Comparing generosity between direct and indirect donations in a dictator game

(1)

COMPARING GENEROSITY BETWEEN DIRECT AND

INDIRECT DONATIONS IN A DICTATOR GAME

Student: BSc Esmée Stoutenburg Student ID: 10358641

Supervisor: A. van der Veen Deadline: 15-01-2018 ECTS 15

(2)

1

Abstract

This paper examines the difference in the level of generosity between donations to a homeless person (direct donations) and to an organization helping homeless people (indirect donation). An experiment was conducted using two groups. The participants of the first group earned money by performing a real-effort task. The participants in the second group received an endowment without doing anything. Both groups played a dictator game twice: first with the option to offer money to a homeless person, second with the option to offer money to an organization. Generosity was measured via the amount donated by the dictator, the level of direct and indirect donations was compared to each other in order to decide in which dictator game the dictator was more generous. Generosity was significantly greater for indirect than direct donations in the group that received the endowment from the experimenter. No significant difference in generosity was found for the group that earned their money.

Statement of Originality

This document is written by Student Esmee Stoutenburg who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

2

1. Introduction

According to the ‘Centraal Bureau Fondsenwerving’ (2017), the amount of money donated to charities in 2015 for the Netherlands was around 5.7 billion euros, of which 4.4 billion was in private donations by households and individuals. A lot of money is spent on charity, therefore most previous research is done investigating the motivation of individuals to donate money to charity. However, this amount of donations does not contain direct donations to individuals, such as homeless people. Unfortunately, the difference in the level of generosity for donating directly to a homeless person compared to donating money via an organization that helps homeless people has not been investigated. This information is useful for organizations that are helping the homeless. For example, if donors are more generous with regard to direct donations, the organizations can increase donations by connecting the donors to the recipients. This leads to the following research question: Do people become more generous when donating to an organization than to a homeless person?

This paper uses an experiment to study the differences in the level of generosity for direct or indirect donations. The donations in the experiment were actually transferred to the chosen organization or to a homeless person. Due to the experimental design used in this paper, it is also possible to look at differences in the level of donations for various income levels. This is an addition of this paper since previous literature disagreed about the effect of the income level on the level of generosity. The study of McClelland and Brooks (2004) stated that people with a higher income level are more generous when it comes to donating money. The reason for this generosity is that it is easier for wealthy people to earn back the donated money. On the other hand, Côte, House and Willer (2015) found that people with a higher income become less generous in a situation where income inequality is high. They stated that wealthy individuals want to preserve their privileged position. People with a higher income feel entitled to their wealth and their privileged position.

The main result found in this paper is a significant difference between donations to a homeless person and to an organization. Overall, the participants donate more to an organization than to a homeless person. Testing separately for different groups of participants, a significant difference was found for the participants receiving the endowment from the experimenter. For the participants who earned their money by doing a task, no significant difference was found.

(6)

5 The rest of the paper is structured as follows. Chapter 2 contains a literature review discussing the expected difference between direct and indirect donations, the effect of income on the level of generosity and a description of the dictator game. Chapter 3 describes the experimental design. Chapter 4 presents and explains the results. Chapter 5 contains a discussion and conclusion of the research. An appendix with additional information is added after the conclusion.

(7)

6

2. Literature Review

This section gives an overview of all the relevant literature. First, trust is described as a possible explanation for the expected difference in the level of generosity between directly and indirectly donated money. Second, the effect of income on generosity is reviewed. Third, an overview of the dictator game is given. This includes information about the external validity of a dictator game, which influenced the experimental design of this research. The literature review is concluded with a summary and the contribution to previous literature of this paper.

2.1. The impact of trust on generosity

For this paper, a difference in the level of generosity is expected for direct donation to a homeless person compared to indirect donations via an organization that helps homeless people. Bekkers (2003) stated that trust is necessary for donations, since donors cannot control the decisions that are made with the donated money. Donations directly made to a homeless person can be freely spent on the preferences of the receiver, for example, on food and clothes, but also on alcohol, drugs and junk food. These choices can contradict with the preferences of the donor concerning the purpose of their money. Therefore, the donor must trust that the homeless person will (partly) spend the money well. An alternative to donating to a homeless person is donating to an organization. This organization uses the donations to provide food, a place to sleep, clothing and personal hygiene products. This way the donor knows the purpose of the donated money. The problem caused by lack of control for the donor over the expenditures of a direct receiver could be solved by donating to an organization. The chosen organization usually has the same standards as the donor. These standards could be based on preferences such as religion or opinions about healthy food and humanity. Since the organization has similar standards as the donor, the donor feels connected to the organization (Uslaner, 2003).

2.2. The effect of income on generosity

The experimental design of this paper makes it possible to explore the level of generosity for different income values. McClelland and Brooks (2004) found that the amount of money donated differs per income level. People with a higher income donated more money (in absolute terms) to charity. The reason for this is that individuals with a higher income have more financial resources available to donate to others (McClelland & Brooks).

Additionally, the study by Erkal, Gangadharan and Nikiforakis (2011) used an experiment to investigate the differences in donations across income. The measurement of

(8)

7 income depended on the earnings in the experiment and was originated from the experimental design. The experimental design is as follows: to earn money, participants were divided into groups of four and were asked to do an individual real-effort task. After the task, the participants were ranked within this group, based on their relative performance. Based on the relative ranking, the participants received a payment. The participant with the highest score also received the highest payment. In the second part of the experiment, participants had the option to offer a part of their earnings to another participant. The highest offers were not made by the participants with the highest income level in a group, but by the participants with the second highest income level in that group. According to Erkal et al., participants with the highest earnings wanted to maximize their profits. On the other hand, participants with the second highest earnings pitied participants with lower earnings because the earnings were partly based on the performance of the group members. Therefore, money transfers depend on the reason for the income inequality.

To further investigate income inequality the paper of Côte et al. (2015) is discussed, which investigated the effect of economic inequality on the size of donations. They found that if economic inequality1_{is high, people with a high income are less generous than people with}

a low income. Côte et al. concluded that high-income people are afraid of losing their privileged position. The gap between wealthy and poor individuals is large, so high-income people have relatively more to lose. On the other hand, when economic inequality is low, people with a high income are more generous. This concludes that the level of generosity for different income levels not only depends on the level of income but also on the economic inequality.

