35.56 tokens during part 3 and 4, while the participants who earned less than their matched opponent sent on average 51.95 tokens during part 3 and 4. With a p-value of 0.034 this is a significant difference. So with a significance level of 5%, it can be concluded that participants who earned more during part 1, send less tokens to other participants during part 3 and 4 than the participants who earned more during part 1, which is according to the hypothesis.

This shows that people are already influenced in their social behavior towards others, with only having more (less) economic resources.

Hypothesis 2: The participants with an ex-ante economic disadvantage during the first part

of the experiment show more prosocial behavior towards others with the same disadvantage than to those with an ex-ante economic advantage during the first part, while participants with an ex-ante advantage during the first part do not differentiate in their behavior towards others with the same advantage and those with the afore mentioned disadvantage.

Since this is a measurement between treatments for the same participants, a Wilcoxon Signed Ranks test was conducted to test the hypothesis. Players A sent on average 42.78 tokens to other players A and 26.67 tokens to players B, which is on average 16.11 tokens more to players A. With a p-value of 0.134 it cannot be concluded that this difference is significant. For players B a one-sided Wilcoxon Signed Ranks test was conducted. Players B sent on average 61.67 tokens to other players B and 43.89 tokens to players A, which is on average 17.78 tokens more to players B. With a p-value of 0.0015 this difference is significant. So with a significance level of 5%, it can be concluded that participants with an ex-ante economic disadvantage during part 1 (the players B) send more tokens to other participants with an ante economic disadvantage (other players B) than to the participants with an ex-ante economic advantage during part 1 (players A).

Figure 5 shows a negative relation between the amount of tokens earned in part 1 and the average of the cumulative amount of tokens sent during part 3 and 4, so it is interesting to do a regression to test if this negative relationship is significant. Although the difference in average amount of tokens sent during part 3 and 4 between the different player types is not significant, it does seem to have some effect on the average amount of tokens sent during part 3 and 4. So in order to study which effect (mainly) drives prosocial behavior it is necessary to

(29)

29

also analyze the effect of player type. To test this, a multiple linear regression model is used of the following form:

γi = α + β1Χ1i + … + βkΧki + εi , where i = 1 ,… , n

For this multiple regression the average of the cumulative amount of tokens sent in part 3 and 47 will function as dependent variable and the “cumulative amount of tokens earned in part 1” and the “player type” as independent variables. However, it is also possible that the variables influence each other. The amount of tokens earned could have a different effect on a players with an advantage than on those with a disadvantage. To test whether this interaction effect is significant, a two-way ANOVA test was conducted. The results for this test (F = 0.215, df = 1, p = 0.647) show that the interaction effect is not significant, so for the regression the interaction term will not be included. Therefore the following model was used:

γi = α + β1Χ1i + β2Χ2i + εi, where i = 1 ,… , 36

Since the maximum cumulative amount of tokens that was earned during part 1 was 250 (so the minimum was 0) and participants could only win per 50 tokens, the distribution of X1 = ϵ{50, 100, 150, 200, 250}. The variable “Player Type” is a dummy variable that represents 1 if the player type is B and 0 if the player type is A so X2 = {0, 1}. The results of this

regressions are shown in table 1. Table 1

Variable Coefficient Std. Error P-Value N = 36

R² = 0.218 F = 4.6 (p-value: 0.017) Constant 70.735 17.04 0.000 Cumulative Payoff Part 1 -0.224 0.098 0.029 Player Type 1.912 11.449 0.868

This table shows a negative relation between tokens earned during part 1 and average tokens sent during part 3 and 4. These results show that each extra token earned in part 1, will result in 0.224 tokens sent less on average during part 3 and 4 (while keeping the other variables constant). With a p-value of 0.029 (t = -2.278) this is significant. If for example someone earned 200 tokens during part 1, he will send on average 44.8 tokens less during part 3 ánd 4 than a participant of the same type that earned 0 tokens during part 1. The variable Player Type, with a p-value of 0.868 (t = 0.167), does not have a significant effect on the average

Does everyone have a piece of Scrooge McDuck hiding in them? : the influence of money on social economic behavior : a study focused on how people their social economic behavior may be influenced by their relative amount