Jumping the Queue
Final Version August 15, 2016
ABSTRACT
Recent literature examined the antisocial behavior of people. From different laboratory experiments we learn that such behavior is quite common in such settings. The important question that arises from this evidence is whether this nasty behavior is consistent with real life behavior. This study tests the external validity of these results by comparing Dictator Game outcomes with real life behavior among children. This papers shows an absence of a correlation between lab and field behavior, and therefore, laboratory results are not found to be predictive for this sample.
Keywords: Social preferences, Dictator Games, external validity, field
behavior. JEL
classification: C91, C93, D03.
_________________________________
Master Thesis in Business Economics (15 ECTs) Author: Olaf van der Wardt
Student ID: 5957648
Supervisor: Dr. Jeroen van de Ven University: Universiteit van Amsterdam
Track: Managerial Economics and Strategy
Statement of Originality
This document is written by Olaf van der Wardt who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
In the past few decades, laboratory economic experiments are increasingly used to examine human decision behavior. Experimental results show that people tend to deviate systematically from the conventional economic framework. As a consequence, theorists adapted standard models by incorporating psychological aspects and social considerations people have, to predict the anomalies seen in experiments. Famous alterations include altruism (Levine, 1998); fairness (Rabin, 1993); inequity aversion (Fehr & Schmidt, 1999); and reciprocity (Andreoni & Miller, 2002). Lab experiments offer many advantages, such as controlling desired variables for control and treatment groups under ceteris paribus conditions. Therefore, these experiments are seen as the golden standard to research and test economic theory. However, critics state that experiments conducted in laboratories are remote and therefore do not reflect real life situations (e.g. Guala, 2012; Haley & Fessler (2005); and Levitt & List 2007). Due to the abstract nature of laboratory experiments, the external validity of the results obtained in these experimental settings have been of debate.
` Levitt &
List (2007) discuss the generalizability of laboratory results. One of their main objections towards reliable inferences outside of the laboratory environment is the scrutiny of actions by other participants and experimenter. They argue that it is often unclear how participants perceive their actions to be observed, and therefore obtained results can be influenced by factors other than participants’ true preferences (e.g. signaling fairness, meeting experimenter demands).1 Furthermore, self-selection of volunteers poses another problem for the results to be a good indicator for real world behavior. That is because subjects in laboratory experiments are mostly volunteering students. Next to the fact that these students differ in age and attained education with respect to the general population, the self-selection characteristic itself might also be an indication that the used subject pools don’t offer the opportunity to generalize results without caution. The last objection raised by the authors is that in order to make clear-cut interpretations of the results, the laboratory design may become too3 abstract (context free). Despite the advantage of controlling the desired
1 Also Zizzo (2010) suggests the presence of social and cognitive experimenter demand effects in laboratory
experiments. Subjects may feel obliged to some extent to meet the “demands” of experimenters. Several factors of artificial experiments may trigger such behavior, such as indicating variables of interest, available option sets, or the offered context.
stimulus in such a setting, it is at the expense of the external validity since participants don’t face real life experiences. The above casts doubts on how we have to treat the results from artificial experiments, in particular for those involving other-regarding preferences.
Perhaps the most famous experiment to measure such social preferences is the Dictator Game (DG) introduced by Kahneman, Knetsch & Thaler (1986), and later adopted by dozens of other researchers. The game involves a dictator and an anonymous receiver, in which the dictator is endowed with an amount of money. The dictator then has to decide how much of the endowment, if it all, (s)he is to give to the recipient. Standard economic theory predicts no dictator will grant any of the money to the recipient to maximize its own monetary payoff. However, empirically it is shown that dictators do give money to their anonymous receivers. From a meta study performed by Engel (2011), it is seen that in 129 Dictator Game published studies in the preceding 25 years, roughly a third of the entire endowment was granted to the recipient, with two thirds of the individuals willing to share an amount of the endowment. This behavior – accompanied by the simplicity of the game that leaves little room for misunderstanding the task under experiments’ subjects – has been seen as proof for people having other-regarding preferences instead of purely maximizing own wealth (e.g. Camerer, 2003). In light of the internal validity of these kind of experiments, Bardsley (2008) and List (2007) altered the regular Dictator Game to test the consistency of the altruistic like behavior seen in so many experiments. Their variation endows both the dictator and the partner with money, instead of only the dictator. This alteration still lets the dictator control the outcome of the game, but now (s)he is also offered the opportunity to take (some of) the endowment from their partners, next to the options of sharing (an amount) of the endowment, as well as opting for both players to keep their initial endowment. Their results are based upon between-subjects variation and show that many fewer dictators are willing to share money in the “take-version” of the DG than those involving in the regular DG. Strikingly, many participants playing the role of the dictator opted for the take option. For instance, in the game in which the dictators are endowed with unearned money in List (2007), almost half of them took the maximum of their partners, while 35% took a nonzero amount when the money was earned. Results from the action sets used in Bardsley (2008) resemble the non-altruistic proportions seen in List (2007).