The effect of economic inequality on the size of donations was experimentally confirmed by Sands (2016) who executed a field experiment. Sands described homelessness as the most visible representation of economic inequality. People who were unaware that they were participating in an experiment were exposed to an experimenter who pretended to be either a homeless person or a wealthy person. The subjects were then asked to vote on a regulation to redistribute wealth; the majority voted against wealth distribution after having been confronted with a poor person. This means that visibility of economic inequality negatively affects the willingness to redistributed income.

1_{To define the difference between high and low economic inequality Côte et al. (2015) portrayed economic}

situations to the participants. They did this via processed data from previous research. Every participant viewed a chart displaying an economic situation. The participants did not differ in characteristics from each other. Therefore, the difference in high or low economic inequality was the only difference.

(9)

8 To further investigate the effect of income on the size of donations, Chowdhury and Jeon (2014) explored donations for a donor and a receiver in the same financial group. Chowdhury and Jeon found that individuals donated more to a receiver in the same financial group when they both had a high financial status2_{. This means that people with a higher income}

are more willing to share their money with people who have similar income levels instead of participants with lower income levels.

2.3. The dictator game

A motivator for donating to charity is altruistic behavior. Altruism consists of two components, pure altruism and warm glow altruism. When an individual helps another individual without any personal gain, it is called pure altruism. Warm glow altruism means that an individual helps another person to gain a positive feeling (Bekkers & Wiepking, 2010). Surana and Lomas (2014) described three positive feelings acquired from warm glow altruism. First, donating to charity leads to an increase in self-image, leading to an increase in the self-esteem of the donor. However, Wiepking and Maas (2009) claim the opposite causation and state that more empathy and a higher self-esteem increase charitable behavior. Second, the ability to donate money means that the donor’s earning is higher than somebody else’s. This is an indicator of greater skills and leads to an increase in the donor’s sense of competence. Third, a positive feeling arises when the donor got praised by others for donating money.

A general measurement for generosity is a dictator game, which is a two-player game with a dictator and a receiver (List, 2007). The dictator is given an endowment that he can redistribute by offering some amount to the receiver and keep the rest of the endowment. The receiver must accept the offer of the dictator. From the amount offered, it is possible to measure whether individuals are altruistic or rational. Individuals are rational when they maximize earnings by keeping the entire endowment (Chowdhury and Jeon, 2014).

The amount of money offered by the dictator in the experiment executed in this paper is high compared to the experiment of Hoffman, McCabe, Shachat and Smith (1994). They found that donations in an anonymous dictator game are around ten percent of the endowment. However, the donations in this research are above 50 percent of the earnings, to find a reason for this the research of Brañas-Garza (2006) is added. They found that adding the information

2_{The financial status was determined via the show-up fee in an experimental setting. Individuals with a high}

(10)

9 that the receiver is poor led to offers above ten percent of the endowment. More specifically, it led to unequal distributions benefitting the receiver. In addition, the donations increased after informing the dictator about the purpose of the money. In the experiment of Brañas-Garza, the specified purpose was that the donated money was used to purchase medicines for low-income receivers.

In conclusion, information about the low-income level of the receiver, as investigated by Garza (2006), leads to an increase in generosity. However, in the research of Brañas-Garza the dictator received three five-dollar bills and was restricted to only using five-dollar bills in his offers. It could be that this restriction affected the amount offered by the dictator, resulting in an increase in generosity.

2.3.1. External validity of the dictator game

According to Franzen and Pointer (2013), various elements influence the external validity of the dictator game. First, a potential self-selection bias of participants could influence the results. The self-selection bias leads to more selfish participants because they participate for the monetary reward (Falk, Meier & Zehnder, 2013). Another consequence of the self-selection bias is that people self-select into an experiment to help the researcher. This means that the participants have higher prosocial preferences. However, Falk et al. did not find any difference between the group that was self-selected into the experiment and non-participants3_{. Thus, a}

self-selection bias does not seem to influence the experimental results. This is an important finding for this paper, where a lot of participants self-selected into the experiment as a favor to the experimenter.

Second, because the endowment is received for free, the distribution of money in a dictator game is not comparable to a real-life situation. Receiving the money from an experimenter leads to an increase in the dictator’s generosity (Gąsiorowska and Helka, 2012). This flaw in external validity could be reduced by adding a real-effort task (Benndorf, Rau and Sölch, 2014). With the real-effort task, the dictator would have to do a task to earn the money that is used to make an offer to the receiver later on. A problem not solved by the real-effort task is physically holding money (see appendix B for additional information about the influence of money holding and money priming on the external validity of the dictator game).

3_{Data of non-participants is obtained via surveys and actual donations. The surveys were gathered at the university,}

where every student had to complete a questionnaire. This questionnaire included questions about prosocial behavior.

(11)

10 Finally, subjects in a dictator game are aware of the fact that they are participating in an experiment. This can lead to an experimenter demand effect. An experimenter demand effect is where subjects adjust choices to the expectations of the researcher (Zizzo, 2010). This could affect the results in these experiments.

Last, Franzen and Pointer (2013) explored the differences in external validity between a lab experiment and a field experiment in a dictator game. In the field experiment, the “misdirected-letter technique” is used. The individuals were not aware of their participation in the experiment. They received a letter intended for another person. This letter contained an amount of money, around ten dollars. It was tested whether the participants returned the letter and the corresponding amount of money. Since the field experiment had similar results as the lab experiment, Franzen and Pointer concluded that external validity is confirmed.

However, contradictory results were found in the research of Winking and Mizer (2013). In their experiment, participants were also unaware of participating in the experiment. The participants received an endowment from a passerby in the form of casino chips. The participants had the option to share the chips with a stranger, which was a collaborator in the experiment. Winking and Mizer found that none of the participants shared their chips with a stranger. Therefore, they concluded that there is a problem with the external validity of the dictator game.

A difference in the experimental design for the above researches is money entitlement. In the experiment of Franzen and Pointer (2013), money entitlement is low because participants received an endowment intended for somebody else. In the experiment of Winking and Mizer (2013), the participant is entitled to the received endowment given in form of casino chips because the chips were a gift to the participant. When people feel entitled to money, they would rather keep it for themselves. This lowered the offered amount to the other participant.

