University of Amsterdam Department of Economics
Endowment origin in public good experiments
Renza Dosker 0420182
Master Thesis; field Public Economics August 19th 2009
First supervisor: Dr. K. Abbink
2
Economists have adapted their view of human decision making radically after findings from economic experiments. Because of this, it is of upper importance that the
fundamentals of these economic experiments are as thoroughly studied as the field that these experiments explain. To contribute to this, recently some economists shifted their focus to the endowments which individuals use to play the respective economic
experiments with. The endowments used during experiments in the lab are sudden and kind of out-of-the blue and this may influence the way individuals behave regarding their money and others. To contribute to the research of this issue of unearned endowments in economic experiments, this thesis describes an experiment that focuses on the effect of the provision of earned endowments on other-regarding behaviour in a one-shot standard public good game.
3
Acknowledgements
I would like to take this opportunity to thank those who made this thesis possible. Firstly, my supervisor Dr. K. Abbink who offered his help, knowledge and guidance during the time of my thesis. I would also like to thank Dhr. Dr. T. Offerman for his passionate way of teaching in game theory which got me interested in this field of economics.
My special gratitude goes to Sietske Dosker for her knowledge and accuracy, for her cheering words when needed and for her energy when mine had a day off and to Jan and Riktia Dosker who supported me during the entire period of my school carrier. Finally I would like to thank Kim Janszen for all the time that we spend at the university studying but mostly I want to thank her for the much needed breaks at our favourite coffee place.
Renza Dosker
4
Table of contents
ACKNOWLEDGEMENTS ... 3
1 INTRODUCTION ... 6
2 BACKGROUND EXPERIMENTAL ECONOMICS ... 8
2.1GENERAL FINDINGS ... 8
2.2PUBLIC GOOD EXPERIMENTS ... 9
2.2.1 Main findings ... 9
2.2.2 Repetition ... 10
2.2.3 Communication ... 10
2.2.4 Group size... 10
2.2.5 Marginal return per capita... 11
2.2.6 Subject pool ... 11
3 STARTING CAPITAL ... 12
3.1CHARACTERISTICS OF STARTING CAPITAL ... 12
3.1.1 Starting capital as heterogeneous endowment ... 13
3.1.2 Starting capital as unearned endowment ... 13
3.1.3 Starting capital as unexpected endowment ... 15
3.2EXTENSION ON CONDUCTED RESEARCH ... 16
4 SET-UP EXPERIMENT ... 18
4.1GENERAL SET-UP... 18
4.2TREATMENTS ... 21
4.3SUMMARY SET-UP EXPERIMENT ... 23
5 DATA ANALYSIS ... 24
5.1DATA DESCRIPTION ... 24
5.1.1 Contributions made to the public good ... 24
5.1.2 Contributions ‘received’ ... 25
5.2THE MODEL ... 26
5.3EXPECTATIONS ... 27
5.3.1 Own endowment ... 27
5.3.2 Total endowment group members ... 29
5.3.3 Standard deviation endowment group members ... 29
5.3.4 Treatment dummies ... 29
5.3.5 Control variables ... 30
5.4SIGNIFICANCE VARIABLES ... 30
5.5TESTS FOR MISSPECIFICATION ... 31
5.5.1 Goodness-of-fit test... 31
5.5.2 Heteroscedasticity ... 32
5.5.3 Autocorrelation ... 32
6 CONCLUDING REMARKS ... 33
5
APPENDICES
I. INSTRUCTIONS PHASE ONE ... 39
I.ISCREEN 1; GENERAL INSTRUCTION FORM PHASE ONE ... 39
I.IISCREEN 2; BASELINE TREATMENT ... 39
I.IIISCREEN 2; WORK TREATMENT ... 39
I.IVSCREEN 2; TALENT IQ TREATMENT ... 40
I.VSCREEN 2; TALENT MUSIC TREATMENT ... 40
II INSTRUCTIONS PHASE TWO ... 41
II.ISCREEN 1; GENERAL INSTRUCTION FORM PHASE TWO ... 41
II.IISCREEN 2; BASELINE TREATMENT ... 41
II.IIISCREEN 2; EARNINGS TREATMENTS ... 41
II.IVSCREEN 2 CONTINUED; ALL TREATMENTS ... 42
II.VSCREEN 3; QUESTIONNAIRE ALL TREATMENTS ... 43
II.VISCREEN 4; EXPECTATIONS ABOUT OWN ENDOWMENT ... 43
II.VIISCREEN 5; REVEAL ENDOWMENT TO PARTICIPANT ... 44
II.VIIISCREEN 6; DECISION TABLE ALL TREATMENTS ... 44
II.IXSCREEN 7; QUESTIONNAIRE EXPECTATIONS... 44
III INSTRUCTIONS END OF EXPERIMENT ... 46
III.ISCREEN 1; ENDOWMENTS ... 46
III.IISCREEN 2; CONTRIBUTIONS ... 46
III.IIISCREEN 3; END ... 47
III.IVSCREEN 4; GENERAL QUESTIONNAIRE ... 47
6
1 Introduction
An endless number of economic experiments have been conducted. From prisoners’ dilemma to dictator game and from individual decision making games to large group experiments. Economists have adapted their view of human decision making radically after findings from these economic experiments. It can be said that one of the most influential, in light of the idea of the economic man, and most accepted ideas, is that individuals exhibit other-regarding behaviour in a quite extensive way.
The focus of experiments has been on explaining and quantifying this other-regarding behaviour. For example, in public good experiments participants tend to put forty to fifty percent of their endowment voluntary into the public good which are mainly intended contributions which cannot be explained by confusion. So individuals not only value their own monetary pay-off but also their relative pay-off and exhibit a notion of fairness. But although it seems reasonable to extrapolate this to the real world, everybody knows his share of ‘other regarding people’, the question remains if decisions in the lab really correspond to decisions making in the real world or that there are important differences between inside and outside the lab.
To address this, recently some of the focus has shifted to the endowments which are given to individuals to play the respective economic experiments with. The
endowments normally used during experiments in the lab are sudden and kind of out-of-the blue and this may influence out-of-the way individuals behave regarding out-of-their money and others. The endowment that is provided in economic experiments is in three ways different from the endowment which individuals face in real life. The given endowment is
homogeneous, unearned and unexpected.
This thesis will focus on this second characteristic of endowments that are provided in experiments. The fact that individuals do not have to put any time, effort or human capital in acquiring the endowment may cause the individuals to value this endowment in a different way and change their notion of fairness.