Despite the fact a large portion of people maximized purely their own wealth in both studies, not all participants chose the most selfish outcome. Hence, this evidence cannot disregard the fact that social preferences are taken into consideration by subjects while making a decision on how to distribute the money in a DG.
The generalizability of results obtained in a DG to a real-world setting have not been extensively explored yet. Up to date, only studies examine the external validity of Dictator Games. As we will see in the literature section, most of these studies implement a context-free laboratory experiment accompanied by one field experiment – which is trivially not context free. For my research, I want to further investigate how well dictator game outcomes observed in an artificial setting are associated with behavior in the field. More specifically, I will be interested in whether pro-social behavior displayed in a laboratory Dictator Game is externally valid. My main contribution is that I will make use of field data that is obtained from observations regarding social behavior for years of the participants. This unique dataset may offer an interesting opportunity to see how well these data match with a simple social experiment in a more systematic way.
To examine the generalizability of (anti-)social behavior seen laboratory outcomes, this study links real life (i.e. field) behavior of children to the results of a take-variation of a Dictator Game. Children tend to be more generous than the results obtained with adults in previous literature. However, results show that these outcomes are not externally valid since no correlation exists with the field data.
The structure of the rest of my thesis consists of five sections. First, I will briefly summarize the main similar studies that document comparisons between behaviors in the lab with field data. Next, in the methodology section, the experimental setup, used datasets and the hypothesis are described. After that, the results are discussed. Accordingly, I will offer my conclusions after considering the limitations of my research.
Up to date, natural-field experiments are still relatively scarce, and comparisons between social behavior in and outside the lab are even less plentiful2. A natural-field experiment is an experiment in which subjects are not aware of them taking part in it – and therefore believe their actions are not observed – and in which experimental contexts don’t play a role (Harrison & List, 2004). Ideally, such an experiment duplicates a laboratory setting – in which one can control the desired stimuli and randomization of subjects – into a field setting. However, the scarcity seen from natural-field experiments can be explained by the facts that these settings are typically hard to conduct perfectly due to factors as costs, finding volunteering companies (plus dealing with the problem of self-selection), and complexity of tasks. Nevertheless, there are several papers that are relevant for my research, in that they are investigating behavior in and outside the laboratory. I will discuss the studies performed in non-Western and Western societies separately. In the last paragraph, I will motivate my research by explaining my contribution to the existing
literature. The first
non-Western study is that of Karlan (2005). He conducted an experiment with indigenous Peruvians. He tests the external validity of (pro-)social behavior by conducting both a Trust Game and a Public Goods Game and monitoring the repayment of microfinance loans. Only for the Trust Game he finds that people behaving pro-socially also are less likely to default. Another study with indigenous inhabitants by Gurven & Winking (2008) show no consistency between in lab and field outcomes. In their study, they let Bolivian participants play Dictator and Ultimatum Games. They find that their sharing behaviors of the everyday distributed food do not correspond to their behavior in those games. Remarkably, the authors find that outcomes of both games are more self-interested than is seen in the same experiments in Western societies. A third study with South American subjects on the consistency of cooperativeness is that of Fehr & Leibbrandt (2011). They let fishermen of shrimps participate in a Public Goods Game and measure the in-field cooperativeness by the traps used. Local fisherman use plastic bottles with little holes in it as shrimp trap. The smaller the holes are, the more infertile shrimps are caught from the lake, causing
2 Card, DellaVigna & Malmendier (2011) gathered all papers documenting field experiments that were
published in one of the top five economic journals in the preceding 35 years. They counted 84 studies of which only eight involved social preferences.
exploitation of the free accessible lake. Their outcomes show no correspondence between the field and laboratory data. Another lab-field study performed in a non-western society is that of Barr & Zeitlin (2010). Their study is on the behavior of teachers in Uganda. Teacher absenteeism has been a major problem in the East-African country.3 The authors show that teachers who are more altruistic in a Dictator Game are also spending more time lecturing, thereby verifying external validity of altruistic behavior in this setting. From the discussion of these papers, it is seen that most – but not all –of these studies show (at least) a moderate association between behavior seen in the lab and in the field.
Next to these studies conducted in non-industrialized countries, also the external validity of in-lab games with Western participants have been published recently. The first example of such a study is that of Engelbrecht-Wiggans, List & Reiley (2005). In his paper he reports on his findings with respect to sports card traders’ behavior in a controlled gift-exchange game in a laboratory and at a sports card trading market. While the traders’ behavior reflects pro-social behavior in the laboratory game, their actions in the natural trading market approaches outcomes predicted by self-interested motivations. It appears that social preferences only surface due to reputational concerns: suppliers only deliver high quality products when these products can be checked for certification by the buyer. Also, the possibility of future interactions between buyer and supplier play a more vital role in trading decisions than social considerations do in the sports card trading market.