Moreover, there is a difference in money valuation between both experiments. Franzen and Pointer (2013) used a monetary reward, which is more reliable than the endowment in casino chips used by Winking and Mizer (2013). People may have different valuations of casino chips, while money has a constant valuation. Money entitlement influences the level of generosity in the dictator game.

2.4. Summary

Based on the described literature, the amount of money donated to a charitable cause is influenced by two main factors: the level of income and trust of the donor. Erkal et al. (2011)

(12)

11 found in their experiment that the participants with the second highest earnings offered the highest donations to another (poor) participant instead of the participants with the highest income. The inconsistency of the effect of income in the level of donations is supported by Côte et al. (2015). Another important factor influencing generous behavior is the level of trust that the donor has. Trust is necessary for donating money, especially when the donations are made towards strangers. On the other hand, trust could increase after charitable giving. Additionally, Wiepking and Maas (2009) found that people with higher cognitive abilities are surrounded by a broader social group. This has an influence on charitable giving since people with a broader social group are more generous compared to people with a less diverse social group. Therefore, trust is higher for individuals with high cognitive abilities as well as offered monetary donations.

Finally, there is also a connection between the level of trust and economic inequality, described in the previous paragraph. Uslaner (2000) found that trust is influenced by attitudes toward the future. These attitudes are pessimistic when economic inequality is higher. According to Côte et al. (2015), trust is low when economic inequality is high. Low amounts of trust lead to less generous individuals. An advantage of a more equal economy is that it could increase generalized trust and therefore leads to an increase in generosity.

Since previous studies about the dictator game have mainly focused on offers made by the dictator directly to another participant who can freely spend the money, it is interesting to see what happens to the amount of money offered when the donation is not directly given to the receiver but via a more transparent third party. In this paper, a dictator can directly make an offer to a homeless person or indirectly by donating to an organization that helps homeless people. Donations made directly to a homeless person could be spent freely, including on alcohol, drugs and unhealthy food. On the contrary, an organization declares the purpose of the donated money on forehand. This makes the objective of the donated money more transparent, which may lead to an increase in the amount of money offered by the dictator.

(13)

12

3. Hypotheses

The literature shows that different levels of generosity can appear in a dictator game. When looking at the donations made to a homeless person or to an organization, differences in generosity can occur. Brañas-Garza (2006) found that adding information about the specific purposes for the offered money increased the offers. For donations made to a homeless person the expenditures are not further specified, this occurs for donations made to an organization. This leads to the second hypothesis:

Hypothesis 1: There is an increase in the level of generosity when indirectly donating to an

organization compared to directly donating to a receiver, represented in this case by a homeless person.

Furthermore, the literature described the effect of money entitlement in an experimental design. Participants were less generous when they felt entitled to the earnings. When participants felt less entitled to the earnings they were indifferent about the purpose. Therefore, it is expected that the difference in donations made to a homeless person or an organization is influenced by the way in which the earnings were obtained. This is implemented in the experiment, since the design differentiated between participants who earned their money via a real-effort task and the participants who received the endowment for free. This leads to the following two hypotheses:

Hypothesis 2: The participants preforming a real-effort task would rather donate money to an

organization instead of a homeless person.

Hypothesis 3: No difference will occur between donations made to a homeless person

compared to the donations made to an organization when the participants received an endowment.

Previous literature expected that donations made by participants who obtain their earnings via a real-effort task are less generous than those issued by participants who received an endowment. Therefore, the fourth hypothesis is as follows:

Hypothesis 4: Doing a real-effort task leads to a lower level of generosity compared to

(14)

13

4. Methodology

This section describes the experimental design used to obtain the data. The experiment is an online experiment designed in Qualtrics. After the description of the experiment, certain features of the experiment are explained in more detail.

4.1. The experimental design

The dataset consists of the data of 85 participants4_{with different educational backgrounds}

recruited via e-mail and Facebook. The experiment started with 95 participants however, 10 participants did not finish the experiment and were excluded from the experiment. A between-subjects-design was used, where the participants were divided into two groups. The sessions for both groups consisted of two parts. In the first part, participants would obtain money either by doing a real effort task or as an endowment. This full amount of money was used twice, once in both dictator games.

In the first part, the 34 participants in the first group earned money by doing a real-effort task. The 51 participants of the second group received the initial endowment from the experimenter5_.

The second part of the experiment occurred one week after the end of the first part. At the start of the second part, participants were informed about their earnings obtained in the first part of the experiment. In the second part, participants were asked to play two dictator games. In both dictator games, they used the money obtained from part one. In the first dictator game, part of the money could be given directly to a homeless person. In the second dictator game, the money is offered indirectly to homeless people via an organization that helps homeless people. The donations made in both dictator games were paid in real-life. In both cases, the participant had the option to keep the full amount of earnings. A summary of the above described groups is given in Table 1.

4_{The participants consist of friends, family, fellow students and strangers.}

5_{The unequal distribution between the real-effort task and the endowment group is caused by the used}

randomization program. Participants who for-fitted the task were counted as group member by Qualtrics. This data was unusable. More individuals for-fitted the real-effort task, therefore the groups were unequal.

(15)

14

Table 1: Overview of the two different groups

Notes: There are two different groups. The first group is doing a real effort task. The second group is not doing a task and receives the money from the experimenter; this group is labelled as endowment. There are two different receivers, receiver A is a homeless person and receiver B is an organization. In total there are 85 participants, 34 participants for treatment 1 and 51 participants for treatment 2.

The real-effort task used in part one of the experiment is an encryption task. This encryption task is based on the one used in the experiment of Benndorf, Rau and Sölch (2014). Participants had to transform letters into a numerical sequence. The participants were confronted with given random combinations of four letters. Each letter corresponded to a three-digit number as given in a table shown to the participants. The letters in the table were not alphabetically ordered and the letter order was randomly assigned and randomized for each new letter combination. The three-digit numbers matched to the letters also changed for every new letter combination (see Appendix B for the instruction of the encryption task).