To contribute to the research on this issue of unearned endowments in economic experiments, this thesis will focus on the effect of the provision of earned endowments on other-regarding behaviour in a one-shot standard public good game. As endowments are
7
earned, players have heterogeneous endowments. However because this is not the main characteristic of interest, instead of totally heterogeneous endowments only three levels of endowments are used. As many researchers have focused on the dictator game, this thesis will focus on public good experiments. The main question is: “Do property rights1 alter the
behaviour of participants in public good experiments?” But as people earn their endowment in different ways in real life it is also examined if the way in which these property rights are earned, has an effect on the behaviour displayed by participants. This thesis has the following set-up; in the second chapter, the background of experimental economics is discussed. In this chapter, the main findings from experimental and in specific public economics are discussed. In chapter three, the main findings of experimental economics with regard to the effect of endowment origin are discussed. In this chapter one can find the research that is already done in this field and more attention will be paid to research that can still be done. In chapter four, the experimental set-up of the experiment is given. The expectations about the results of this experiment are discussed in chapter five, as well as the way in which the collected data could be reviewed. Finally, in chapter six a short concluding remark is given.
1 The term ‘property rights’ is defined as entitlement obtained by performing a (working) task (Fahr and
8
2 Background experimental economics
2.1 General findings
Since researcher have been intrigued by the possibilities of laboratory experiments to give more inside in economics in the 1940s a lot of experiments have been done (Davis and Holt, 1993). One can make inferences about behaviour and preferences of people by looking at the several different types of experimental games that have been conducted. To give a more comprehensive overview of the insights in behaviour that are found in
previous experiments, this paragraph will briefly discuss the main findings from
experimental games in general and of the extreme case of the dictator game. After this, the focus will shift to findings from public goods experiments.
The two main general findings from laboratory experiments are that participants are not just concerned with their own monetary behaviour and that the behaviour of the participants in the experiment is not static. Although one can assume from economic theory that people always prefer more money than less (Bolton, Katok and Zwick, 1998, p. 270) it appears from experimental economics that also relative pay-offs, monetary pay-offs of others and notions of fairness may play a role in observed behaviour. Even when all incentives to display social behaviour, as in the one-shot anonymous dictator game, are eliminated and economic theory of self-interested people predicts that rational dictators will offer nothing to others, still some other regarding behaviour can be observed2 (Cherry
and Shogren, 2008, p. 69).
With regard to the evolving behaviour of the participants, one can assume that behaviour is adjusted when participants gain experience and learn about the games they play and the behaviour of other participants (Ledyard, 1995). Participants may get insight in the preferences of other players and can use this to adjust their strategy, but also they might just learn to understand the rules of the game better so confusion and mistakes are reduced (Houser and Kurzban, 2002, p. 1067).
2
However another finding is that when faced with the choice of either giving more or less than they would when they could freely choose, dictators choose less. This shows that although people are more generous than originally predicted, they are no saints either (Bolton, Katok and Zwick, 1998, p. 285).
9 2.2 Public good experiments
As the main focus of this thesis is on public good experiments, it is also of importance what the findings in this specific case are with regard to social behaviour. A public good is characterized by non-excludability and non-rivalry. Non-excludability means that it is either impossible or undesirable to exclude consumers so benefits from cooperation are shared by everyone, even those who defect (Buckley and Croson, 2006, p. 936 and Sell and Wilson, 1991, pp. 107-108). Non-rivalry means that the use of the good by an additional consumer has no effect on the utilization of the good by other consumers (Buckley and Croson, 2006, p. 936).
2.2.1 Main findings
One robust finding in public good experiments is that when confronted with a Voluntary Contribution Mechanism (VCM), a significant proportion of participants contribute towards a public good while the individually pay-off-maximizing choice would be to free ride on others’ contribution. This incentive to free ride follows from the non-excludability characteristic of the public good (Sell and Wilson, 1991, pp. 107-108) and the fact that the marginal rate per capita (MRPC), is smaller than zero. A marginal return smaller than zero means contributing to the public has a lower utility than keeping the money in the private account when assuming pure self-interest (Buckley, Corson, 2006, p. 937).
Contributions to the public good in one-shot3 and finitely repeated games are
generally halfway between the Pareto-efficient level (full contribution) and the free riding level (zero contribution). Findings on the specific number differ somewhat but generally lay between forty and sixty percent of the original endowment. On average players display conditional cooperative behaviour. This means that the contribution is increasing in the total contribution of the other group members.
Several explanations have been given for the found contribution rates. Part of the contributions can be explained by mistakes or confusion, but a large part is also intended. Reasons for these intended contributions can be altruism or the warm-glow feeling from giving. Altruism is related to pure unselfishness while the warm-glow feeling means that you derive utility from giving for example because you prefer to maximize the aggregate income level, an impure form of unselfishness.
10
Since researchers have found that people do not behave in a purely selfish and rational way, many experiments have been conducted to examine if this finding is robust for different settings. Among others it has been investigated if communication, culture, group size, individual characteristics, marginal returns and repetition alter the finding of social behaviour.
2.2.2 Repetition
The influence of repetition on the contribution rate can be seen as the second major finding from conducted economic experiment (Ledyard, 1995). The combination of the presence of free-riders and conditional co-operators with a self serving bias, explains the decline of contributions observed in public good games in the following way. Participants react on average conditionally cooperatively on other participants’ contributions with a bias in the selfish direction. Therefore positive but deteriorating contributions will be observed as conditional co-operators adjust their contributions in response to each other and the presence of free riders (Fischbacher, Gächter and Fehr, 2001, p. 403).
2.2.3 Communication
It appears that face-to-face communication improves the rate of contribution. This can be explained by looking at human evolution and socialization. During evolution, face to face communication was the only way of communication available. Besides this we are
socialized in small groups and are used to face to face communication. So both human evolution and socialization have taught us to trust on those with whom we communicate face-to-face (Brosig, Ockenfles and Weiman, 2001, p. 16).
2.2.4 Group size
Several experiments have shown that the finding that people do not behave purely selfishly is robust for larger groups as well. According to Isaac, Walker and Williams (1994, p. 1) participants deviate from free riding even more often in large groups as they do in smaller groups. Public goods appear to be provided even more efficiently in large groups (forty to hundred participants). Other researchers indicate that there is no effect found from group size (For example Marwell and Ames, 1979). Although exact numbers remain difficult to find, there is a consensus about the fact that larger groups will not eliminate social behaviour. See for a more comprehensive overview Public goods: a survey of experimental research (Ledyard, 1995, pp. 151- 155).
11
2.2.5 Marginal return per capita
To study the robustness of findings in public good experiments, also the effects of different marginal rate of returns to investments or marginal return per capita (MPCR) in the public goods have been examined. It appeared that the intuitive idea that contributions to the public good increase with increased marginal rates of returns is indeed true (Ledyard, 1995). The effect of different MPCR is closely related to group size. When a public good is imperfectly non-rival it means that the utility for one person can slightly decrease when group size increased. Isaac and Williams (1988) investigated the effect of these two variables, MPCR and group size and found that larger groups have more difficulty with efficiently providing public goods but that this is caused by diminishing MPCR (Isaac and Walker, 1988). However it has been found that even when using very low MPCR, social behaviour can not be entirely eliminated (Ledyard, 1995).