Another study by Benz & Meier (2008) with students of the University of Zurich argue that a correlation exists between a donation game experiment in-class and a natural occurring donation game with students. For both experiments students were endowed with money in advance. The authors controlled for an order-effect and find that the correlation of 0.28 for students who first participated in the in-class experiment and later were part of the in-field donation game. The correlation for students following the opposite order was 0.40. Results show that for the in-class experiment donations amounted approximately 65% of the endowment and almost 80%, respectively. However, it likely seems that the study suffered from anonymity problems since the amount of donations made is quite high. Also, since students were
3 Chaudhury, Hammer, Kremer & Muralidharan (2006) report a survey in which 27 per cent of teachers in
endowed with money for the natural occurring donation opportunity, one can figure out the purpose of such an event. For these reasons, at least some concern is present in believing the claims the authors make for the consistency seen in their study.
Baran, Sapienza & Zingales (2009) conducted a similar experiment with MBA students of the Chicago Business School, although they used a Trust Game to measure social preference in an in-class setting. The results on both the game and the donations in the field show a correlation of 0.29, which is quite similar to that of Benz & Meier (2008). Also for Baran et al. (2009), it is seen that subjects display seemingly high levels of pro-social behavior. That is, for both the Trust Game and the donations made to an alumni campaign of the students’ university, roughly 90% of subjects return or donate in the respective experiments. Therefore, care has to be taken into account interpreting the results as a proof for external validity of altruistic behavior. Englmaier & Gebhardt (2016) performed a Public Goods Game in the field and in the lab with students of the University of Munich to test the generalizability of social behavior. In their field experiment they documented the behavior of students with respect to their free riding behavior registering books for the university library. Each student was grouped with three other students to perform the task. All students were payed the same amount of money for the task and were allowed to leave after they registered 680 books combined. Students were unaware of being part of an experiment since the task was framed as a voluntary job at the university. Most of the students taking part in the registering job also participated in a laboratory Public Good game. Results show that students who are less free riding in the library job were also less free riding in the Public Good game. The results were controlled for ability (speed) and show a correlation of 0.40.
Another technique implanted by researchers to test for external validity of experimental outcomes concerning pro-social behavior is the misdirected letter technique. In four experiments, Stoop (2014) was first to test the generalizability of social preferences by means of a pretended misdirected letter. The letters were “intended” to thank volunteers of the University of Tilburg for their services. The stamped and addressed envelopes were transparent to make sure the thank you card plus two notes of five euro were noticed by the receivers. In four different double-blind settings in order to control for subject sample (students vs. citizens) and scrutiny (lab vs. home), receivers face the option of sending the
card - including part or all of the cash – to the rightful owner or keeping the letter themselves. Results show no difference between the four subject pools with respect to generosity. For each setting the behavior is the same: approximately 50% of the subjects return the letter including the ten euros in it. The outcomes suggest that altruistic behavior is unaffected by the experimental context. Franzen & Pointer (2013)4 added on this study by letting subjects play a standard DG and misdirecting them letters – in the same way as it was conducted in Stoop (2014) – on a later date. The obtained results show a moderate association between the choices made in each setting.
Shortcomings of Stoop (2014) and Franzen & Pointer (2013) are that both did not implement a standard Dictator Game to test for external validity concerning (pro-)social behavior. Instead of making use of a windfall endowment, they let their subjects participate in a DG in which the endowment will most likely be perceived as one with property rights for the receiver (i.e. in the misdirected letter setting, the receiver had the “right” to receive the money enclosed in the envelope). Several studies show that perceived recipient rights with respect to the Dictator’s endowment may lead to different outcomes than the standard counterpart of the game (e.g. Cherry, Fryckblom & Shogren, 2002, and Oxoby & Spraggon, 2008). The above shows that evidence for external validity of social laboratory behavior is mixed. However, most studies show at least a moderate correlation between lab and field outcomes. Although we need to consider that it is more likely that studies are published when such a correlation is found and not vice versa. Above that, most studies conduct context-free lab experiments and compare the results with the outcomes of one context-dependent field occasion. The lack of systematic studies offers me the opportunity to investigate the external validity of the DG outcomes further by using a dataset that consists of multiple observations of social behavior of subjects in everyday life. In this way, I can deal with the problem of investigating only one-to-one correlations between a lab game and a particular field setting.
3 Experimental Setup
4 The study of Stoop (2014) was already available in 2010. I preferred to refer to the most recent publication
3.1. Recruitment and background characteristics
Permission was granted to conduct experiments at two primary schools in Hilversum and Nederhorst den Berg, both located at the center of the Netherlands. All children forming classes six, seven and eight at both schools (aged between 10 and 12 years) – and whose parents gave permission to let their children volunteer – participated in the experiment. Parents, teachers and schoolboards were all informed about the basic content and purpose of the experiment. The children were not informed about their task ex ante.
Next to the experimental outcomes on the children’s (anti) social behavior, access was gained to a large database (ParnasSys) that is used by teachers both as a student tracking system and as a student administration system at the two schools. This database provides huge amounts of information on the children ranging from test scores to social behavior scores. Of particular interest for the experiment were the scores on social behavior, which are updated at least four times each schoolyear for each individual – starting from the moment they begin their educational career at one of the two
schools. The experiment has
been performed with 49 children in Hilversum and 36 children in Nederhorst den Berg, which sums up to a total sample of 85 children.