Before the encryption task started there was a trial round, which consisted of five random letter combinations.6_{The trial round was immediately followed by the encryption task when the}

participant had answered the question in the trial round correctly, if not they could not continue with the experiment. In the encryption task, participants had five minutes to arrange as many random letter combinations as possible. After these five minutes, part one of the experiment ended. Participants were notified about their earnings at the beginning of the second part of the experiment7_.

The second part of the experiment consisted of two dictator games. In one dictator game, the participant could offer a part of the earnings to a homeless person. In the other, donations

6_{In contrast to the rounds in the main task, the trial round has no time constraint. A timer counting up instead of}

down is added to make the layout completely equivalent to the main task. A trial round was conducted to make sure all the participants understood the experiment. There was no feedback in the trial round, however participants were excluded from the experiment when they could not complete the trial round.

7_{Payments in the experiment itself are expressed in points instead of euros; participants are aware that this is not}

the amount they get paid at the end of the experiment. The paid amount is the amount at the end of the second part of the experiment when points are converted into euros and one point was equal to one euro.

Receiver

Homeless person (A) Organization (B)

Dictator Real-effort task (N = 34) Treatment 1A (N = 34) Treatment 1B (N = 34) Endowment (N = 51) Treatment 2A (N = 51) Treatment 2B (N = 51)

(16)

15 were made to ‘de Regenboog Groep’, an organization that provides homeless people with food, hygiene products, clothes and shelter. Participants could also choose another organization instead of donating money to ‘de Regenboog Groep’. The participant had to come up with the alternative organization. However, the participant could only choose an organization that also helps homeless people. This choice was allowed in order to ensure that distrust toward ‘de Regenboog Groep’ would not influence the offers made to an organization.

To control for a possible order effect8_{, the order of these two dictator games was randomly}

decided for every participant. To distract participants from wondering about the research question ten questions about the participants’ characteristics were added (see Appendix B for the personality trait questions). This is important to reduce the experimenter demand effect (Zizzo, 2010). This means that the choices made by the participants were influenced by what participants think would be useful for the experimenter. To conclude, every participant answered twelve randomized questions, two dictator game questions and ten personal characteristics questions. These questions were followed by a questionnaire that collected information about feelings of jealousy and trust and general preferences for donating money.

4.2. Rationale for the experimental design

The division into two groups (real-effort task and endowment) was based on the expected level of generosity described by the hypotheses. For the group who received the endowment, no difference was expected in the level of generosity for donations to a homeless person or to an organization. For the group who did the encryption task, donations to an organization were expected to be higher than donations to a homeless person. The different expectation between both groups is based on the literature about money entitlement, which is the claim the individual has on the money. Doing a task increased the money entitlement. Expected is that the participants who earned the money via a real effort task donate less of their earnings, especially towards a homeless person.

To reach participants with different educational backgrounds, the instructions of the experiment are in Dutch and English. For instance, native Dutch-speaking-participants with a lower level of education might have dropped out of the experiment if the instructions were only provided in English. To control for education in the data processing, participants were asked to state their achieved educational level.

(17)

16 Prior to the start of the second part, the participants are notified of the amount of points they earned in the first part of the experiment. In the first part the participants were asked to give their e-mail address. This way the participants could be provided with the information about their earnings in part two of the experiment, which was conducted one week after the first part of the experiment. To prevent for any mistakes about unreported earnings, instructions were provided with a message about the earnings. Participants received a notification that asked them to contact the experimenter if a mistake occurred and the participants did not receive their amount of earned points (see Appendix B for the instructions of the second part of the experiment).

At the start of part one, some general information about the gender, age, educational level and income of the participants was gathered. The second part of the experiment ended with a survey. In the first question of part 2, participants were asked whether jealousy influenced the way they distributed the endowment. If the participant answered yes, the question was followed by a Likert scale9_{(from 1 to 7) to measure the amount of jealousy. The second question asked}

whether the participants trusted the experimenter to transfer the money to a homeless person. This is followed by several questions about the preferences of participants with respect to donating money. The first of these questions asked whether participants would rather help more people in a general way or one person in a specific way. The second asked how often the participant directly helped others, besides family and friends. The third asked whether the participant helped others via an organization. The last question asked how much money the participant had donated in the past year.

4.3. Reward system

The earnings in the first part of the experiment were based on a tournament ranking scheme. To implement a ranking, participants were randomly divided into groups of four. The groups that executed the encryption task were ranked based on the number of right answers compared to the other members of their group. The group that received the endowment from the experimenter was randomly ranked. Based on their rank, participants received 20, 15, 10 or 5 points. The highest ranked participant received the highest amount of earnings, while the lowest ranked participant received the lowest possible amount. In the instructions, participants were informed that one point equaled one euro.

(18)

17 Additionally, the instructions informed the participants about the reward system to be applied at the end of the experiment. Four people would be randomly drawn for payment. They got paid for only one of the two played dictator games; either the game where an offer is made to a homeless person or the game where the offer is made to an organization was randomly chosen. In case the donation was made to an organization, the experimenter transferred the donation to the organization and provided the winning participant with a receipt of the donated money. When the donation was made to a homeless person, the experimenter donated the money to a homeless person and provided the winner with a picture of the donation as proof that the donation took place. It was possible that the homeless person would not agree to being on a picture. Therefore, the offers made to a homeless person partly depend on the participant’s trust in the experimenter. Therefore, the survey question was added asking the participants whether they trust the experimenter to transfer the offers to a homeless person.

All four participants who were selected for payment received an e-mail with the request to send their bank information. Out of the four participants selected for payment, one of them got paid for the dictator game where the money was offered to a homeless person. This participant kept the full endowment. The other three players got paid for the dictator game where offers were made to ‘de Regenboog Groep’. The winners received the money they wanted to keep from the endowment to their bank account and they received an e-mail with the receipt of the transaction to ‘de Regenboog Groep’.

(19)

18

5. Results

In this section, the results of the experiment are presented. First, a comparison in generosity between donations to a homeless person and an organization is given. This is followed by the effect of income on generosity. Additionally, results are represented more precise with the help of regressions, taking the control questions into account. The results are concluded with some possible limitations regarding the interpretation of the results.

