The possible crowding out effect of rewarding

Bachelor Thesis

Your name: Pauline Dekker

Your student number: 6285724/10001974

Specialization: Organizational Economics

Field: Incentive Pay & Crowding out

Number of credits thesis: 12 ECT

Title of your research proposal: The possible crowding out effect of rewarding

The thesis coordinator will assign a teacher to supervise your thesis. Assigned supervisor (to be filled in by thesis coordinator):

Eszter Czibor

If a teacher has already accepted to supervise your thesis, please provide the name. Name of supervisor:

Abstract

The crowding-out effect and its possible existence is a much discussed subject, this effect occurs when people are demotivated by external rewards. Several papers find evidence for the existence of this phenomenon, such as in the papers of Johannesson and Mellstrom, 2008; Gneezy and Rustichini, 2000A; Frey and Oberholzer-Gee, 1997 and several others. This thesis adds to the literature examining crowding out by the introduction of the raffle of a gift card introduced to test the crowding out hypothesis among third year Economics & Business students by comparing a treatment and control group. The students who were asked to fill out the questionnaire without mentioning compensation served as a control group while the students who received the questionnaire with the mentioning of compensation served as the treatment group. In short the results indicate no significant crowding out effect among students and no gender effects are observed either. However there were some remarkable difference between the treatment and control group: it turned out that respondents from the treatment group were more likely to already be blood donator and that they had ever collected money/raised funds for a good cause more than respondents from the control group.

1) Introduction

Several papers are written on the topic of motivation and its effect on peoples performed effort. In an influential paper the term motivation is defined: ‘To be motivated means to be moved to do something’ (Deci and Ryani, 2000A, p 54). Probably the most renowned author on the topic of motivation splits motivation into two parts, intrinsic and extrinsic motivation (Deci, 1972).

Intrinsic motivation is best defined as the urge to perform a certain task without receiving any compensation. Your personal passion and personal preferences are a part of your intrinsic motivation. Extrinsic motivation is defined by Deci (1972, p 113) as:

‘performance of an activity because it leads to external rewards’. In short it means that people perform a certain task just for the compensation they will receive to do that task. In a research on the motivation of accepting socially unwanted projects, Frey and Oberholzer-Gee (1997) state that intrinsic motivation might weaken the price effect,

meaning that when you introduce a price incentive, you might partially destroy intrinsic motivation. Intrinsic motivation in this case is partially substituted by extrinsic motivation. Instead of performing a task because you feel it is a good idea or you preferred to perform a certain task over another, it is replaced by your motivation to perform in return for

compensation. Only one paper also mentioned a third aspect of motivation, namely the image part. Ariely, Bracha and Meier (2009) state that a person’s motivation is also driven by the perceptions of others, which they call image motivation.

The fact that intrinsic motivation might be overshadowed by external motivation is known as the crowding out effect, first added to the principle-agent theory in standard neoclassical economics by Frey (1993). When you use standard economic reasoning, an increase in financial reward leads to an improvement of performance. A lot of papers have proven that this is often not the case because of the crowding out effect (Johannesson and Mellstrom, 2008; Gneezy & Rustichini, 2000A; Frey & Oberholzer-Gee, 1997 and many other papers). There is also some ambiguity in the crowding out field, Fehr & Falk (2002) still view some ambiguities in the interpretation of the data and question whether crowding out is always associated with monetary compensation.

Altogether there seems room for a Dutch research, since none of the mentioned papers has explored the crowding out effect among Dutch students. Therefore the following research questions come up: Does there exist a crowding out effect among 3th year

Economics & Business students in participating in experiments/surveys? And does this effect differ between male and female students?

The importance of finding an answer to the research question is that it will give a (small) insight of a Dutch student’s way of thinking. Based on Veldheer et al. (2012), there might be country related differences because Holland is a welfare state, for Dutch citizens it is generally accepted that individuals have to pay a considerably high amount of their

income to the government. Their income is then redistributed by the government to support the less fortunate; even though Dutch citizens have to pay a lot of taxes, there is still a lot of generosity in supporting charities. From the Dutch CBF (2012) website it follows that in the year 2012 nearly 2 billion euro’s were collected for charities. This is not particularly

distinctive for Holland but for a lot of (North)-European countries, although the most previous literature is focused on America & Asia. Therefore there could be demonstrated differences that haven’t showed up previously.

The distinction between men and women is made because Croson and Gneezy (2009) found that men and woman have different characteristics, like their degree of risk aversion and their social preferences. Women are usually more risk averse then man and they also are more sensitive to signals and small differences in experiments. The difference between the treatment and control group is exactly such a small difference that changes the

perception of the behavior of women. By using the survey the desired result is that the possible difference in crowding-out between men and women comes forward.

Personally it is interesting to see if such a phenomenon like crowding out also exists within your vicinity. The conducted research fits in the existing literature because a lot of previous articles write about a crowding out effect in all different areas among different references groups. Like the blood donation experiment from Johannesson and Mellstrom (2008), this paper also investigates whether a small fee/reward will influence people’s willingness to participate. They also mention in their paper that it would be interesting to investigate if the outcomes are different in one particular subject group of the three groups they use. Which is what this paper is investigating, whether a crowding out effect exists among Economic & Business students?