3.2. Methodology
All children were taken one by one from their classes. I told them they were allowed to participate in an experiment. It was made sure that each individual participating in the experiment understand his/her task, by asking them questions on the outcomes of different possible decisions to be made, beforehand. Furthermore, it was intended to hand the children the task without framing an objective of the task at hand, in order to avoid manipulation of the children’s behavior. Above that, participants were instructed that they were not allowed to communicate anything of the experiment to their fellow students after they finished the task and reentered class.
To measure anti-social behavior in the field, I made use of the famous Dictator Game design. Each individual took part in a standard DG and a variation of the game, that is similar to the one introduced by List (2007). The standard game is used to verify robustness of the variation of the game. The game used by
List (2007) offers the advantage of a more symmetrical treatment. That is, dictators are given the opportunity to take all units from another player as well as giving all to the other player. This experimental setup allowed me to get rid of a “moral” cost function, which might be present in asymmetrical Dictator Games (in which dictators are allowed to take less units as they are allowed to give). Also, I made use of different endowments for this experiment than the money that is used in List (2007). Instead, I let the children take, give, or hold their endowment in the form of poker chips. Each poker chip represented a small toy bought in a toy store. After finishing the task, each individual was allowed to collect its earnings according to the outcome of the experiment. Furthermore, instead of using strict Dictators and Receivers, each child played both roles in the two variations of the game. They were paired up with a random classmate, and payoffs were determined by both games – in which they are either a dictator or a receiver. Both the classmate on the receiving end and the decisions made during the experiment were kept anonymous. Earnings were handed out in absence of other children at the end of the day.
In the standard DG, both the Dictator and Receiver were endowed with five poker chips. After receiving these poker chips, three more chips were given to the Dictator. The Dictator then had to decide how to distribute the additional chips. That is, the Dictator chose the amount of chips, s ∈ [0, 3], to give to the Receiver. As a result, the Dictator ended up with 8 - s chips, while the anonymous classmate received 5 + s. Note that s was an integer since it represents the amount of poker chips to be distributed. The payoff domain for both players is [0, 8]. Also for the variation of the game, both players received five poker chips in the initial stage of the game, after which an extra three poker chips were handed to the Dictator. Different from the standard version, the Dictator now had an additional opportunity to take from the other player. More specifically, the action set of the player who was in charge of the outcome of the game comprised s ∈ [-5, 3], in which s were integers again. Consequently, the payoff schemes of both the Dictator, 8 – s, and the Receiver, 5 + s, remained the same. The difference with the standard version of the game – due to the take option – is that the domain of payoffs increased for the Dictator and Receiver, which is now [0, 13].
Due to the access to the ParnasSys database, a large amount of data is available for comparing laboratory behavior and field behavior. Instead of tracking behavior for a single field event, this database provides information on social behavior from the beginning of the school career. Information on several social behavioral traits are listed and updated regularly. For this research the following traits (with description) are of importance:
• Social initiative The ability of a child to get into contact with other children in a competent way. The child is able to approach other children from their own initiative. Also, (s)he is capable to contribute in social gatherings and is skilled in telling by itself.
• Social flexibility The capability of a child to switch flexibly between social situations. (S)he is able to compromise with other children in case of a fight or other evident occasions. Furthermore, (s)he is open minded for new ideas and activities, and has the ability to let go of own ideas or thoughts for other ones. Above that, the child is in possession of a healthy dose of resilience.
• Social autonomy The child is able to form his/her own opinions in social interactions. (S)he knows how to give their opinion, to stand up for their self, to make own choices, and to take care of their own business.
• Impulse control A child’s ability to control and correct its behavior in social intercourse. The child does not speak out of turn, thinks before it undertakes, and is in control if its behavior. In addition, (s)he doesn’t drain too much attention, and acts according to the rules.
• Empathy The ability to understand or feel the thoughts and feelings of another child or person. (S)he is able to adapt to the needs of others. Furthermore, (s)he listens attentively to others, shows interest for others, is being polite to others, and helpful towards others.
All of the social traits mentioned are rated on a scale of 0-100 by the respective teachers for each child. Each teacher is trained to rate social behavior traits for children in a structured way. For these specific schools, teachers follow the guidelines that are included in the ParnasSys system. The ParnasSys system makes use of the Zien! method as a guideline for
rating children on different aspects of their school career (e.g. test scores, social development, lingual skill development, etc.) The Zien! method has been approved by COTAN (Dutch education commission for testing affairs)
3.4. Hypothesis
To provide an answer to the research question whether there exists a correlation between lab and field behavior in a within-subjects design, the following null hypothesis will be tested:
H0: Antisocial behavior in experiments is not predictive for antisocial behavior in the
field
Since papers reporting on lab-field relationships suggesting a (moderate) correlation seem to have the upper hand over studies not reporting a correlation, I expect the alternative hypothesis to be more likely:
H1: Antisocial behavior in experiments is predictive for antisocial behavior in the field
4 Results
First, some summary statistics of the experiment are presented in Table 1. A sum of 85 children, of which 40 girls (46%) from two different primary schools participated in both treatments. 23 children (27%) were ten years old, 30 children (36%) were of age eleven, and 31 children (37%) were aged twelve. The average number of poker chips given to the other player in the standard DG is around 1.20, and ranges from 0 to 3 (which were also the minimum and maximum amount of chips to give, respectively). Girls do not tend to behave differently than boys in the Standard Treatment (Mann-Whitney test, Z = -1.535, p = 0.125). Also, no differences were seen between the different ages.