5.1. Descriptive analysis

Initially, a comparison between the average donations to an organization and to a homeless person was considered, taking all the participants into account. The first observation shows that participants donated an average of 7.2 euros to a homeless person versus 8 euros to an organization. The second observation is that average donations after doing a real-effort task are higher to an organization (7.3 euros) compared to the donations made to a homeless person (6.6 euros). This same comparison is made for the participants who received an endowment. Here donations made to an organization were also higher (8.4 euros) compared to donations to a homeless person (7.5 euros). The last observation is that the average donations after doing a task (6.6 and 7.3) were lower than the donations by the group that received the endowment (7.5 and 8.4). For the means of offered donations, see Table 2.

Table 2: Average main variables

Treatment Homeless Person Organization

Real-effort task and Endowment group 7.1667 (5.8120) 7.9647 5.8587) Real-effort task 6.6061 (5.9421) 7.3382 (5.7134) Endowment 7.5294 (5.7562) 8.3824 (5.9729)

Note: The total average earnings for all participants together is 12.70588 euros, with a standard deviation of 5.64595. The table distinguishes between different donations in different treatments. The corresponding standard deviations are in parentheses.

(20)

19 Next, the spread and median for both treatment groups are displayed through a boxplot in Figure 1. All subgroups have the same median of donating, namely five euros, which is not central. This means that the data is not symmetrically distributed. However, the spread for donations to an organization is wider compared to the spread of donations to a homeless person. This applies for both groups and the total amount of participants. This means that the dataset containing the offers made to an organization is more variable. Also, the donations to an organization were higher than the donations made to a homeless person.

Figure 1: Boxplot representing the scatter of the donations

Notes: Homeless (Both Treatments) refers to the donations to a homeless person for all participants. Organization (Both) refers to the donations to an organization for all participants. Homeless (Task and No Task) are the results per treatments, similar for Organization (Task and No Task). The minimum and maximum values are indicated by the bottom and top line of the boxplot. The middle line is the median. All possible donations are between 0 and 20 euros.

5.2. Comparison of generosity

The data was further explored with a Wilcoxon signed rank test10_{, since the gathered data was}

not normally distributed. A Wilcoxon signed rank test measures whether the median of the differences is zero for not normally distributed data using a within subject design. There are

10_{The Wilcoxon signed rank test is a nonparametric test to calculate the difference between two groups with the}

(21)

20 two variations of the Wilcoxon signed test. The sign rank test assumes that both distributions are the same and the sign test has no assumptions concerning the distribution sets (Stata manual).

5.2.1. Hypothesis 1

There is an increase in the level of generosity when indirectly donating to an organization compared to directly donating to a receiver, represented in this case by a homeless person.

For the sign test, the null hypothesis is that the distributions of donations to a homeless person are equal to the distribution of donations to an organization. The standardized z-value was 3.005 (p-value of 0.0027). This means that the results are significant and that the null hypothesis is rejected at a significance level of 5 percent11_{. Out of 84 observations, 24 showed a difference}

in donations between donations made to a homeless person or an organization. Out of these 24 observations, 19 donations were greater for an organization than a homeless person. This shows that the difference found in the offers made by the dictator to a homeless person or an organization is in favor of an organization. Therefore, indirect donations lead to increased generosity compared to direct donations.

5.2.2. Hypothesis 2

The participants preforming a real-effort task would rather donate money to an organization instead of a homeless person.

Additionally, the difference in donations to a homeless person or to an organization for the group who did a real-effort task is included. First, a sign rank test was performed for the difference between donations to a homeless person and to an organization for the participants who did the real-effort task. The two-sided p-value of 0.1203 is not significant; the null hypotheses cannot be rejected. There is no difference between donating to a homeless person and an organization for the participants doing a real-effort task.

5.2.3. Hypothesis 3

No difference will occur between donations made to a homeless person compared to the donations made to an organization when the participants received an endowment.

11_{This significant result is also found with a t-test for the difference in means, between donating to a homeless}

person versus an organization, with a p-value of 0.0006. Participants give on average more to an organization than to a homeless person seen from the one-sided t-test whereby the results for the difference in means are smaller than 0 is significant with a p-value of 0.0003.

(22)

21 Furthermore, the difference in donations to a homeless person and to an organization for the endowment group is explored. For the group receiving the endowment for free, a significant difference for donations to a homeless person compared to donating to an organization was found, with a p-value of 0.0094. The null hypothesis, distribution in donations to a homeless person is equal to the distribution of donations to an organization, can be rejected.

Summarized, the null hypothesis is rejected for the group who did the real effort task. However, the null hypothesis is not rejected for the endowment group. Thereby, the average differences in donations represented in Table 2 are almost the same. For the group who did a real-effort task the donations to an organization are 0.73 euros higher compared to the donations to a homeless person. For the endowment group the donations to an organization are on average 0.85 euros higher than to a homeless person. An explanation for the difference in significance is the difference in sample size. The real-effort task group consists of a smaller sample size (34 participants) compared to the endowment group (51 participants).

5.3.4. Hypothesis 4

Doing a real-effort task leads to a lower level of generosity compared to individuals who received an endowment.

To confirm the reliability of the above result, the difference in donations to a homeless person between the real-effort task group and the endowment group must be taken into account. Also, the difference in donations to an organization between a real-effort task group and the endowment group must be taken into account. A Wilcoxon rank sum test was used. This is a nonparametric test for between subject designs. All results concerning differences in performing a task or receiving the endowment for free are not significant. This means that implementing a real-effort task has no influence on generosity compared to receiving the money for free.

5.3. The effect of income

Furthermore, the differences in the level of generosity were compared to each other across different income levels. Examining the means, participants with an income of 20 euros donate on average 9.1 euros to a homeless person and 10.4 euros to an organization. Participants with the second highest earnings of 15 euros donate on average 10.6 euros to a homeless person and 11.3 euros to an organization. The group with the second highest earnings donated the most

(23)

22 when the donations were expressed in absolute values. The corresponding means for different income levels are given in Table 3.

There is a significant difference in average donations between a homeless person and an organization for the group who earned 20 euros12_{. Consequently, the null hypothesis—that}

there is no difference in donations—can be rejected. The difference between donations to a homeless person and an organization for participants earning 15 euros is not significant at a 5 percent level. However, the results are significant at a 10 percent level13_{. For earnings of 10 and}

5 euros, no significant results were found, meaning that there is no difference in donations between a homeless person and an organization14_.