2.2.6 Subject pool
Besides experiments to test for the influence of repetition and communication, researchers have also questioned whether the found contribution rate is due to the fact that in almost all experiments university students are used and that most experiments are done in western countries (Konow, 2003). To check the existence of social behaviour across different cultures and to check the influence of individual characteristics, Henrich et. al. (2005) undertook behavioural experiments in a range of small-scale societies. Their main findings are that the selfish model fails in all societies studied, but that considerable differences in the level of social behaviour do exist. Another finding is that individual-level economic and demographic variables do not have a consistent influence on the presence of social
behaviour. Hence the presence of other regarding behaviour is robust for culture and individual characteristics (see also Konow, 2003).
12
3 Starting capital
3.1 Characteristics of starting capital
Although a lot of robustness tests are done it should still be questioned whether the findings that are described above give a good indication of the level of selfishness and generosity that can be found in real life. The ambiguity of the lab results and economic theory has already led several researchers to question whether either endowment
heterogeneity or heterogeneity of endowment origin could affect how people adhere to the predictions of rational theory. As this is a more recent field of interest, the research that is done is less extensive and work remains to be done.
In public goods experiments people are given the opportunity to contribute a part, or all, of their endowment to the public good. It is a standard practice in economic
experiments to provide participants with an initial endowment of money, a starting capital, which can be used for these contributions. But in the real world people make decisions considering their own earned income. This may cause a difference in how participants regard their money (Clark, 2002, p. 223).
There are three characteristics of starting capital which can cause this type of endowment to influence the decisions made in following rounds, the unexpected nature of the endowment, the fact that the endowment is unearned and the fact that the endowments are, often, homogeneous. On the other hand in real life endowments are often
heterogeneous, earned and expected. In the table below, these characteristics can be found on top of the columns while several hypotheses about how people regard their money are listed vertically4. These hypotheses about preferences of people all interact with one of the
characteristics of endowments after an earning phase.
4 The theories that are described by Konow in the ‘Utilitarianism and Welfare Economics’ category are
not listed in the table as these theories do not interact with one of the characteristics of endowments after an earning phase. ‘Utilitarianism and Welfare Economics’-theories assume that people prefer efficient outcomes, outcomes which maximizes aggregate outcome. These theories would thus assume
13
Heterogeneous Earned Expected
Equality and Need √
Equity and Desert √
Context √
PI Hypothesis √
3.1.1 Starting capital as heterogeneous endowment
The effect of heterogeneous endowments is explored by several researchers. See for example the article of Cherry, Kroll and Shogren (2005). In this article an experiment is designed to explore the impact of heterogeneous endowments and earned endowments on observed contributions in a linear public good game. The results suggest that the
contribution levels were significantly lower when groups had heterogeneous rather than homogeneous endowments. This finding was independent of the origin of the endowment.
This contradicts what can be expected in light of the ‘Equality and Need’-principle. Theories of ‘Equality and Need’ are based on the idea that outcomes in which people are treated equally and the basic needs of all are satisfied are preferred (Konow, 2003, pp. 1194-1200). However this would implicate that there is no reason to contribute to the public good if participants phase homogeneous outcomes and that in the heterogeneous case there would be a positive relationship between endowment and contribution.
However these results are in line with theories from the category ‘Context’ which can also predict differences between contributions when endowments are heterogeneous or homogeneous. For example if one looks at the influence of group identification one can argue that group identification is less when contributions are heterogeneous and
consequently contributions will lower.
3.1.2 Starting capital as unearned endowment
Theories of ‘Equity and Desert’ are based on the idea that outcomes in which endowments are in line with entitlement or property rights are preferred. This would mean that no contributions will be made in homogeneous cases and that a negative relationship between endowments and contributions can be expected in the earning treatments. Additionally, not all entitlements are treated in the same way. Within this group of theories consensus in reach about the fact that endowment heterogeneity based on effort is legitimized, while endowment heterogeneity based on luck is not.
14
No consensus in reached about endowments based on birth characteristics, for example intelligence level (Burrows and Loomes, 1994, p. 203, Konow, 2003, pp. 1206-1215 and Fahr, Irlenbusch, 2000, pp.277-278). It generally can be said that there is a tendency for a consensus that property rights are stringer when the perceived degree of responsibility is higher. However the perceived degree of responsibility remains subjective (Isaksson and Lindskog, 2006).
No conclusive results can be found to support this theory. In some experiments it appeared that group members contributed about the same to the public good whether their endowments were earned or not (Cherry, Kroll and Shogren, 2005, pp. 257-262). Although in other papers this view is contradicted and the absence of this effect in other studies is attributed to the symmetric nature of the games studied (Kroll, Cherry and Shogren, 2007). Some authors even find an inverse found money effect in which participants who earned their endowment contributed more when they played with participants who got their endowment from the experimenter. Note that in this experiment subjects in one group have different endowment origins (Spraggon, Oxoby, 2009).
Also Cherry, Frykblom and Shogren (2002) have been intrigued by endowment origin in economic experiments. In their experiment the authors investigate whether other-regarding behaviour can be repressed in a dictator game if people bargain over earned wealth instead of unearned wealth granted by the experimenter. The experiment consisted of two stages, earnings and bargaining. In the first stage dictators could earn either ten or forty dollars5. In the second stage the dictators bargained over their earned wealth. All
bargaining games are one-shot and players had complete information (Cherry, Frykblom and Shogren, 2002, pp. 1218-1219).
The baseline treatment resembled the normal dictator game in which the experimenter assigns an initial endowment to the dictator. In the earnings treatment dictators could earn money and recipients were informed that the dictators would receive money in a prior round. In the double blind treatment the set-up was the same as in the earnings treatment, except for the fact a procedure was followed that makes sure no one but the dictator will be able to find out how much he gave to the recipient (Cherry, Frykblom and Shogren, 2002, p. 1219).
In the above mentioned experiment the authors found that other-regarding
behaviour is greatly diminished when bargaining involves earned wealth and that this other-regarding behaviour is nearly eliminated when earned stakes are combined with anonymity.
5
15
In the earned-wealth treatment, isolated dictators acted as predicted by self-interested game-theory in 95 percent of the cases. So when asset origin is legitimized by effort and strategic concerns are controlled by isolation, other-regarding behaviour became the exception and self-interested behaviour the norm. As the authors indicate in their conclusion, asset origin could affect the degree of self-interested behaviour in other experimental settings as well and that the use of a starting endowment might explain the lack of free-riding in the provision of public goods in laboratory settings (Cherry, Frykblom an Shogren, 2002, pp. 1219-1220).
Also Oxoby and Spraggon (2008) conducted an experiment on earned wealth effects in dictator games. In this experiment either the receiver or the dictator could earn money in the first phase of the experiment (ten, twenty or forty dollar). After this, dictators would make a distribution decision based on the endowment after phase one. In the baseline treatment the pair was given a particular endowment (again ten, twenty or forty dollar) by the experimenter and the dictator would decide on how to divide this amount (Oxoby and Spraggon, 2008, pp. 704-705).