Likewise, Table 1 also reports the results of the Take Treatment. The average amount of poker chips given is in the range of 1.40. This amount is less compared to the Standard Treatment, which is an expected result given the fact that Dictators are given the opportunity to take chips in the variation of the game. With respect to chip distribution, no significant differences are seen between boys and girls
(Mann-Whitney test, Z = -1.593, p = 0.111). Also here, ranges concerning the minimum and maximum amount of chips by both sexes don’t differ.
Table 1 – Summary statistics, game distributions
category average stand. dev. min. max. N. obs. Standard Treatment All 1.20 0.69 0 3 85 Boys 1.13 0.73 0 3 45 Girls 1.28 0.65 0 3 40 Take Treatment All 0.39 2.20 -5 2 85 Boys 0.44 2.21 -5 2 45 Girls 0.36 2.22 -5 2 40
4.1
Distribution of endowments
The distributions of poker chips between the Dictator and Receiver in the Standard Treatment are listed in Figure 1. We learn that half of the participants display altruistic preferences in the sense that they give one of the possible three poker chips to the other player. Strikingly, almost a third of the children gave more to the other player than to themselves. Only a small fraction of the sample didn’t show any altruistic behavior by giving nothing to a random classmate. Although the results are more or less in line with most studies seen in the literature (with adult participants), a greater part of the children show, on average, a greater amount of other regarding preferences. Normally, two third of the adult participants in a DG show altruistic preferences, in which they tend to give a third of the endowment on average.
Figure 2 shows the result for the Take Treatment. Also in this treatment, most children showed altruistic behavior. More specifically, more than half of the sample gave – in their role of Dictator –at least one chip to the Receiver. Almost the same amount of children gave two chips to the other player compared to the Standard Treatment. The same consequences hold for the Take Treatment with respect to giving two chips away, in that such a decision causes the other player to end up with more chips than the Dictator itself. Approximately a third of the sample gave at most nothing of the additional chips to the other player. The amount of children showing less other-regarding preferences is higher than in the Standard Treatment. The addition of a take option let some children to switch to a less altruistic decision with respect to their decision in the Standard Treatment. This behavior is consistent with that of List (2007). However, the amount of participants displaying such behavior is small in this sample compared to the one in List (2007). In his study, almost half of the sample chose the least altruistic distribution between themselves and another player. In addition, roughly a tenth of the volunteers gave more to the other participant than to themselves. In conclusion with respect to the distribution of
endowments, we see that children used in this study more or less follow the same trend that is seen in preceding literature, although they tend to be significantly more generous toward fellow players. 0 10 20 30 40 50 0 1 2 3 Per c en ta g e Money transferred
4.2
Generalizability of Dictator Games
In order to test whether the Dictator Games outcomes are predictive for real life behavior, a correlation analysis is used to determine a potential relationship. Table 2 contains the correlations between game decisions (i.e. chip distributions) and the social trait scores (Spearman’s ρ). Strikingly, none of the results are as expected initially. That is, no significant positive correlations are present, which would indicate potentially predictive power of (any of) the social trait scores. The experiments with children and the scores on their social behavior in school don’t show any sign of such a relationship. The scores on social traits are to no extent of predictive power for the behavior in Dictator Games in this setting.
Table 2 – Correlations between game decisions and social trait scores (Spearman’s ρ) category Social initiative Social flexibility Social autonomy Impulse control Empathy Standard Treatment All 0.05 0.03 0.03 -0.05 0.02 Boys 0.08 -0.03 0.06 -0.08 -0.05 Girls 0.10 0.06 0.05 -0.10 0.07 Take Treatment All -0.04 -0.06 -0.08 -0.19* -0.19* Boys 0.07 0.07 -0.01 -0.17* -0.11 Girls -0.13 -0.24* -0.14 -0.25* -0.30* 0 5 10 15 20 25 30 35 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 Per c en ta g e Money transferred
*, **, and *** indicate significance levels at 10%, 5% and 1%, respectively.
Most correlation are in the region of +/- 0.10. These results indicate that most decisions are completely random with respect to the social scores available from the database. Only the correlations in the Take Treatment with the behavior during the game and both impulse control and empathy scores are showing a seemingly negative correlation. These will most likely be spurious correlations since only these six results show a moderate relationship. Above that, also these results are not significant, and therefore, do not interfere with the overall conclusion of the absence of any relationship between the social trait scores and the behavior during both variations of the Dictator Game.