Table 3: Average donations across income

Income Average donations to a

Homeless Person Average donations to an Organization

Earnings 20 9.1364 (7.3179) 10.4090 (7.2221) Earnings 15 10.6364 (5.2784) 11.3478 (5.2363) Earnings 10 5.0000 (3.9299) 5.3158 (4.1373) Earnings 5 3.4286 (2.0389) 4.0952 (1.3074)

Note: The table above gives the average donations in euros for different income levels. Standard deviations are in parentheses.

In addition to the absolute monetary donations, the percentage of the endowment donated to an organization or homeless person is shown in Figure 2. This shows that the lowest ranked group, who earned 5 euros, donated the highest percentage of their earnings, followed by the participant who earned 15 euros. Participants who received 10 euros ranked third and the lowest relative offers were by the participants with the highest income level of 20 euros.

12_{This is confirmed with the Wilcoxon signed rank test, finding a p-value of 0.0251.} 13_{Tested with the Wilcoxon sign rank test, with a p-value of 0.0834}

14_{A Kruskal-Wallis test is performed, the donations to a homeless person are significant different across income}

with a p-value of 0.0003. The donations toward an organization are also significantly different across income at a level of 0.0001.

(24)

23

Figure 2: Percentage of endowment donated

Notes: This graph shows relative differences between donations, expressed in percentages.

5.4. Control variables

Even though the data for this experiment is not normally distributed, some regressions which take the control question into account are done. First, the participants’ trust that the experimenter actually transferred the donations made to a homeless person is measured. The corresponding value had no significant influence on the donations. This means that the donations to a homeless person were not influenced by a potential lack of trust in the experimenter. Second, the participants were asked whether they prefer to help one individual in a specific way or to help more people in a general way. A significant15_{result was found;}

participants preferred helping an individual in a specific way over helping more people in a general way. However, this did not influence the donations made in the experiment, since the results found in Table 416_{were not significant. Furthermore, earnings of 20 or 15 euros had a}

significant influence on the amount donated to a homeless person as well as to an organization. This conclude that participants with a higher income (20 or 15) donate more money.

Additionally, a regression is done on the amount donated divided over the earnings (represented in Table 4). Here is found that participants who earned 20 euros donated significantly less to a homeless person than participants who earned 5 euros. For donations made to an organization, participants who earned 10 or 20 euros donated significantly less than

15_{Wilcoxon sign rank test, a significant difference found in preferences with a p-value of 0.0057}

16_{Table 4 does not contain all the control variables used in the regression, for the complete table refer to Appendix}

A. The experiment also controlled for jealousy. Out of 85 participants, only one experienced jealousy. Therefore, they did not answer the additional questions referred to jealousy.

0 10 20 30 40 50 60 70 80 90

Earnings 20 Earnings 15 Earnings 10 Earnings 5

P er ce n ta ge o f en d o w me n t d o n ate d

(25)

24 participants who earned 5 euros. The percentage donated by the participants earning 15 euros is close to the percentage donated by participants who earned 5 euros therefore no significant difference is found (this is also seen in Table 3 where the participants who earned 15 euros donated the highest absolute amount of money compared to the other earning groups). The significant results found based on the regressions are confirming the results that are shown in Figure 2.

Table 4: Regressions on the amount donated

(1) (2) (3) (4) Donations to a homeless person Donations to an organization Ratio donations to a homeless person Ratio donations to an organization Task -0.754 -0.568 -0.008 0.072 (1.106) (1.226) (0.088) (0.118) Gender 2.298+ _2.134 _0.198* _0.174 (1.174) (1.303) (0.094) (0.126) Age 0.022 0.016 0.001 0.000 (0.059) (0.065) (0.005) (0.006) Earnings (€10) 2.280 0.874 -0.138 -0.465** (1.557) (1.727) (0.124) (0.166) Earnings (€15) 8.028** _6.587** _0.100 _-0.255 (1.514) (1.680) (0.121) (0.162) Earnings (€20) 5.375** _4.930** _-0.245* _-0.553** (1.507) (1.671) (0.120) (0.161) Trust 0.060 0.254 -0.007 0.0127 (0.469) (0.520) (0.037) (0.050) Helping individuals 0.326 0.350 0.024 0.012 (0.340) (0.377) (0.027) (0.036) Helping generally 0.230 0.025 0.010 -0.021 (0.279) (0.309) (0.022) (0.030) Euros donated 0.000 -0.002 0.000 0.000 (0.002) (0.002) (0.000) (0.000) Constant -1.029 1.702 0.358 0.941** (3.038) (3.370) (0.243) (0.325) Observations 80 80 80 80

Standard errors in parentheses

+_{p < 0.1,}*_{p < 0.05,}**_{p < 0.01}

Notes: The variable Task is a dummy variable (Task =1 and Endowment =0). The variable Gender is a dummy variable, with value 1 for a male participant and value 0 for a female participant. The variable ‘earnings’ is a dummy variable for the different amount of earnings, 20, 15, 10 or 5 euros. The variable ‘Trust’ represented the trust the participant had in transferring the donations to a homeless person. This is measured with a Likert scale running from 1 till 5. The average trust in the experimenter is 4.1 out of 5; there is no reason to assume that people donated less because of distrust in the experimenter. The variable ‘Helping individuals’ and ‘Helping generally’ consisted of a Likert scale (From 1 till 7). The lowest value 1 represented totally disagree (so the participant had no preferences in helping an individual in a specific way or helping more individual is a general way). The highest value of 7 represented totally agree (the participant preferred to help an individual in a specific way or for the other question to help more people in a general way). ‘Euros donated’ is the amount in euros donated by the participant to charity over the past year.

(26)

25

6. Discussion and Conclusion

6.1. Conclusion

To answer the research question whether people are more generous in donations made to an organization compared to a homeless person, an experiment is executed. The overall donations to a homeless person were lower than to an organization. This means that participants prefer to help homeless people indirectly via an organization.