In the baseline treatment part of the dictators made the theoretically predicted zero offer. This percentage was lower when the endowment was higher. In the
dictator-earnings-treatment all dictators made a zero offer and in the receiver treatment the offers were significantly higher than in the baseline treatment (except in the ten dollar case). In the receiver-earnings-treatment even offers of one hundred percent were made (Oxoby and Spraggon, 2008, pp. 706-708).
From the fact that only in the twenty and forty dollar receiver-earnings-treatment offers were significantly higher than in the baseline treatment, it can be seen that only when the dictator could notice with certainty that the receiver had made any effort (pay-offs above the minimum pay-offs were achieved) higher offers were made (Oxoby and
Spraggon, 2008, p. 709). So indeed from this experiment it seems again that when earnings are legitimized by effort, behaviour is adjusted towards this.
3.1.3 Starting capital as unexpected endowment
The last characteristic is the unexpected nature of starting capital which may cause participants to not fully consider the opportunity cost of the endowment. This may have two effects. The first possible effect is that participants may have a higher propensity to consume and are thus willing to purchase more non-pecuniary goods such as fairness. With regard to the public good experiment this would mean that people are willing to buy more
16
public-spiritedness than they would have done if they had to earn their own income. Weak evidence for this can be found in the experiment of Oberholzer-Gee and Eichenberger (1999) but cannot be found in other experiments (Clark, 2002, p. 226). The second possibility is that the starting capital may cause individuals to engage in more risk-seeking behaviour because of mental accounting (Clark 2002 and Kroll, Cherry and Shogren, 2007, Thaler, 1999). In a public good experiment this can mean that individuals contribute more in earlier rounds to evoke contributions from group members in following rounds.
This hypothesis is tested by Clark (2002) by comparing VCM contribution rates when participants supply their own endowments, so endowments ‘brought from home’, with the contribution rates when endowments are provided. In both the treatments the distribution of promised earnings is kept constant (Clark, 2002, pp. 226-227). In his experiment, Clark is not able to find any house money effects, which suggests that the use of a free initial endowment does not distort contributions made in a VCM environment. However the conclusions of this article are contradicted in Harrison (2007). In this article it is argued that when using the appropriate statistical methods, the use of house money does have a significant effect on the contributions made in a VCM environment. The argument is based on the fact that the overall level of contributions might not be influenced but that this hides that the propensity to free ride is influenced.
3.2 Extension on conducted research
Although some things have been done already, the effects of endowment origin are still not entirely understood. In the article by Clark (2002) it is mainly investigated whether the unexpected character of endowments provided by the experimenter have an influence on contributions made to the public good. However, besides the unexpected character of the endowments provided, also the question if endowments are generated by some form of effort can influence the motivations to contribute to the public good. So in this case earned wealth compared to unearned wealth is the field of interest. This thesis will try to
contribute to this field of interest and can be seen as complementary to the research already done by Clark (2002).
The above described effect did receive some attention. In the article of Cherry, Frykblom and Shogren (2002) the authors research the effect of unearned wealth, though this is in a dictator game. This experiment basically investigates the main field of interest of this article in the case of the dictator game. But the incentives in the dictator game are not entirely the same and the dictator game is a very extreme game with regard to asymmetry.
17
It remains unclear whether the conclusions drawn from this experiment can be extended to the public good experiment and other experiments.
Another difference between this thesis and the article of Cherry, Frykblom and Shogren (2002) is that in their article, as in most other articles, only one way of earning your endowment is used while, as stated before, in this thesis three different ways of earning endowment will be explored. This is an extension that is not initiated before.
This thesis will also extend research already done by Oxoby and Spraggon (2008). In the mentioned article it remains unclear whether recipients are informed about how much the dictators earned in the first phase. If it is indeed the case that recipients are not informed about the earnings of the dictators, the dictators might take this lack of
information as an incentive to give the recipient next to nothing because the recipient is unaware of the size of the total amount anyway. In the results of the article of Oxoby and Spraggon (2008) one can therefore not detangle the effect of a lack of information and the effect of endowment origin.
The article of Cherry and Shogren (2008) also examines the dictator game. And also in this experiment only one type of earned money is examined. The findings of this article would point in the direction of finding endowment origin effects in public good
experiments. The experiment described in this thesis would be able to shed more light on this issue.
18
4 Set-up experiment
4.1 General set-up
This experiment will consist of a public good experiment with a voluntary contribution mechanism. The experiment consists of two phases. When the experiment starts participants are neither informed of the fact that the game consists of two phases nor is incorrectly told that the game consists of only one phase. This is done to avoid anticipation effects of future phases, such as possible effects on effort in phase one.
In phase one, participants are either able to earn their endowment or they just receive a particular endowment as done originally in economic experiment. This is the phase that differs considerably across treatments. The second phase consists of a normal one shot public good experiment with a voluntary contribution mechanism. Participants decide on how to divide their endowment between their private account and the public account. The size of the public good is given by the total of the individual contributions to the public good. The marginal pay-off to a participant is 0.4 tokens. The pay-off from phase two is given by the following pay-off function:
Π = Endowment phase one – gi + 0.4∑gj
Π = pecuniary payoff
gi = individual contribution to the public good
gj = total contributions to the public good
Only one round of the experiment is employed to avoid strategy choices. By using only one round and making sure all subjects are aware of this, it is avoided that reputation formation and other types of repeated game considerations play a role in determining the displayed behaviour (Fischbacher, Gächter, Fehr, 2001, p. 400).
19
After the earnings phase, participants are asked to fill in a contribution table. This contribution table states the ten possible6 endowments the other group members can have
after phase one and asks how much participants are willing to contribute to the public good if the stated endowments are earned. The set-up with the contribution table makes it possible to see if different player types can be distinguished and if endowment origin has an effect on the distribution of these player types.
In total participants are asked four times to answer a question or a set of questions. The first questionnaire is done after the introduction of the experiment and is used to check whether participants understand the incentives of the experiment. The second time, only in the earning treatments, is after participants have engaged in the earning part of the experiment. They are then asked about how they think they did compared to others and thus how high their endowment will be. The third time again a questionnaire is given right after the contribution table is filled in but before the endowments and contributions of the other group members are revealed. Participants are asked about their expectations of these endowments and contributions of the group members. The last questionnaire is done in the end. The answers of these questionnaires can be used as control variables in the regression that is run on the contributions made by participants. Among others, participants are asked about age, education and sex.
Groups will consist of four participants to make comparison with other experiments easy. In the standard public good game people receive a homogeneous
endowment so participants have full information about the endowment received by others. The use of the contribution table replicates this by asking how much the participants would be willing to contribute for different endowments of their group members.
In treatment two and three it will be announced that on basis of the score of phase one, participants will be either given a high, medium or low endowment and that these groups will all consist of one third of the participants. After the first phase, participants are asked to fill in the contribution table and are then randomly assigned to groups of four. After the division into groups, participants will be informed of the endowments of their group members. Their pay-off is then calculated based on the endowments of their group members, the corresponding contribution according to their contribution table and the contributions of their group members.