This research also has a different angle; it links the potential crowding out effect to individual characteristics obtained by the questionnaire. These individual characteristics include gender, age, general income aspects and characteristics about donating, recycling and collecting behaviour. Other relevant articles also search/find a similar effect when introducing fees or rewards. To see whether the crowding out also exists among Dutch students, the research design is as follows; two similar questionnaires are made, where one mentions the possibility of a small reward and the other doesn’t mention any compensation. Students are randomly assigned to receiving one or the other questionnaire.

In the next section the related literature and the resulting hypotheses are discussed, followed by the discussion and motivation of the questionnaire. The questionnaire is

followed by the experimental design, including the procedure, experimental treatments and information on the subjects. After that an interpretation of the result and a conclusion will follow.

2) Literature review

Besides the great interest the topic received in psychology, a large number of researches have been conducted in Economics on the subject of the possible crowding out effect of extrinsic motivation. This section gives a short overview of the previous literature and its findings.

The paper that influenced this thesis the most is that of Johannessen and Mellstrom (2008), who inquire whether blood donating is influenced by monetary rewards. They use three groups, a no payment group, a small payment group and a donation group.

Beforehand the initial fraction of women who donated was significantly higher than the fraction of men. They found that women actually are less likely to donate when they get paid for donating blood then when they do not receive any compensation. In this research the subjects were undergraduate students at Gothenburg University, but the student pool varied between the sessions.

Gneezy & Rustichini (2000B) also used students as their subject pool; they use a group of 160 students at the University of Haifa to test if monetary compensation influences the individual performance in answering questions on an IQ test. Subjects all got 60 NIS (approximately 12.50 euro) to participate in the experiment and depending on the group they either received or not a complementary compensation. The results were remarkable; the students who got paid 10 cents of NIS (approximately 2 eurocent) per question answered 5 questions less right than the students who did not receive an additional payment. The striking aspect is that this effect disappears when students get paid 1 or 3 NIS (less than half a eurocent) per correct answer; these students gave 11 more right answers than the

students paid 10 cents. Because a small payment leads to poorer performance but a high compensation leads to better results, they conclude that the amount of compensation is crucial to the existence of a so called crowding out effect.

Frey and Oberholzer-Gee (1997) also experience a considerable crowding out effect in their research about using price incentives to master support for unwanted projects. For their research purpose they use projects that are socially desirable, but which puts a burden to the direct neighborhood. They conducted 305 interviews, asking people if they were willing to accept a nuclear waste repository in their vicinity, with and without mentioning compensation in exchange for their acceptance. Their results were remarkable; without

mentioning any reward, 50.8 percent of the respondents agreed to accept the repository, but this level dropped to 24.6 percent when compensation was introduced. So they concluded that price incentives crowd out civic duty.

Another viewpoint comes from Bolle and Otto (2010), who mention that the value of an individual’s services is not exactly known, but that they act on the basis of an estimate. So when you introduce a payment it gives a signal, which might be interpreted as appropriate to the value of the activity or not. Then the consequence of the adoption of compensation only leads to crowding out when the individuals own estimation was considerably higher than the offered compensation. This results in a crowding-in effect for people with small initial beliefs and resulting in a crowding-out effect for the people with high initial beliefs. Their main conclusion is that no payment is quite different than a (very) small payment. Even though the next paper is from a later date, Bolle and Otto (2010) have laid the theoretical foundation for the research of Gneezy and Rustichini (2000A), who conducted a field study in a group of day-care centers, where parents came late even more often after introducing a late-coming fee. The study was conducted in ten private day care centers in Haifa. They came to the conclusion that introducing a fine caused a significant increase in late-coming parents and that this number of parents did not decrease after removing the fine. Similar to the previous article you can conclude that a fine puts a price a tag to an action people that people can view as appropriate.

Another social view on the subject comes from Ariely, Bracha and Meier (2009), who introduce a new variable in the crowding out theory; they claim that image motivation is decisive when you want to know whether someone is performing a certain task because he wants do good or whether that person is extrinsively motivated. In their experiment “Click for Charity”, subjects could donate by clicking two keys. Participants were randomly assigned to two groups, one group that made their choices public and one group who kept these private. According to the authors extrinsic incentives dilute the signal of pro-social behavior in public contributions but if no one is watching, defined as private contribution here, the incentives do increase pro-social behavior. So the crowding out effect only shows up in public donations, in case of private donations a crowding-in effect is experienced.

Eckel, Grossman and Johnston (2005) also introduce a new variable to the crowding out framework, namely fiscal illusion. They conducted a laboratory experiment that

examined whether third-party contributions influence private contributions to a charity. The 6

researchers gave the subjects 20$ to distribute and depending on the group they framed a tax of 2$ or 5$ of the 20$, which they indicate as forced contributions. Their results indicate that these taxes did not crowd out private giving when the source of funding wasn’t

apparent with regard to charity giving. This means that people did not donate less when they did not know where the 2$ or 5$ was going. But there was a negative effect when the source was apparent to the subjects. Which means that people do donate less when they know where the 2$ or 5$ tax is used for. The amount of taxes did not significantly influence the level of giving. From this paper follows that the way questions are framed is very important. Two other papers view the crowding out effect from a theoretical viewpoint;

Bénabou and Tirole (2004) mainly focus on the pro-social effects of actions; they call the negative effect of extrinsic motivation ‘the spoiling effect’, meaning that your motivation for good behavior is overshadowed by its compensation. And Kreps (1997) agrees on this spoiling effect of compensations, he mentions that extrinsic incentives can create distaste for the required effort.