4.3
Regression Analysis
In order to test whether the Dictator Games outcomes are predictive for real life behavior, also a regression analysis is performed. This analysis is necessary to find out whether a more social child i is also displaying more altruistic preferences (greater amount of poker chips given to other player) in the Dictator Games:
𝑌𝑌𝑖𝑖 = 𝛼𝛼 + 𝛽𝛽𝛽𝛽𝛽𝛽𝛽𝛽𝛽𝛽𝛽𝛽𝑠𝑠𝑖𝑖+ 𝜃𝜃𝑋𝑋𝑖𝑖 + 𝜖𝜖𝑖𝑖,
where Traits stands for the behavioral traits listed earlier; and X is a vector of attributes (gender and school).
Table 3 shows the regression results. We clearly note that the results are insignificant with respect to the aim of this study. That is, for each social trait the results are not significant. Also, the expected sign for the coefficients is not present for each social trait. More specifically, Empathy and Impulse control have a negative sign, while a positive one was expected (i.e. more altruistic children will also score higher on the social traits). The only significant results are those of Gender and School for the Take Treatment and Standard Treatment, respectively. These are expected to be a spurious relationship, since no theory can explain why girls tend to perform significantly different in the Take Treatment of the Dictator Game, and why children attending the school in Nederhorst den Berg only do significantly different than their fellow primary school pupils in Hilversum in the Standard Treatment.
Standard Treatment Take Treatment Social initiative 0.002 0.007 (0.41) (0.61) Social flexibility 0.001 0.012 (0.21) (0.88) Social autonomy 0.002 0.009 (0.41) (0.82) Impulse control -0.003 -0.011 (-0.76) (-1.09) Empathy -0.000 -0.022 (-0.11) (-1.63) Gender -0.119 -1.453** (-0.69) (-2.86) School 0.294* 0.485 (1.76) (1.01)
Test-statistics between parentheses. *, **, and *** indicate significance levels at 10%, 5% and 1%, respectively.
5 Conclusion and Discussion
In this study, the external validity of the famous Dictator Game is tested in a more systematic way than is seen for most studies, in which they compare one laboratory experiment with one field context. More specifically, data on social behavior of children are gathered from a database used by teachers to store their findings and a take variation of the Dictator Game is used to obtain experimental data. The objective of this study is to see whether the results obtained in the experiment will explain (some) of the behavior displayed by the children during their attendance at their respective schools.
The clear-cut conclusion of this study is that the results of the experiment do not explain the data on every day social behavior of the children to any extent. That is, each of the five social traits are not significantly explained by the behavior displayed in the Dictator Games played by the volunteering children. In contrast with several other studies showing at least a moderate association between lab and field experiments, not even a weak association between the
two different sources of observing and obtaining social behavior is found in this study. Therefore, these results are more in line with the view of Laury and Taylor (2008). They argue that one should be cautious to extrapolate in lab findings on social preferences to less abstract contexts. However, also some caution needs to be taken into account when interpreting the results of this study as well. Although much attention is payed to the experimental method and procedures to conduct an internally valid experiment, this research is also prone to some limitations of concern. Perhaps the most obvious one concerns the experimenter scrutiny problem. For this experiment, children needed to make their decisions clear to me as the experimenter. Despite letting them know that each decision (i.e. each decision on the spectrum of altruistic to selfish) is neither a good one nor a bad one, the children may have felt obliged to make distributions between the other player and themselves more altruistic than desired. Therefore, the results in the Dictator Game could have been biased towards a more altruistic outcome. On the other hand, one can argue that a more social child will also be more aware of his behavior towards the experimenter during the game. In such case, the results will be biased, but the more social child will also display more social behavior during the experiment. Therefore the outcomes may be biased in absolute terms, but not in relative terms. Another
potential limitation to the experimental setting is that children were endowed with so-called “windfall gains”, whereas in real life settings one has to deal with earned money. For instance, Cherry, Frykblom, & Shogren (2002) show that people bargaining over earned wealth are significantly more selfish than other studies show involving the same bargaining game with unearned money. In addition, they show that when people involved in these games face complete anonymity, they are more or less behaving like the theory of self-interested agents suggests them to do (i.e. abandoning selfless behavior). These results of Cherry et al. (2002) implicate for my study that the children may have displayed more social behavior in this setting as they would have shown in a setting with earned wealth. But, also here, one can argue that the more selfish child will be more selfish in either setting (earned vs. unearned money), and therefore the results may not be relatively biased.
A third limitation to this study is that since using children (aged 10-12 years old) for examining the external validity of social behavior in laboratory studies, one should be cautious to extrapolate these findings towards
adults. Nevertheless, Murnighan & Saxon (1998) show that children aged 12 and 15 years, approximately display the same behavior as college students do, although they were to some small extent less selfish. Another study by Harbaugh & Krause (2000) finds that children aged 6 to 12 years participating in a one-shot Public Goods experiment overall display the same amount of altruism as their adult counterparts. In line with these studies, Webley (2005) supports the view that children that are 11-12 years old are shifting into an adult definition of economic understanding (p.62). Although some concerns about the external validity of child studies to adult studies may stay present, the testing of the external validity of a Dictator Game experiment with children with the dataset at hand in this study is on itself interesting to discover.