Furthermore, two different groups of participants were used. For the participants who received the money from the experimenter, significant differences in the donations towards an organization versus a homeless person were found. For participants who did the real-effort task, this difference was not significant. The fact that a significant difference was found between donations to a homeless person compared to donations to an organization only for the group who received an endowment cannot completely be attributed to the difference in doing a real-effort task or receiving an endowment, because the average generosity across both treatments was not significantly different. A possible explanation is the difference in sample size between both groups. Further research is needed to confirm this finding.

In addition, the difference in the level of generosity across income whilst donating to a homeless person or to an organization was taken into account. A difference between donations made to an organization or to a homeless person was found for the higher levels of income (20 and 15). This means that the participants with higher incomes preferred to offer money to an organization rather than to a homeless person. For the participants with lower incomes, the difference between donating to a homeless person or to an organization was not significant. Additionally, the absolute amount donated by the participants with different income levels was taken into account. A significant difference was found for the different income levels between donations to a homeless person and to organizations. Donations from the second highest income group were highest in an absolute sense, which corresponds to the results provided by Côte et al. (2015). However, in percentage terms, the lowest ranked group donated most. This last result could suffer from a response bias as participants might have used multiple sources outside of the experiment to make a decision. These decisions are not completely based on the changed variable in the experiment. Therefore, five euros can be seen as a small amount of money compared to a real-life income. This could have caused participants to donate their full endowment. This redistribution option could never be implemented in real life, as giving all of

(27)

26 your income to less wealthy people would leave you incomeless. Therefore, these results might be unreliable.

6.2. Limitations

This experiment consisted of an online experiment designed via Qualtrics. A problem of an experimental design executed via Qualtrics is that participants who dropped out in part one of the experiment were still registered to one of the groups. After dropping out, it was possible to participate again because the e-mail address was not blocked yet. Due to this flaw, a lot of participants dropped out when they were classified in the real-effort task group and participated again into the group who received their endowment from the experimenter. The people who did the real-effort task did not mind to invest time. They probably did not experience this as putting in effort to earn their money.

Even though Falk et al. (2013) did not find any differences in the result due to a self-selection bias, a lot of the participants are friends, family or acquaintances of the experimenter who want to help the experimenter to receive the ‘right’ results. In other words, an experimenter demand effect (Zizzo, 2010) might have influenced their natural choice behavior. The plausibility of this bias having occurred is increased because of anonymity flaws. In the experiment, participants had to fill in their mail address to participate at all. A lot of the e-mail addresses contained surnames or last names, which made it easy to identify participants. Furthermore, participants made high offers compared to the average of previous standard dictator games. Some participants even donated all their earnings. These higher donations were in line with the results of the experiment of the paper by Brañas-Garza (2006), whereby added information about the low-income level of the receiver led to higher donations. Additionally, the earnings used in this experiment were low compared to a real-life situation. The higher earnings were explained as follows by the participants: they would not donate all their money if the endowment was higher. They even saw the experiment as an opportunity to donate money to a charity17_.

Finally, for the differences found in this paper a Wilcoxon sign rank test is used. The Wilcoxon sign rank test only measures a difference between two groups. However, the test does not specify what the difference is between two groups, an increase or a decease. Neither does it report the amount that the two groups differ from each other. Therefore, the differences found in this paper cannot be specified.

(28)

27

6.3. Discussion

The results of this research indicate that there is a significant difference between donations to a homeless person compared to donations to an organization. This means that people would rather help homeless people via an organization to manage the donations. This increase in generosity could be due to an increase in control, which was described by Bekkers (2003) as a substitute for trust. Trust is needed for donations towards a homeless person, because there is no control over their spending pattern.

Another result is a difference in donations for participants with higher values of income (20 and 15)18_{. This could be compared with the research of Chowdhury and Jeon (2014), stating}

that individuals prefer to offer money in a dictator game to participants with the same income level. A higher income means that a homeless person is no longer part of the same financial group. People prefer to donate to an organization instead. Thus, people with higher real-life incomes could accomplish a significant difference between donating to a homeless person versus an organization. However, this is a generalization, as there is not enough data and in this experiment to confirm the conclusion about higher real-life income.

The results in the experiment differed from the answers to the control questions. First, participants answered that they prefer to help one person directly rather than helping more people in a general way. This is not found in the experiment where people donated more to an organization (helping people in a general way) than to a homeless person (helping somebody in a specific way). However, in the control question used in the experiment, the individual is not specified as a homeless person. The question asked for the preferences of the participant to help an individual in a specific way. Therefore, the participant can freely interpret characteristics about the individual. This could have increased the preference to redistribute with the individual.

This paper focused only on monetary generosity. Further research should take the effect of non-monetary generosity into account and should also focus on performing a real-effort task that is more time consuming, so participants consciously perceive the invested time and effort in earning the money. Besides that, the earnings should be increased to get more results about significant differences in generosity for incomes above 20 euros.

The most important recommendation for further research is to conduct a field experiment and research the level of generosity when individuals can directly donate money to

18_{For the lower incomes, the difference between donating to an organization or a homeless person, are not}

(29)

28 a homeless person where the purpose of the expenditures of the homeless person are known. When the purpose of the money donated to a homeless person is clarified, a homeless person would be more comparable to an organization. When researching this difference, the only difference between a homeless person and an organization is direct or indirect donating money. The expenses of a homeless person could be controlled and restricted via pin transactions instead of cash.

(30)

29

References

Bekkers, R. (2003). Trust, accreditation, and philanthropy in the Netherlands. Nonprofit and Voluntary Sector Quarterly, 32(4), 596-615.

Bekkers, R., & Wiepking, P. (2011). A literature review of empirical studies of philanthropy: Eight mechanisms that drive charitable giving. Nonprofit and Voluntary Sector Quarterly, 40(5), 924-973.

Benndorf, V., Rau, H. A., & Sölch, C. (2014). Minimizing learning behavior in experiments with repeated real-effort tasks.

Bjørnskov, C. (2007). Determinants of generalized trust: A cross-country comparison. Public choice, 130(1), 1-21.

Brañas-Garza, P. (2006). Poverty in dictator games: Awakening solidarity. Journal of Economic Behavior & Organization, 60(3), 306-320.

Chowdhury, S. M., & Jeon, J. Y. (2014). Impure altruism or inequality aversion?: An experimental investigation based on income effects. Journal of Public Economics, 118, 143-150.