A final remark has to made, there is also some disagreement in the literature on the topic of crowding out. Fehr & Falk (2002) come to the conclusion that there are still some ambiguities in the interpretation of the data and question whether crowding out is always associated with monetary compensation.

3) Hypothesis

Based on previous literature this research focuses on the following two hypotheses.

Hypothesis 1: intrinsic motivation might be overshadowed by extrinsic motivation, causing a so called crowding out effect.

This hypothesis is supported by a lot of previous literature (Johannesson and Mellstrom 2008; Gneezy and Rustichini 2000B; Ariely, Bracha and Meier 2009; Eckel,

Grossman and Johnston 2005; Gneezy and Rustichini 2000A; Frey and Oberholzer-Gee 1997), who all conclude that extrinsic motivation crowds out intrinsic motivation at least to some extent. This research will use a small fine to test this hypothesis, which is supported by some of the previous articles, who reveal that the size of the compensation also plays a big role. A small fine usually causes a crowding-out effect while a big compensation leads to a

crowding-in effect. The fact that women and men do not respond the same to different ways 7

of compensating is conclude by the research of Johannesson and Mellstrom (2008) who find a significant effect for women, but not for men; which leads to the second hypothesis.

Hypothesis 2: the possible crowding-out effect will be gender related.

The existing of gender differences is also supported by Gneezy and Croson (2009) who find that men and women differ in three important aspects; risk preferences, social preferences and reaction to competition. Their first finding is that men and women have different levels of risk aversion. Usually men are less risk averse that women, unless lotteries are framed as losses. Their explanation of these differences is that women have different emotional experiences of the possible outcomes and men are possibly more overconfident in their success. In terms of social preferences they find that women are not more or less socially oriented but they differ in their preferences per situation and men are more

constant in their preferences. So women are more sensitive to signals and small differences in experiments and it might be that women are more likely to be influenced by the

introduction of a payment in the incentivized questionnaire compared to the

non-incentivized questionnaire. Because woman are not more socially oriented, there shouldn’t be a difference is response rates between men and women in the control group. Finally, Croson and Gneezy(2009) find that woman behave more prudent in competitive situations; and that this could be because of what is taught to individuals or because of what is

ingrained in their culture. Because of these findings it is expected that women will respond more to the introduction of a small fee, manifesting in a decrease of female response to the compensation questionnaire.

The overall decrease in response to the compensation questionnaire is also

supported by Ariely, Bracha and Meier (2009), who find that pro-social behavior might be diluted by monetary compensation. Filling out a questionnaire for a fellow student to help complete their thesis might be seen as pro-social behavior.

A small note should me made here, because a survey will be used and the response rates are usually not very high, the effect might not show or might be insignificant, also because the individual’s actual behavior is not measured but a hypothetical situation is sketched using a non-incentivized and an incentivized questionnaire where the answers don’t have financial consequences for the respondents. For both the existing of a crowding out effect and the gender differences it holds that literature that actually finds a significant

effect is more likely to be published, which might result in a slightly distorted view on both the events.

4) Motivation of the questionnaire

In general pro-social behavior in real life in the absence of financial incentives is caused by intrinsic motivation, people do good because they feel they should do. But in the presence of financial incentives these motives might be overshadowed, causing a decrease in willingness to participate. In order to test the previous mentioned hypothesis two questionnaires were conducted, one with, the other without compensation for participation.

In the questionnaire respondents were first asked to fill in some general questions about age, gender and nationality. The next questions asked if they have a part-time job; if their parents provide some monthly contribution; and if they still live with their parents, these questions were provide was to get an insight of the student’s financial position. These questions were posed to examine if their financial position might influence the way they respond to the questions. It could be that students with a part-time job respond different because they might already have some more money to spend, which makes them less responsive to financial incentives.

Based on the blood donation experiment from Johannesson and Mellstrom (2008), the next two questions were; if the respondents are blood donators and if their decision would change when they get paid to donate blood. Based on their conclusion it’s expected that less people would donate blood. It is possible that this effect might not show up

because people only answer what they would do; it is not an observation of what they would really do if they had the choice in real-life.

The next two questions are about recycling; according to Gneezy and Rustinchini (2000B) returning (beer) bottle is seen as socially good behavior and introducing a fine has a negative effect. Because recycling beer bottles in Holland is already common the expected finding is that people will recycle a little less after withdrawing the fine. This would be in line with Gneezy and Rustinchini (2000A) other article in which they conclude that after

withdrawing a fine for late coming parents the parents actually get later more often. But also in this case the real-life action could differ from what respondents say they would do. Based on this same article the eleventh and twelfth question asked if student (normally) arrive on

time at college; and if this would change if they had to pay for the fact they were to

late(assuming they already made the decision to go to college). Because in this case the fine wasn’t introduced yet, the late coming should actually increase.