Another shortcoming of this study is that only one experiment is conducted to test the external validity of laboratory experiments. Although the dataset comprises of a systematic bundle of information on the children’s social behavior outside of the lab, the in-lab information on their behavior consists only of one experiment. This may raise concerns about the internal validity of the study, since chances are – with only conducting one experiment – that a potential relationship found is a spurious correlation instead of a causal correlation (Galizzi & Navarro-Martínez, 2015). Further research could tackle this problem by implementing several social preference games into a study in which these results are compared to real life behavior. Also, Galizzi & Navarro-Martínez (2015) argue that there is no clear one-to-one relationship between a specific in-field situation and an in-lab game experiment, which is another plausible argument for implementing more systematic approaches with respect to the amount of games involving in studies for this subject of interest.
References
[1] Andreoni, J., & Miller, J. H. (2008). Analyzing choice with revealed preference: is
altruism rational? Handbook of Experimental Economics Results, 1, 481-487.
[2] Baran, N., Sapienza, P., & Zingales, L. (2009). Can we infer social preferences
from the lab? Evidence from the Trust Game. Unpublished manuscript, University of
[3] Bardsley, N. (2008). Dictator game giving: altruism or artefact? Experimental
Economics, 11(2), 122-133.
[4] Barr, A., & Zeitlin, A. (2010). Dictator games in the lab and in nature: External
validity tested and investigated in Ugandan primary schools (No. 2010-11). Centre for the Study of African Economies, University of Oxford.
[5] Benz, M., & Meier, S. (2008). Do people behave in experiments as in the field?—
evidence from donations. Experimental economics, 11(3), 268-281.
[6] Camerer, C. (2003). Behavioral game theory: Experiments in strategic interaction.
Princeton University Press.
[7] Card, D., DellaVigna, S., & Malmendier, U. (2011). The role of theory in field
experiments. The Journal of Economic Perspectives, 25(3), 39-62.
[8] Chaudhury, N., Hammer, J., Kremer, M., Muralidharan, K., & Rogers, F. H.
(2006). Missing in action: teacher and health worker absence in developing countries. The Journal of Economic Perspectives, 20(1), 91-116.
[9] Cherry, T. L., Frykblom, P., & Shogren, J. F. (2002). Hardnose the dictator. The
American Economic Review, 92(4), 1218-1221.
[10] Engel, C. (2011). Dictator games: A meta study. Experimental Economics, 14(4), 583-610.
[11] Engelbrecht-Wiggans, R., List, J. A., & Reiley, D. H. (2005). Demand reduction in Multi-unit auctions: Evidence from a sportscard field experiment: Reply. The
American Economic Review, 95(1), 472-476.
[12] Englmaier, F., & Gebhardt, G. (2016). Social dilemmas in the laboratory and in the field. Journal of Economic Behavior & Organization, 128, 85-96.
[13] Fehr, E., & Leibbrandt, A. (2011). A field study on cooperativeness and impatience in the tragedy of the commons. Journal of Public Economics, 95(9), 1144-1155. [14] Fehr, E., & Schmidt, K. M. (1999). A theory of fairness, competition, and
cooperation. Quarterly Journal of Economics, 817-868.
[15] Franzen, A., & Pointner, S. (2013). The external validity of giving in the dictator game. Experimental Economics, 16(2), 155-169.
[16] Galizzi, M. M., & Navarro Martínez, D. (2015). On the external validity of social-preference games: A systematic lab-field study.
[17] Guala, F. (2012). Reciprocity: Weak or strong? What punishment experiments do (and do not) demonstrate. Behavioral and Brain Sciences, 35(1).
[18] Gurven, M., & Winking, J. (2008). Collective action in action: Prosocial behavior in and out of the laboratory. American Anthropologist, 110(2), 179-190.
[19] Haley, K. J., & Fessler, D. M. T. (2005). Nobody's watching? Subtle cues affect generosity in an anonymous economic game. Evolution and Human Behavior, 26, 245–256.
[20] Harbaugh, W. T., & Krause, K. (2000). Children's altruism in public good and dictator experiments. Economic Inquiry, 38(1), 95-109.
[21] Harrison, G. W., & List, J. A. (2004). Field experiments. Journal of Economic
literature, 42(4), 1009-1055.
[22] Kahneman, D., Knetsch, J. L., & Thaler, R. H. (1986). Fairness and the assumptions of economics. Journal of business, S285-S300.
[23] Karlan, D. S. (2005). Using experimental economics to measure social capital and predict financial decisions. The American economic review, 95(5), 1688-1699.
[24] Laury, S. K., & Taylor, L. O. (2008). Altruism spillovers: Are behaviors in
context-free experiments predictive of altruism toward a naturally occurring public
good?. Journal of Economic Behavior & Organization, 65(1), 9-29.