6 All three the same endowment: three possibilities, two have the same endowment and one has another
endowment: 3 x 2 = six possibilities and all have a different endowment: 3 NcR 3 = one possibility. In total ten possibilities.
20
The advantage of the above described set-up, with endowments based on relative scores instead of absolute scores, is that it is avoided that the work and talent treatments might not be exactly comparable as endowment distributions may differ between
treatments. However, the option of letting people compete over a position in a pre-set distribution can also have important disadvantages. In this setting participants participate in phase one and on basis of their results get a particular ranking and get an endowment based on this ranking. As distributions are set a priori, treatments would be perfectly comparable. The disadvantage might be that it cannot be said with certainty that the use of a ranking instead of an absolute pay-off does not influence the incentives given to
participants. However, in this particular set-up the incentives are not changed as it does not matter why a certain effort level is chosen but that a certain level is chosen.
On top of this, the use of a contribution table will make sure that a complete preference set of all participants is collected and that the results of the different treatments are comparable. As participants are unsure about the endowments of the other group members after phase one, they have to take all options seriously and supply a full contribution table in which all possible scenarios are played out.
As groups consist of four participants, a multiple of four participants is required to play each treatment. One should also keep in mind that a part of the participants might not be willing to engage in the first phase of these treatments. Especially in the talent
treatment, participants might not want to participate in the quiz. In total sixteen
participants are needed to conduct each treatment once, so a multiple of sixteen is required to conduct the experiment.
The expected pay-off for participants should exceed the average wage that could be earned by students outside the lab. The whole experiment, including start-up and payout will probably not take more time than one hour so an expected total pay-off of fifteen Euro with a minimum of around five Euro will attract enough students to the experiment.
The lowest possible endowment is an endowment of Emin after phase one. And a
minimum pay-off of (Emin - Emin) + 0,4 * (Emin + 0) = 0,4 Emin at the end of the experiment.
However, as contributing to the public account is not mandatory, participants have full power to make at least Emin. Consequently the following possible endowments are
proposed for all treatments: Emin = 6, Emed = 12 and Emax = 18.
The above proposed endowments are chosen for several reasons. First, if all possible endowments are all multiples of three, the average of the endowment of the three group members of a participant is always a round number which makes calculation easier.
21
Second, if all the endowments are equal numbers, all participants have an unequal number of possible contributions to the public good which means all participants have a middle option. Third, Emin is an acceptable lower bound on the pay-off from the experiment.
Fourth, for consistency the difference between Emedand Emin and the difference between
Emax and Emed is equal7.
4.2 Treatments
The experiment consists of four treatments; the baseline treatment, the work treatment, the IQ talent treatment and the music talent treatment.
Treatment 1 = Baseline. In the baseline treatment participants receive their
endowment from the experimenter at the start of the experiment. The experimenter randomly assigns one third of the group to low endowment, one third of the group to medium endowment and one third of the group to high endowment. This set-up is comparable to original standard public good experiments. While not emphasized,
participants are allowed to read a magazine or a book or spend their time studying during the passing time.
Treatment 2 = Work. In the work treatment participants receive an endowment based on
work done, the possible endowments are the same three levels as in the baseline treatment. Individuals are given a minimum endowment of Emin and a maximum endowment of Emax
depending on the relative outcome of the work that is done. To make sure that not all participants choose the same effort level one should make sure that the task that has to be done is unpleasant to a certain extent.
Participants will see a number appearing on their screen and the will have to type in exclamation marks according to this number, followed by an Enter. So if the number appearing on the screen is two, participants will type ‘!’ ‘!’ ‘Enter’. The numbers change but will always be larger than zero but smaller than six. The computer will give the numbers in the same order to everybody and follows a list, made by and analogue prior to the
experiment. It is expected that all participants have equal abilities in performing this task so
7
In the experiment design also the number of participants is of great importance as this influences the significance of the results. In the end the number of participants depends on the budget of the
experimenter. The average endowment will be 12 Euro across all treatments. Taking into account that one run off all treatments consists of 12 participants we can make the following calculation: Expected costs per run = 12 Euro * 12 = 144 Euro. As there are ten different possibilities of endowment distributions it is preferred that at least all possibilities are played out ones, even though also contribution tables are collected. The expected costs per ten runs = 10 * 144 Euro = 1.440 Euro (excluding an additional amount for unexpected costs). When possible, more runs of the experiment are preferred as this increases the reliability and the significance of the results.
22
the outcome of the work task only depends on effort8. Participants have twenty minutes
for the task.
Treatment 3 = IQ talent. The endowment in the talent treatment is based on an
IQ test, where the outcome of the IQ test is a consequence of intelligence level which is generally accepted to be largely exogenous (Plomin, Spinath 2004). To make sure the endowment is indeed based on intelligence, a widely accepted IQ test will be used, the General Admission Test (GMAT)9. Participants will have twenty minutes to complete ten
questions from the sample test and are given a minimum endowment of Emin and a
maximum endowment of Emax depending on the relative number of correct answers.
If participants do not want to participate is this part of the experiment because they do not want to display their intelligence level, individuals can just give random answers to the questions. In this case they know the experimenter can see that they decided not to display their intelligence level and they also know they will only Emin.
Treatment 4 = Music talent. In this treatment participants will receive an
endowment based on their relative performance in Seashore’s Test of Musical Talent (Seashore, 1919). Besides this, the treatment is the same as treatment 3. So instead of endowments based on intelligence level, endowments are now based on musical talent. Adding this treatment can give further inside in what people regard as strong entitlements and what not (see the discussion in paragraph 3.3).
The Seashore test consists of six parts which all resemble a subtest. The subtests are pitch, rhythm, loudness, time, timbre10 and total memory. In each subtest participants
face a two-alternative forced choice discrimination task in which participants are asked to judge whether the second sound/sequence is different from the sound/sequence that was heard before (Musacchia, Strait and Kraus, 2008, p. 36). Because of the forced
two-alternative questions the data of this test is easily to compare and record, which makes this test very useful for this experiment.
8 This work task is inspired on the work task of Silammaa, 1999. See for other work tasks Bridger and
Long, 1984, Bosman and Van Winden, 2002 and Fahr and Irlenbusch, 2008.
9 The use of this test follows previous work (see for example Hoffman et al., 1996, Clark, 1998, Kroll,
Cherry and Shogren, 2008, and Cherry, Frykblom and Shogren, 2002). The GMAT is used by many graduate schools of business to measure the qualifications of their applicants. The ten multiple choice questions of the test used by List and Cherry, 2000 will be used in this experiment. The participants in the experiment of List and Cherry have 45 minutes to complete ten multiple choice questions and seven open questions.