The last three questions are based on the paper about charities of Ariely, Bracha and Meier (2009); they asked whether you have ever collected for a good cause, if you would collect if you get paid a certain percentage of your collected amount; and if you would donate if no-one knew it was your contribution. In their paper they conclude that money incentives depend on visibility, so in fact the respondents are expected to answer no on the last question because no one will know they donated. They also concluded that introducing a fine decreases the signal of pro-social acts, so the expected answer on the question: ‘Would you collect if you get a percentage of the money you collect? Paid by the government, not by the good cause?’ should be no, because now you get paid and your act is not seen as social anymore. Of course all these questions are hypothetical and the real life actions people will undertake would probably differ from what they answer.

All the questions were the same for the two compensation groups; the only

difference was the mentioning of the raffle of a 10 euro gift card among respondents in the introduction of the questionnaire. All these answers are not only used to test the findings of other papers but also to check whether respondents in the two treatment groups are different along these dimensions.

5) Experimental design 5.1 Procedure

This research was carried out among students of the University of Amsterdam, situated in the capital city of Holland. The University of Amsterdam is one of the two universities in Amsterdam and of the twenty-one universities of Holland, with almost 30.000 students it is one of the largest general research universities in Europe. With the assistance of the

secretary of the Faculty of Economics and Business, students were emailed with the question if they could help completing a bachelor thesis by filling out a questionnaire. The

questionnaire contained general questions about age, nationality and financial situation and

some specific questions about their view on donating, volunteering and recycling. See appendix A for the questionnaire.

5.2 Treatment

Two types questionnaires have been sent, one to each treatment group; the first questionnaire mentioned nothing about compensation and the second questionnaire mentioned the raffle of a 10 euro bol.com gift voucher among respondents. Aside from the introduction text before the start of the questionnaire and the option to leave your email address if respondents wanted to participate in the raffle of the gift voucher, the two surveys were identical. All the questions were the same for both the control and the treatment group. The actual questionnaire can be found in appendix A.

5.3 Subjects

All subjects were undergraduate Economics and Business students with 120 or more ECTS; this is the Dutch term for credits. Subjects were recruited by email, addresses were provided by the secretary of the faculty of Economics and Business for this research purpose. Only employees of the secretary of the UvA have access to these addresses. Some were personal email addresses and others were university mails. Students were randomly assigned two one of the two treatment groups, their student numbers were sorted ascending in value and the first half of the list of student numbers were sent the non-incentivized questionnaire and the second half was sent the incentivized questionnaire.

The first questionnaire, the one without mentioning compensation, was send to 437 students on Monday morning between. It was sent on a Monday because in the weekends students might not check their university mail. Thirty-two addresses found not to exist, they were returned with a delivery status notification failure. One email was responded with an automatic email that the concerned person was on vacation. Because his date of return was after the closing of the questionnaire the address was left out of the sample. This results in a control group consisting of 405 students. The second questionnaire, the one with

compensation, was send to 438 students on the same day and time. Unfortunately there were a lot of non-existing emails in this group; sixty-five emails were returned with a delivery status notification failure, resulting in a treatment group of 373 students. For both the

questionnaires a reminder was send on the following Wednesday of the first email around 11

noon to also reach the students who read their mail around lunch and to enlarge the response rates.

6) Descriptive statistics 6.1 General findings

In this section the general results of the questionnaires are reviewed. In the following text, respondents should be interpreted as the students who actually responded to one of the questionnaires. In the treatment group 373 emails were sent and 55 students responded and in the control group 405 emails were sent and 77 students responded, which totals to 132 filled out questionnaires. The general information like age, gender and nationality is only available for the respondents, unfortunately not for everyone who was sent an email to. This means there can only be made comparisons between the two groups of students that

responded to the questionnaire and no comparisons between the students who did and did not fill out the questionnaire.

From the general questions it followed that there were more male than female respondents, 59.09% percent of the students who filled out the question were men. Women are under-represented among the Economics & Business students, so this might explain why there are less female respondents in general. The average age was 22.33 years and the most of the questioned has a Dutch nationality, namely 91.67%. About three quarters, 66.67%, of the respondents have a part-time job and 64.39% receives a monthly contribution from its parents to pay for their rent, college tuition, groceries or other monthly payments.

6.2 Literature-related findings

In this section the differences in answers on the literature related questions are reviewed. Based on the blood donation experiment from Johannesson and Mellstrom (2008), the next two questions were about blood donating. Based on their conclusion it is expected that less people would donate blood when a compensation is introduced. From this research it

appears that 6.06% of all respondents is already a blood donator. From the 8 students that initially donated, 7 would also donate when compensation is introduced and only 1 student wouldn’t donate anymore. So of the 39 students who would change its blood donating decision, there was 1 student who wouldn’t donate anymore after the introduction of compensation and 38 students, who didn’t initially donate, would start donating. These findings are the opposite of the expected findings.