Côté, S., House, J., & Willer, R. (2015). High economic inequality leads higher-income individuals to be less generous. Proceedings of the National Academy of Sciences, 112(52), 15838-15843.

De Winter, J. C., & Dodou, D. (2010). Five-point Likert items: t test versus Mann-Whitney-Wilcoxon. Practical Assessment, Research & Evaluation, 15(11), 1-12.

Erkal, N., Gangadharan, L., & Nikiforakis, N. (2011). Relative earnings and giving in a real-effort experiment. The American Economic Review, 101(7), 3330-3348.

Falk, A., Meier, S., & Zehnder, C. (2013). Do lab experiments misrepresent social preferences? The case of self-selected student samples. Journal of the European Economic Association, 11(4), 839-852.

Franzen, A., & Pointner, S. (2013). The external validity of giving in the dictator game. Experimental Economics, 16(2), 155-169.

Gąsiorowska, A., & Hełka, A. (2012). Psychological consequences of money and money attitudes in dictator game. Polish Psychological Bulletin, 43(1), 20-26.

Geven in Nederland 2017: 20 jaar cijfers en trends over gevend Nederland. (2017, April 24). Retrieved January 06, 2018, from https://www.cbf.nl/nieuwsbericht/geven-in-nederland-2017-20-jaar-cijfers-en-trends-over-gevend-nederland

(31)

30 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.

List, J. A. (2007). On the interpretation of giving in dictator games. Journal of Political economy, 115(3), 482-493.

McClelland, R., & Brooks, A. C. (2004). What is the real relationship between income and charitable giving?. Public Finance Review, 32(5), 483-497.

Reutner, L., & Wänke, M. (2013). For my own benefit or for the benefit of others: Reminders of money moderate the effects of self-related versus other-related persuasive arguments. Social Psychological and Personality Science, 4(2), 220-223.

Sands, M. L. (2017). Exposure to inequality affects support for redistribution. Proceedings of the National Academy of Sciences, 201615010.

Smeeding, T. M. (2005). Public policy, economic inequality, and poverty: The United States in comparative perspective. Social Science Quarterly, 86(s1), 955-983.

Smith, C., & Hill, J. P. (2009). Toward the measurement of interpersonal generosity (IG): An IG scale conceptualized, tested, and validated.

Stata Manual. Consulted from: https://www.stata.com/manuals13/rsignrank.pdf

Surana, P. K., & Lomas, T. (2014). The power of charity: Does giving away money improve the wellbeing of the donor?. Indian Journal of Positive Psychology, 5(3), 223.

Uslaner, E. M. (2000). Producing and consuming trust. Political Science Quarterly, 115(4), 569-590.

Uslaner, E. M. (2003, March). Trust in the knowledge society. In Conference on Social Capital, Cabinet of the Government of Japan (pp. 24-25).

Vohs, K. D., Mead, N. L., & Goode, M. R. (2006). The psychological consequences of money. science, 314(5802), 1154-1156.

Wiepking, P., & Maas, I. (2009). Resources that make you generous: Effects of social and human resources on charitable giving. Social Forces, 87(4), 1973-1995.

Winking, J., & Mizer, N. (2013). Natural-field dictator game shows no altruistic giving. Evolution and Human Behavior, 34(4), 288-293.

Zizzo, D. J. (2010). Experimenter demand effects in economic experiments. Experimental Economics, 13(1), 75-98.

(32)

31

Appendices

Appendix A

Table 5: Information about the participants

Treatment N Dutch Male Average

Age Educational Level Amount of right answers Real effort Task 34 22

(61.1%) 17 (47.2%) 28 5.6 13.3 (53.2%) Endowment 51 35 (68.6%) 20 (38.5%) 30 5.5 - Total amount of participants 85 57 (67.0%) 37 (43.5%) 29.2 5.6 -

Notes: N refers to the number of participants in the specific group; Dutch refers to the participants who chose Dutch as languages for the instructions. The Educational level consists of 8 options:1= None 2= Elementary school, 3= High school, 4= MBO, 5= HBO, 6= University Bachelor, 7= University Master and 8= PhD. Important to notice is that none of the participants had educational level 1, 2 or 8. In the table is shown that the average education level is located between an HBO and a Bachelor diploma. The percentage of the total amount is shown between brackets.

For this experiment, a total amount of 85 participants is used, distributed over two groups. The first group exist of 34 participants doing a real effort task. The second group consists of 51 participants receiving an endowment from the experimenter. Of the total amount of 85 participants, 37 were male (43.5%). The average educational level of the participants is between a HBO and a university Bachelor degree. Fout! Verwijzingsbron niet gevonden. shows an overview of the background information of the participants.

(33)

32

Table 5: Summary of all variables

Variable Mean Std. Dev.

Right Answers 13.32353 out of 25 4.96485

Earnings 12.70588 5.64595

Trust in the experimenter 4.08642 out of 5 1.1958

Treatment Keep Homeless

Person

Organization Real effort task and No Task

5.511905 (6.072839) 7.166667 (5.81198) 4.741176 (5.870306) 7.964706 (5.858723)

Real effort task 6.121212 (7.231691 6.606061 (5.942113) 5.455882 (6.888855) 7.338235 (5.713448) No Task 5.117647 (5.229329) 7.529412 (5.756224) 4.264706 (5.098385) 8.382353 (5.972929)

(34)

33

Table 6: Corresponding Earnings to the amount of right answers for every participant

Participant Amount of right answers Earnings

1 4 5 2 5 5 3 5 5 4 5 5 5 9 10 6 9 5 7 10 10 8 10 5 9 11 10 10 11 10 11 11 10 12 12 15 13 12 15 14 12 15 15 12 5 16 13 20 17 13 15 18 13 15 19 13 5 20 14 15 21 14 15 22 15 20 23 15 20 24 15 5 25 16 20 26 16 20 27 17 20 28 17 20 29 17 10 30 18 10 31 19 15 32 20 15 33 25 20 34 25 20

Notes: an overview of the amount of right answers and the corresponding earnings. Even though the earnings are not ascending there was no jealousy

Comparing generosity between direct and indirect donations in a dictator game