10 “Timbre is widely understood to be the sound quality which can distinguish sounds with the same pitch and loudness (e.g., the quality of a trumpet versus a violin)” (Musacchia, Strait and Kraus, 2008, p. 40).
23 4.3 Summary set-up experiment
o Welcome and explanation phase one o Phase one: earning/assigning endowment o Explanation phase two
o Questionnaire 1: check if experiment is understood o Questionnaire 2: expectations about own endowment11 o Reveal endowment to participants
o Phase two: fill in contribution table
o Questionnaire 3: expectations about endowment en contributions group members o Reveal endowments and contributions group members
o Calculate payoff participants
o Questionnaire 4: general information o Payment participants
24
5 Data analysis
5.1 Data description
There are two seemingly similar questions that can be asked in this research which can be both deducted form the data of contributions that are being made. The first is how participants play against participants with a particular work/IQ/music rating and the second is how participants with a particular work/IQ/music rating play. These two ways of looking at the data are described separately.
5.1.1 Contributions made to the public good
After the data is collected, to give a first indication of the results, the table below can be given. On the vertical endowments all possible endowments of the group members are given, as well as an average of all possible endowments together and on the horizontal axe, the treatments are given. To give a further inside in the data, also a column with the average contributions of the three earnings treatments are given, as well as an average of the two talent treatments. This can be extended by dividing the columns into four columns, one for the three possible own endowment levels after phase one and one overall column.
In this table the upper left cell will give the average contribution that is made by all participants across treatments. The cell at the right of this one gives the average
contribution made in the baseline treatment. On basis of earlier experiments it can be expected that this percentage will be between forty and sixty percent.
25
Table 1: Average contributions to public good per treatment in percentage of endowment
All treatments T1 T2 T3 T4 Earning treatments (T2-T4) Talent treatments (T3-T4) All possible endowments Average contribution Exp. 40-60 6 – 12 – 18 6 – 6 – 6 12 – 12 – 12 18 – 18 – 18 6 – 6 – 12 6 – 6 – 18 12 – 12 – 6 12 – 12 – 18 18 – 18 – 6 18 – 18 – 12
Another way to represent the collected data before a regression is run, is to make a scatter plot per participant and one per treatment with the possible endowments of the group members on the horizontal axe and the contributions as a percentage of the own endowment on the vertical axe. Again a further distinction can be made by making a diagram per treatment per own endowment level. With these diagrams it can be seen whether particular player types exist and whether the distribution of these player types differs per treatment.
5.1.2 Contributions ‘received’
The second way of looking at the data is to see how participants play against participants with certain endowments. In the ten possible endowment sets there are six sets with at least one participant with a low endowment, six sets with a participant with a medium
endowment and six sets with at least one participant with a high contribution.
By adding the contributions made in these cases one could see whether contributions made in these cases differ. However in this case a set with one participant with a low
contribution is treated the same as a set in which there are two participants with a low contribution. To overcome this, one can just compare how participants play against the set
26
6-6-6, 12-12-12 and 18-18-18. A table can be made with the own endowment in columns and the endowments of the group members in rows. By comparing the contribution percentages in the first column it can be seen whether participants with a certain
endowment receive different contributions to the public good. This can be simply tested with a non-parametric test.
All
endowments
Low Medium High
6 – 6 – 6 12 – 12 – 12 18 – 18 – 18
5.2 The model
With regard to the experiment and the collected data, it can be expected that the contribution of participant i (Ci,) to the public good depends, besides on some control
variables, on own endowment, the endowment of the group members and the treatment to which the participant is assigned. The endowment of the group members can be
characterized by two variables, the total of the endowments of the three group members and the standard deviation12 of these three endowments13.
A priori it cannot be defended to just include three14 treatment dummies in the regression analysis as there is no reason to believe that there is just a fixed difference in contributions between all treatments15. It might for example well be that participants in the
work treatment behave differently in response to their own endowment. Hence to include this into the regression analysis, besides standard dummies for the different treatments, also slope dummies are included.
Consequently the model of interest can be summarised as follows:
%Ci = β0 + β1 T1E + β2 T1EG + β3 T1VEG + β4 T2E + β5 T2EG + β6 T2VEG + β7 T3E
+ β8 T3EG + β9 T3VEG + β10 T1 + β11 T2 + β12 T3 + β13 EG + β14 VEG +
β. control variables + u
12 Standard deviation summarizes how far away from the average the data values typically are. Standard deviation = s = √s2 with s2 = ∑((xi- xaverage)2)/(n-1) (Cooper and Schindler, 2003, p. 475).
13
When only the variable total endowment would be used, some of the possible sets of group
endowments would be characterized by the same number. For example the endowments 6, 12 and 18 and 12, 12, 12 both amount to 36. To be able to distinguish between these sets of endowments also a variation variable is included.
14
As there are four different treatments, three dummies have to be included to be able to represent all treatments. The baseline treatment will be represented by setting all three included dummies to zero. 15 Just including standard treatment dummies would mean that the model is only able to work with vertical shift between regression lines for the different treatments.
27
In which %Ci stands for the contribution of participant i in ℅ of the endowment, T stands
for treatment, E for own endowment, EG for the total endowment of the group members and VEG for the variation in the endowment of the group members. As Ci is a percentage
of endowment, the range of possible contributions lies between zero and one hundred percent. As there are no limitations to the predictions of the above described model, mathematical translations are in order to make sure the model only gives %Ci between zero
and one hundred. To do this a functional logit model will be used16.
5.3 Expectations
5.3.1 Own endowment
The first variable that is included in the model is the own endowment. As
suggested in Van Dijk and Grodzka (1992, p. 331), contributions to a public good provide participants with an option to redistribute wealth. This would mean that overall,
participants who are given a high endowment will contribute a relatively large part of their endowment to the public account and participants with a low endowment will contribute a relatively small part of their endowment to the public account, independent of the
treatment they are in. This resembles behaviour in line with the ‘Equality and Need’-theories in which participants strive for equal outcomes and follow the rule of proportionality (Cress and Kimmerlee, 2008 and Konow, 2003).
When following the ‘Equity and Desert’-theory, one can argue that when endowment differentials are assigned by chance, as in the baseline treatment, group members will strive for equal outcomes. However in light of this theory this relationship might change if endowments are earned, this is why there is not just one variable for own endowment but one for each of the four treatments. In the case of earned endowments it can be expected that participants with a higher endowment feel that this money belongs to them more than to the group as a whole, which would mean that the above described positive relation between own endowment and contributions will not be found. According to the ‘Equity and Desert’-theory, people prefer and expect distributions of wealth to be in line with distributions on related dimensions.
16 When Xβ resembles all explanatory variables the following reasoning can be followed:
-∞ < Xβ < ∞, 0 < exp(Xβ) < ∞, 1 < 1 + exp (Xβ) < ∞, 0 < 1/(1 + exp (Xβ) < 1.