The questions that followed were two questions about recycling beer bottles. According to Gneezy and Rustichini (2000B) returning (beer) bottle is seen as socially good behavior and introducing a fine has a negative effect. Because recycling beer bottles in Holland is already common the expected finding is that people will recycle a little less after withdrawing the fine. 81.82%, 108 of the 132 students who filled out one of the

questionnaires, indicated that they were already collecting beer bottles. Of the 108 people who initially recycled beer bottles, 39.81% indicated that they wouldn’t recycle anymore after the removal of the fine. Of the initial 28 non-recycling students, 9 would start recycling because of the removal of the fine. In total the amount of recycling students decreases from 108 out of 132 to 74 out of 132, which is underpinned by the previous research.

The next questions about being on time for college were based on Gneezy and Rustinchini (2000A) paper about daycares, because in this case the fine wasn’t introduced yet, the late coming should actually increase. 112 students, 84.84%, answered that they were usually on time for college, 41 out of 132 claimed that their previous answer would change by the introduction of a late coming fee. Of these 41 students that would change their decision, 24 students, 58.54% would start coming late after the introduction of the fine and 41.46% would start arriving on time while they were initially usually too late. This finding is remarkable, after the introduction of the fine, more than half of the subjects that would change their decision would start arriving to late after the introduction of a fine. This finding is in line with Gneezy and Rustinchinis (2000A) paper because initially 20 students answered they were usually too late for college and this amount increases to 27 students after the introduction of the fine.

The last three questions are based on the paper about charities of Ariely, Bracha and Meier (2009) about collecting for a good cause, in their paper they conclude that money incentives depend on visibility, so in fact the respondents are expected to answer that they wouldn’t donate anonymously. From their paper respondents are also expected not to be

influenced by a percentage of the collected money because collecting should be seen as a social responsibility. The results were the following: 64.39% of the students answered that they have already collected money/raised funds for a good cause before, 43.94% said they would change their initial answer by the introduction of a percentage of the raised money as compensation. Of these students 50 claimed that they would not collect anymore after the introduction while they did before, only 8 students who did not collect/raised funds before would start doing this after the introduction of the fine.

These findings prove that students are definitely aware of the social aspect of acting for a good cause, the percentage collectors decrease from 64.39% to 32.58%, a decrease of almost 50%. To the question if students would donate anonymously, strangely enough, 85.61% of the surveyed people answered yes. This finding is definitely not in line with the findings of Ariely, Bracha and Meier (2009) who predicted that no one would like to donate anonymously because then no one would know that you did something good. Perhaps these students are more social responsible than the average subject pool the authors had in mind. From all previous questions it can be concluded that even by only using a

questionnaire with hypothetical situations, no real life experiments, most of the previous predictions can be shown to be proved. Except from the blood donating and donating anonymously predictions, all other predictions are supported to be true in this research. 7) Statistical analysis

7.1 General findings

In this section the difference in response to the different questionnaires is used to review the existing literature. For all the averages and number of respondents, see appendix B. For the clearance, in the treatment group 373 emails were sent and 55 students responded and in the control group 405 emails were sent and 77 students responded. All between group comparisons are in percentages and all with the corresponding p-value calculated with the two sample proportion test. The p-value is usually defined as the largest value you would reject the null hypothesis under, one often rejects the null hypothesis when the p-value is less than a predetermined level. In the following text not statistically significant means, not significant at the 90%, 95% or 99%-level. Rejecting the null hypothesis means that there is no

relationship between two measured phenomena or that a potential treatment has no effect. In the following, respondents from the incentivized and the non-incentivized treatment are compared in terms of their characteristics. From the general questions it followed that there were not significantly more male responders than female, with a p-value of 0.209 the difference in gender was not significant in both the treatment and control group. With a p-value of 0.209 the gender difference is not significantly different between both groups. For the treatment group the percentage of males was slightly lower (52.73%) than the percentage in the control group (63,64%), both with an average age of the students of approximately 22 years. For both the groups most students were Dutch citizen, 94.45% and 89.61%, respectively for the treatment and control group, with a p-value of 0.312 there was no significant difference.

In the control group the percentage of students with a part-time job was quite a bit lower than in the treatment group, 61.82% compared to 71.43%. This difference however is not significant with a p-value of 0.213. Between 61-66% receives a monthly contributions from its parents, this variable has a p-value of 0.601 which is not significant. And the biggest part of the surveyed individuals do not live with their parents anymore, with a p-value of 0.642 the 36.36% of the treated group and 32.47% of the control group that live at their parents do not significantly differ.

7.2 Answering the research question

In this section hypothesis I and II are reviewed. The first hypothesis is that intrinsic

motivation might be overshadowed by extrinsic motivation, causing a so called crowding out effect. This would imply a lower response rate in the treatment group than in the control group. From the table at the top of the statistics part of the research it follows that for the control group there is a response rate of 19.01% compared to 14.75% for the treatment group. This difference is not significant at the 90, 95 or 90% level, with a P-value of 0.113 it was slightly above the 90% level, which follows from the two sample test of proportions. The distinct values for the different level of significances can be found in appendix D. From the above it is concluded that the first hypothesis is not supported and thereby that the response rates between the groups do not significantly differ from each other.

The second hypothesis is, based on the article of Croson and Gneezy (2009), that this 15

possible crowding-out effect will be gender related, which would imply a higher percentage of women in the treatment group compared to the percentage of women in the control group. The perfect test would be to compare the response rates by groups, between all students who received an email, for men and women separately but unfortunately the data to do this is not available, only the gender of the students who actually responded is

available. Duo to the absence of this data a strong assumption is made, namely that the randomization process was successful and the gender ratios in the complete treatment and control group are not significantly different from each other.

Among the respondents in the control group 28 out of 77 were female students; a percentage of 36.36%, among the respondents in the treatment group 26 out of 55 were female students; a percentage of 40.91%. One can immediately see that the percentage women is higher in the treatment group, which is the opposite of the expected effect. Therefore the effect turned out not to be significant for the crowding out effect on all three levels of significance, the P-value of the two-sample proportion test was 0.209. The different values for gender can also be found in appendix C and D. From these tests one can conclude that there is no significant crowding-out effect but from the response rates you can

definitely notice the difference between the questionnaire with and without mentioning compensation. Summarizing, the gender hypothesis is also not supported, the gender effect turned out not to exist in this research. So there were not significantly more women who filled out the questionnaire for the control group than women who filled out the

questionnaire for the treatment group, this could be influenced by the fact that Economics and Business has a surplus of male students.

7.3 Gender-related differences

To find out if there were any interesting differences between men and women, several two-sample proportion tests were carried out. The difference in blood donators between men and women is not significant with a p-value of 0.84, but it is remarkably that more men were blood donators despite of the prediction of Johannesson and Mellstrom (2008) who found that women are in general more likely to donate blood. From the comparisons it also turned out that women are a little more likely not to recycle, 20.51% males and 22.22% female answered not to recycle, with a p-value of 0.813 this is insignificant. This finding is in line

with the paper of Croson and Gneezy (2009), who stated that women are not more socially adequate than men. On the contrary men are slightly more likely not to arrive on time for college, 19.23% male and 12.96% female initially stated they were usually too late for college, with a p-value of 0.342 this difference is also insignificant. Another difference that strikes is the percentage of students who did not collect money/raised funds before, male students were more likely not to have done this before, 39.74% compared to 27.77% female students, with a p-value of 0.156 this is marginally significant. The last difference between men and women that showed up was that male students are less likely to donate

anonymously, 17.95% of the male students answered not to be willing to donate

anonymous, while only 9.26% of the female students were unwilling to do this, with a p-value of 0.162 this difference however was not statistically significant.

7.4 Literature-related findings

In this section the literature related findings are compared between the treatment and the control group. The first two literature related questions were about blood donating. From this research it appears that in the control group only 2.60% is a blood donator and 32.47% would change his previous answer if a compensation is introduced, for the treatment group these were respectively 10.91% and 25.45%. With a p-value of 0.048 there is a significant difference at the 95-level for the amount of blood donator between both groups. From which you can conclude that in the treatment group there are significantly more initial blood donators than in the control group, the difference between both groups after the

introduction of compensation was not significant with a p-value of 0.384. The expected decrease in donating behavior after introducing a fine is not observed here.

The questions about blood donation were followed by two questions about recycling beer bottles. Among the control respondents 83.12% was already recycling compared to 80% in the treatment group, this difference is not significant with a p-value of 0.647. After the removal of the fine 41.56% of the control and 36.36% of the treatment group changes its decision to recycle beer bottles, which is also not statistically significant with a p-value of 0.547. The expected (small) decrease in recycling behavior is observed because most of the students who would change their decision were students who initially recycled and would stop after the removal of the fine.

The next questions were about being on time for college. In this research student in the control group 84.42% said they were always on time, compared to 85.45% in the treatment group with a p-value of 0.870. Only 27.27% in the first group would change his decision if a fine will be introduced and 36.36% in the second group, this doesn’t differ significantly with a p-value of 0.266. The prediction by Gneezy and Rustinchini (2000A) did turn out to be true in the control group because initially only 14 students came too late for college and after the introduction 17 students answered that they would come too late for college. And also for the treatment group the prediction comes true, first 8 students answered they usually are too late for college and after the introduction of a fine 14 students responded they would come too late.

The last three questions are based on the paper about charities of Ariely, Bracha and Meier (2009) about collecting for a good cause. The results of these questionnaires tell that 58.44% of the control group and 72.72% of the treatment group already collect but that 42.86% and 45.45% would change their decision to collect when a percentage on top of the money collected is used as compensation. The between group comparison of percentage of collectors is significant at the 90%-level, meaning that in the control group significantly more students have already collected money/raised funds for a good cause before. The difference between 42.86% and 45.45% is far from significant with a p-value of 0.767. The results of this research are in line with Ariel, Bracha and Meiers (2009) paper because in the treatment group 40 students answered they collected/raised money before and after the introduction of the fee this amount decreases to only 17 students. Also in the control group the

predictions hold, before the introduction 45 answered yes and after the fine introduction just 17 students were willing to raise money for a good cause. This shows that students, despite their limited disposable income, are definitely aware of the pro-social aspect of serving for a good cause.

Last but not least the results for the last question were quite remarkable, 87.01% of the control and 83.63% of the treatment group indicated that they would donate

anonymously, which was definitely not expected based on the previous literature.