So % Ci = 0 < 1/(1 + exp (Xβ) < 1 in which % Ci is always between zero and one, with gives after multiplying by one hundred a percentage between zero and one hundred percent. This transformation can be resembled by running a functional logit regression in Stata (Buis, 2006 and Papke, Woolbridge, 1996).
28
This means people prefer a positive relationship between for example effort (equity theory) and wealth and ability (status-value theory) (Cress and Kimmerle, 2008, Van Dijk and Wilke, 1994, p. 353 and Konow, 2003).
So these theories would predict that when people think that they are entitled to their endowment, they do not want to share their earned money (Van Dijk and Wilke, 1994, p. 353). The redistribution effect of contributing to the public good is in this case not of importance which means that it is expected that participants contribute less. However this effect might be influenced by the way in which the money is earned. The results of this experiment will be able to give more insight in this.
What happens if you are able to earn money by pure effort (the work treatment) and you put as much effort as you can in earning money and end up in a group with all other people having low pay-offs. Do you follow the rule of proportionality because you think it is fair that participants with a higher endowment contribute more, following the equity norm, or do you punish the other players for shrinking in the previous rounds by lowering your contribution? Do you feel solely entitled to your endowment because of the effort put into it? And if you have a low endowment after phase one you might be inclined to look at the difference in endowments without paying attention to the difference in work done and reason that you do not have to contribute because you already have a low
endowment.
The answers to these questions may shift entirely when the money is earned by talent instead of work (the IQ talent treatment and the music talent treatment). Let’s say you have a reasonably high intelligence level and an associated high endowment. As in the prior example you end up in a group with all other participants having low endowments. As low pay-offs are caused by a low intelligence level you might be inclined to contribute more because you think it is fair that the more privileged people contribute more.
So preferences might differ between the work treatment and the talent treatments. This because people can influence effort but not intelligence or musical talent. This might mean that people prefer a more equal outcome when the endowment is earned by use of talent. So in the talent treatments still some redistribution preferences might be present while these redistribution preferences are absent in the work treatment.
29
5.3.2 Total endowment group members
The second variable that is included in the model is the total endowment of the other group members. Different reasoning can be followed. If the other three group members have a high endowment it might be that participants are more likely to think that the others will contribute to the public good which can cause someone to think that it is then not necessary to contribute themselves. However it might also cause people to feel like they have to match these expected contributions and consequently they might contribute more too. Again, as this reasoning might be different for the different treatments not just one variable is included but one for each treatment. If the group members have a high endowment in one of the earnings treatments participants might be inclined to reward those group members by contributing more to the public good than they would in the baseline treatment.
5.3.3 Standard deviation endowment group members
The possible endowment of the group members is represented by the variable of the total endowment of the group members together with a variable for the variation in the
endowment of the group members. Consequently, the third variable that is included is the standard deviation of the endowment of the group members, again for each treatment, one variable is included. As indicated in paragraph 3.2 on page eleven, previous research, for example of Cherry, Kroll and Shogren (2005), with respect to this variable indicates that in case of heterogeneous endowments contributions to the public good were significantly lower.
5.3.4 Treatment dummies
To conclude with, also three standard dummies for the treatments are included. With these dummies it can be seen whether there is a treatment effect on contributions to the public good, for example if people are less generous with regard to a public good if they have to earn their money prior to making a contribution.
Participants in one of the earnings treatments might contribute for example less to the public good because just the fact that they had to earn their money themselves
decreases their other-regarding behaviour. For example, in the baseline treatment participants will just randomly receive their endowment from the experimenter. The possible feeling of deserving this might in this case be well below the feeling of deserving this when the participants earned their money. As there is no previous work on the effect
30
of different ways of earning money, no expectations can be formed about the relation between the coefficients of these three treatments17.
Above expectations are not only based on theories but also on prior findings in dictator games. In these games pay-off distributions are influenced by the source of wealth. Individuals with legitimate claims to certain assets, due to the effort put into earning these assets, receive larger shares than when endowments are determined by the experimenter (see for example Oxoby and Spraggon (2008), Cherry (2001) and Cherry et.al. (2002)). As Oxoby and Spraggon (2008) state: “We find that property rights (created by legitimizing assets) play a crucial role in individuals’ revealed preferences.” In line with the findings in dictator games the expectation is that the contributions after wealth is earned will be closer to the theoretically predicated zero contributions.
5.3.5 Control variables
Additionally, the model will control for sex, political preference and field of study. This to avoid a bias due to fact that the effect of these variables will be included in coefficient of one of the variables described above. However they are not expected to have a significant influence on the contributions (Henrich et.al. 2005). When the number of participants is low, it might be necessary to combine several characteristics. For example, it can be tested if the answers to variable ‘origin’ can be combined to ‘western’ and ‘non-western’ because it might well be that the number of participants with an African origin is too low to create an own group. The same might be true for the variable ‘political preference’. When the number of participants is low or if some groups have too few members, a new
classification can be made with just ‘left’, ‘right’, ‘centre’ and ‘no preference’.
5.4 Significance variables
To develop this general model into a more specific function and to check the validity of this function, several tests should be undertaken after the data is collected. To start with, a simple test can be done to check whether the included variables are significant as a group. This can be easily seen after a regression is run. Two of the automatically calculated
numbers will be the F-statistic which uses the number of variables included and the degrees of freedom and also the P-value of this F-statistic is automatically calculated. The null hypothesis of this test is that all the betas of the variables are equal to zero and the alternative hypothesis is that at least one of the variables has a beta unequal to zero.
31
When p ≤ 0.05 it can be rejected with 95% certainty that all the variables have a beta equal to zero. This means that the model is still in potential a good fit to the data and additional tests can be conducted. When the null hypothesis cannot be rejected it means that the model is not a good fit to the used data and that a different model has to be set-up (Dougherty, 2007, pp. 104-105).
After this first look at the model, one has to look at the whole table that is provided by the statistical program after running a regression18. From the p-values that are provided
we can see which parameters are significant and which are not. The following scale can be used, variables with a p-value of 0.01 or lower are highly significant and can be marked with ***, variables with a p-value higher than 0.01 but smaller or equal to 0.05 are significant and can be marked with ** and variables with a p-value higher than 0.05 but lower or equal to 0.10 are weakly significant and can be marked with *. All variables with a p-value higher than 0.10 are not significant which means that it can not be rejected that they are equal to zero, hence these variables will not be included in the model.
5.5 Tests for misspecification19
5.5.1 Goodness-of-fit test
One way to have an indication of the goodness of fit of the model is to use the R2, which is
automatically calculated by the statistical program after running a regression. The R2 is a
fraction between 0.0 and 1.0 and does not have a unit of measurement. Higher values of R2
indicate that the model fits the data better. Note that the R2 is just an indication of the
goodness of fit and does not prove that the model is “good”. A high R2 only means that
the curve that is drawn by the regression came very close to the data points. But although the R2 cannot be used on it's own to review the model, it is still a measurement of interest
(Dougherty, 2007, pp. 104-105).