From all previous questions it can be concluded that the answers from the students from the treatment and control group do not differ substantially, only the blood donating and the collecting/raising money variables turned out to be significantly different between the two groups. This seems to point in the direction that the treatment group students are

perhaps more social responsible, even though you would expect that the more social

students would be in the non-compensation group. All though it is remarkable that almost all previous predictions from existing literature can be proved right by this research, despite the use of hypothetical situations. Of course one must keep in mind that all these questions are hypothetical and the real life actions people will undertake could probably differ from what the students have answered.

8) Statistic expansion

In the previous sections it turned out that the initially stated hypotheses were not proven right by this research. However there were two remarkable observations: it turned out that respondents from the treatment group were more likely to donate blood and more likely to have collected money/raised funds before. At first sight this seems contradictory, one would expect that the more social students would be in the control group and the less social students in the treatment group. Therefore this section further examines this finding. To see if the effect also shows up when controlling for other variables, several probit regressions were run. First the blood donating variable was further examined, three

regressions were run. In the first regression the variable treatment, gender, part-time job and monthly contribution were the explanatory variables, this resulted in a significant p-value of 0.047 for the treatment variable, the p-p-value was even smaller when treatment, gender, nationality and living situation were used as explanatory variables, namely 0.037. For this same model an OLS regression was run, with a value of 0.092 for the Treatment variable there is a small positive relationship between the variables

BloodddonaterY and Treatment. This means that when the BlooddonaterY value increases, the coefficient for treatment also increases. Compared to Gender (0.009) and HomeY(-0.016) treatment is a larger effect, but the independent variable DcitizenY(-0.14) explains a larger fraction of the model. The only variable that is significant in combination with the treatment value is the nationality variable. The treatment variable says significant when controlling for more than one other variable, concluding that treatment respondents are significantly more likely to be blood donator.

For the variable CollectedY, the dummy variable capturing whether respondents ever collected for charity, it turned out treatment was only significant when controlling for certain

variables, the first regression with treatment, gender, part-time job and monthly

contribution as variables resulted in a p-value of 0.052, which is significant at the 10%-level. However when controlling for treatment, gender, nationality and living situation, treatment is no longer significant with a p-value of 0.143. So it seems that the variables with a

monetary implication influence the ColletedY variable. When running regressions it turns out that the variable treatment is only significant in combination with the PTJobY variable. From which one can conclude that collecting money/raising fund is correlated with having a part-time job, and only when students have a part-part-time job, treatment is significant as an explanatory variable for CollectedY. For the complete regression data see appendix E. For the model with Treatment, Gender, PTJobY and MonthlyCY as explanatory variables, an OLS model was run to get an insight in the size effect. The coefficient for Treatment here is 0.166, which indicates a positive relationship between the variables CollectedY and Treatment. This means that when CollectedY increases, the Treatment variable will also increase. Compared to the Gender (-0.087) and MontlyCY (0.046)

explanatory variables, Treatment is the second most informative variable. The coefficient for PTJobY, 0.269, is the highest and therefore the most informative variable in this regression. Both the OLS-regressions can be found in appendix F.

These findings could be explained by Gneezy, Imas and Madarász (2012), who conducted an experiment among students testing whether guilt plays a big role in a

deception game experiment. They found that a norm violation creates a temporary feeling of guilt, which diminishes over time. This feeling of guilt creates a propensity for more pro-social behavior, which disappears with the converge of time, they call this the conscience accounting effect. The fact that more non-donators and non-collecting are in the control group might be explained by this effect, they try to compensate their less pro-social behavior in their everyday life by helping a fellow student by filling out a questionnaire without

compensation. This means that guilt increases the importance of improving their following social behavior, which implies that pro-social behavior is not a fixed trait but instead people may try to compensate selfish behavior in certain domains by acting nicely to others.

9) Conclusion

The results of this paper indicate that there is no significant crowding out effect among third year Economics and Business students, even though there definitely is a decrease in

response rate observable between the no-compensation and compensation questionnaire. The expected gender differenced turned out not to show up among this subject pool, male and female students had quite similar responses to the questions in the survey and neither is one group more likely to respond to the treatment.

However some unexpected differences between the students from the control- and treatment group came forward, it appeared that the treatment students were more likely to be a blood donator and they were also significantly more likely to have ever collected

money/raised funds for a good cause. Because students have a different spending patterns then the average person; they usually have a low income, they are expected to respond more to monetary incentives, which might have biased the results of this crowding out thesis. Because response rates were quite low, 19.01% and 14.74% the actual respondents numbers were possibly not large enough to experience such an effect.

For further research it might be interesting to explore the effects of a larger sample size. By limited resources it was not possible to actually reward all the students who would fill out the treatment survey, this might also be interesting to further assign the differences among students. For the gender related hypothesis it would be great to have a subject pool with information about all (students), not just the ones who respond, this might further broaden the possible gender related difference in crowding out.

Bibliography

Ariely, D., Bracha, A. and Meier,S. (2009). "Doing Good or Doing Well? Image Motivation and Monetary Incentives in Behaving Prosocially." American Economic Review, 99(1): 544-555. Bénabou, R. and Tirole, J. (2004). "Incentives and Prosocial Behavior." American Economic Review, 96(5): 1652-1678.

Bolle, F. & Otto, P.E. (2010). “A Price Is a Signal: on Intrinsic Motivation, Crowding-out, and Crowding-in.” Kyklos, 63(1): 9-22.

Centraal Bureau Fondsenwerving (2012). “Overzicht baten en lasten”. http://www.cbf.nl/ Cijfers/overzicht-baten-lasten.php

Croson, R., and Gneezy, U. (2009). "Gender Differences in Preferences." Journal of Economic Literature, 47(2): 448-474.

Deci, E.L. (1972). “Intrinsic motivation, extrinsic reinforcement and inequity”. Journal of Personality and Social Psychology, 22(1): 113-120.

Deci, E.L., and Ryan, R.M. (2000A). “Intrinsic and Extrinsic Motivations: Classic Definitions and New Directions”. Contemporary Educational Psychology, 25(1): 54-67.

Desi, E.L., and Ryan, R.M. (2000B). “Self-Determination Theory and the Facilitation of

Intrinsic Motivation,Social Development, and Well-Being”. The American Psychologist, 55(1): 68-78.

Eckel, C., Grossman, P.J., and Johnston, R.M. (2005). “An experimental test of the crowding out hypothesis”. Journal of public economics, 89(8): 1543-1560.

Falk, A., Fehr, E. (2002). “Psychological Foundations of Incentives.” European Economic Review, 46(4–5): 687–724.

Frey, B.S. (1993). "Motivation as a Limit to Pricing." Journal of Economic Psychology, 14(4): 635-664.

Frey, B. S., and Oberholzer-Gee, F. (1997). "The Cost of Price Incentives: An Empirical Analysis of Motivation Crowding-Out." American Economic Review, 87(4): 746-755.

Gneezy, U., Imas, A., and Madarász, K. (2012). “Conscience Accounting: Emotional Dynamics and Social Behavior”. Suntory and Toyota International Centres for Economics and Related Disciplines, LSE, 563: 1-37.

Gneezy, U., and Rustichini, A. (2000A). "A Fine Is a Price." Journal of Legal Studies, 29: 1-17. Gneezy, U., and Rustichini, A. (2000B). "Pay Enough or Don't Pay at All." Quarterly Journal of Economics, 115(3): 791-810.

Johannesson, M., and Mellstrom, C. (2008). “Crowding out in Blood Donation: Was Titmuss Right?” Journal of the European Economic Association, 6(4): 845–63.

Kreps, D.M. (1997). “Intrinsic motivation and extrinsic incentives”. American Economic Review, 87(2): 359-364.

Veldheer, V., Jonker, J., Noije, L and Vrooman, C (2012). “Een beroep op de burger” Sociaal en Cultureel Rapport 2012.1: 1-30

Appendix A: Questionnaires Above Questionnaire:

For my Economics & Business bachelor thesis related to decision making I’m looking for respondents to fill out my survey. It will only take about 1 or 2 minutes and it would really help me to complete my thesis! It’s not necessary to forward my questionnaire; just filling out would be really appreciated. Among the participating students, a gift voucher of bol.com of 10 euro is raffled. Thanks in advance,

Kind regards, Pauline

Question 1: What’s your gender? *Male

*Female

Question 2: How old are you? *18 *19 *20 *21 *22 *23 *24 23

*25 *26 *27 *28 *29 *30

Question 3: Are you a Dutch citizen? *Yes

*No

Question 4: Do you have a part-time job? *Yes

*No

Question 5: Do you get a monthly contribution from your parents to pay for your rent, college tuition, groceries or other monthly payments? *Yes

*No

Question 6: Are you still living at your parents or other family members? *Yes

*No

Question 7: Are you a blood donator? *Yes

*No

Question 8: Would your decision change if you get paid for blood donation? *Yes

*No

Question 9: Do you recycle beer bottles?

*Yes *No

Question 10: Would your previous answer change if you don’t get a deposit? *Yes

*No

Question 11: Are you usually on time for tutorials/lectures? *Yes

*No

Question 12: Would your decision change if you had to pay a small fee for coming too late? (Assuming you already decide to go to college)

*Yes *No

Question 13: Have you ever raised funds/collected money for a good cause? *Yes

*No

Question 14: Would you collect if you get a percentage of the money you collect? Paid by the government, not by the good cause?

*Yes *No

Question 15: Would you donate for a good cause if no one would know it was your donation?

*Yes *No

The second questionnaire only differed in its story above the questionnaire. In addition it mentioned the sentence: ‘Among the participating students, a gift voucher of bol.com of 10 euro is raffled’. All questions in the questionnaire were exactly similar.

Links to the questionnaires:

Without noticing compensation: http://www.thesistools.com/web/?id=354981 With a compensation mentioned: http://www.thesistools.com/web/?id=355121 Appendix B: Averages

Appendix C: Pie charts Total:

Gender

No-Compensation

Compensation

Appendix D: Two-sample proportion tests Response rates:

Gender

Gender

Gender:

Appendix E: Probit regressions BlooddonaterY

CollectedY

Appendix F: OLS Regressions BlooddonaterY

CollectedY