18 Either Ordinary Least Square or Two Stage Least Square, see paragraph 5.4.
19 It is not necessary to test or look for the presence of outliers as all contributions will be in the range of zero to one hundred percent and even these two extreme contributions represent ‘normal’ contributions. Zero is the prediction according to the theory of the economic man while a contribution of one hundred percent of the endowment is the social optimum. It is also not necessary to look for endogeneity.
Endogeneity occurs when an independent variable is correlated with the error term in a regression model. The effect is that the coefficient in an ordinary least square regression is biased and that t-tests and f-tests cannot be used. In this case, the two stage least square method and instrumental variables are more adequate to review the collected data. However endogeneity will not be present as data from the lab is used in stead of empirical data.
32
5.5.2 Heteroscedasticity
One assumption for the use of the Ordinary Least Square method is that the variances are homogeneous (homoscedasticity). To test if this assumption is indeed correct and that Ordinary Least Square method can be used, one has to test for heteroscedasticity. The null hypothesis is homoscedasticity and the alternative hypothesis is heteroscedasticity. When the null hypothesis is rejected, one has to review the model again and see if a different specification can be found. One possible solution to heteroscedasticity is to look for important variables that were excluded before (Dougherty, 2007, pp. 224-239).
The presence of heteroscedasticity can be tested in several ways. One way is to use the Goldfeld – Quandt test. This test is designed to find a specific form of
heteroscedasticity, namely that the standard deviation is in proportion to one of the parameters. The most general test which can be conducted to test for the presence of heteroscedasticity is the White test. Drawback of this test is the low power of the test which means that a large number of observations are needed to be able to use this test (Dougherty, 2007, p. 92).
5.5.3 Autocorrelation
Closely related to occurrence of heteroscedasticity is autocorrelation. Autocorrelation means that the assumption that the disturbance term of an observation is independent of the value of the disturbance term of the other observations does not hold. Positive autocorrelation means that the disturbance term of a term is more likely to have the same sign as the previous disturbance term. Negative autocorrelation means that the sign of a disturbance term is more likely be the opposite than the sign of the previous value (Dougherty, 2007, p. 354). Just as heteroscedasticity, autocorrelation can be caused by omitted variables and can possibly be resolved by including more variables. However, both autocorrelation and heteroscedasticity are more likely to be a problem when time-series data is examined. As the data that is collected in this experiment is not gathered over time, it is unlikely that the model suffers from either of these two misspecifications20.
20 The collected is not really series data but one should keep in mind that, just as in the case of
series data, multiple observations per participant are collected which is related to the collection of time-series data.
33
6 Concluding remarks
Unfortunately the time span and the budget of this thesis did not allow for an extended or even small data collection. On that account, this part is missing in this thesis which means that only the background information and the research proposal is given. However, even without this part the literature research and the set-up and the expectations of the
experiment give a good idea of the topic of interest and the relevance of this subject for the field of public and experimental economics.
As said before, the topic of endowment origin is underdeveloped and will still be after the described experiment although maybe a little bit less. Follow up research questions remain. Possible extensions to the experiment might be the introduction of a phase 1b in which participants can negotiate a group norm about how much, or which percentage of the endowment, to contribute.
Other possible extensions are to include more treatments and capture more circumstances in which money is earned or received in the real world. One can think of a treatment in which the endowment is not only received by chance (as is the case in the baseline treatment in the described experiment) but endowments are generated by a combination of chance and investment. This would, for example, resemble the option of buying a ticket for a lottery. When the lottery is indeed won, you have been very lucky. But you did make an investment to make a chance. Another possibility is adding a treatment in which endowments are based on creativity, in which creativity can be based on the
34
References
Bolton, GE., Katok, E., Zwick, R. (1998). Dictator game giving: Rules of fairness versus acts of kindness. International Journal of Game Theory, vol. 27, (2), pp. 269-299.
Bosman, R. Winden, F. van. (2002). Emotional hazard in a power-to-take experiment. The Economic Journal, vol. 112, pp. 147-169.
Bridger, R.S., Long, J. (1984). Some cognitive aspects of interface design in a two variable optimization task. International Journal of Man-Machine Studies, vol. 21, pp. 521-539.
Brosig, J. , Ockenfels, A., Weiman, J. (2001). The effect of communication media on cooperation. German economic review (forthcoming). Feb 2001.
Buckley, E., Croson, R. (2006). Income and wealth heterogeneity in the voluntary provision of linear public goods. Journal of Public Economics, vol. 90, pp. 935-955.
Buis, M.L. (2006). Proportions as depended variables. http:home.fsw.vu.nl/m.buis
Burrows, P. Loomes, G. (1994). The impact of fairness on bargaining behaviour. Emperical Economics, vol. 19, (2), pp. 201-221.
Cherry, T.L. (2001). Mental accounting and other regarding behavior: Evidence from the lab. Journal of Economic Psychology, vol. 22, (5), pp. 605-615.
Cherry, T.L., Frykblom, P., Shogren, J.F. (2002). Hardnose the dictator. The American Economic Review, vol. 92, (4), pp. 1218-1221.
Cherry, T.L., Kroll, S., Shogren, J.F. (2005). The impact of endowment heterogeneity and origin of public good contributions: evidence from the lab. Journal of Economic Behavior & Organization, vol. 57, (3), pp. 357-365.
35
Cherry, T.L., Shogren, J.F. (2008). Self-interest, sympathy and the origin of endowments. Economic Letters, vol. 101, (1), pp. 69-72.
Clark, J. (2002). House money effects in public good experiments. Experimental Economics, vol. 5, (3), pp. 223-230.
Cooper, D.R., Schindler, P.S. (2003). Business research methods, McGraw-Hill Companies, 8th
edition.
Cress, U., Kimmerle, J. (2008). Endowment heterogeneity and identifiability in the information-exchange dilemma. Computers in Human Behavior, vol. 24, (3), pp. 862-874.
Davis, D.D., Holt, A.H. (1993). Experimental Economics. Princeton University Press.
Dijk, E. van., Grodzka, M. (1992). The influence of endowments assymetry and
information level on the contribution to a public step good. Journal of Economic Psychology, vol. 13, pp. 329-342.
Dijk, E. van, Wilke, H. (1994). Asymmetry of wealth and public good provision. Social psychology Quatarly, vol. 57, (4), 352-359.
Dougherty, C. (2007). Introduction to econometrics. Oxford University Press, 3rd edition.
Fahr, R., Irlenbusch, B. (2000). Fairness as a constraint on trust in reciprocity: earned property rights in a reciprocal exchange experiment. Economic Letters, vol. 66, (3), pp. 275-282.
Fischbacher, U., Gachter, S., Fehr, E. (2001). Are people conditionally cooperative? Evidence from a public goods experiment. Economic Letters, vol. 71, (3), pp. 397-404.
Harrison, G.W. (2007). House money effects in public goods experiments: comment. Experimental Economics, vol. 10, (4), pp. 427-437.