[25] Levine, D. K. (1998). Modeling altruism and spitefulness in experiments. Review of
Economic Dynamics, 1(3), 593-622.
[26] Levitt, S. D., & List, J. A. (2007). What do laboratory experiments measuring social preferences reveal about the real world? The journal of economic perspectives, 21(2), 153-174.
[27] List, J. A. (2007). On the interpretation of giving in dictator games. Journal of
Political Economy, 115(3), 482-493.
[28] Murnighan, J. K., & Saxon, M. S. (1998). Ultimatum bargaining by children and adults. Journal of Economic Psychology, 19(4), 415-445.
[29] Oxoby, R. J., & Spraggon, J. (2008). Mine and yours: Property rights in dictator games. Journal of Economic Behavior & Organization, 65(3), 703-713.
[30] Rabin, M. (1993). Incorporating fairness into game theory and economics. The
American Economic Review, 1281-1302.
[31] Stoop, J. (2014). From the lab to the field: envelopes, dictators and manners.
Experimental Economics, 17(2), 304-313.
[32] Webley, P. (2005). Children’s understanding of economics. Children’s
understanding of society, 43-67.
[33] Zizzo, D. J. (2010). Experimenter demand effects in economic experiments.
6 Appendices
Table 3 – Summary statistics, social traits
category average stand. dev. min. max. N. obs. Social initiative All 76.80 30.68 3 100 85 Boys 5 100 45 Girls 3 100 40 Social flexibility All 69.81 26.32 7 100 85 Boys 13 100 45 Girls 7 100 40 Social autonomy All 73.14 29.76 2 100 85 Boys 2 100 45 Girls 6 100 40 Impulse control All 66.74 31.64 1 100 85 Boys 7 100 45 Girls 1 100 40 Empathy All 76.48 28.34 6 100 85 Boys 14 100 45 Girls 6 100 40
Logbook experiment
• I introduce myself centrally in class and tell the children that I will do a small experiment with each one of them
• I explain them that the experiment will take place outside of the classroom for each kid individually
• After that, I will take place at a small desk outside of the classroom
• On the small desk, I have gathered 13 poker chips: 5 at the side at which the participating child takes a seat; 5 poker chips at the other end of the desk (belonging to the concurrent player); and 3 poker chips at my end of the desk
• Thereafter, I will invite each kid individually to my desk
• Then, I instruct the participating child on the game and make sure s/he understands the task by asking her/him what the consequences are for several possible decisions • I tell the children they play against a random classmate. Furthermore, I tell them
every decision made will be kept anonymously (only I will take notice of each decision)
• I let the first half of the children at each school play game 1 before game 2. The second half of the children play the games in the opposite order
• After playing the game, I let the children know that they are not allowed to communicate anything of the experiment to other classmates before the end of the day
• Now, each child can return to their classroom
• After returning to the classroom by the previous participant, I will invite the next one
• After every volunteering child participated in the experiment, I thank everyone for participating in the experiment centrally
• Then, I will invite each child individually to collect their payments according to their decisions made earlier during the experiment
Instructions Standard Treatment
[Instructions in Dutch. Instructions were given verbally] Welcome (name of child). Today, I want to play a (second) game with you, in which you have to make a decision. Any decision you make is neither a good nor o bad one. It’s all about making the choice of your preference. Every decision you make will be kept exclusively my information. I will not inform any of your classmates or teacher(s) on the decisions you make here today.
I now will instruct you the game. On this desk 5 poker chips are at your end, and at the other end you also see 5 poker chips. The chips at your end belong to you, while the other poker chips belong to a random classmate. I am not going to tell you which classmate. Also, s/he does not know you are playing against him/her. Consequently, s/he also does not know what choices you made between yourself and him/her.
Every chip is worth a present. I do also own another 3 chips myself. I am going to give you these 3 poker chips. You have to decide whether you keep all of the poker chips additionally given to you, or you may decide to give 1, 2, or all of them to your random classmate. When you have made your decision, you divide the chips according to your preference on the desk: your chips on your end of the table, and the other chips at the other end of the table (those belonging to your random classmate). Remember, you are free to make any decision you like.
At the end of the day, you can collect your reward at my desk. Your reward depends on the decisions you made earlier with respect to the amount of poker chips you have distributed yourself. Also, your random classmate receives his/her reward depending on the choices you made.
At the end of the game, I want you to be silent on the game and your decisions towards your classmates when you reenter your classroom!
Instructions Take Treatment
[Instructions in Dutch. Instructions were given verbally] Welcome (name of child). Today, I want to play a (second) game with you, in which you have to make a decision. Any decision you make is neither a good nor o bad one. It’s all about making the choice of your preference. Every decision you make will be kept exclusively my information. I will not inform any of your classmates or teacher(s) on the decisions you make here today.
I now will instruct you the game. On this desk 5 poker chips are at your end, and at the other end you also see 5 poker chips. The chips at your end belong to you, while the other poker chips belong to a random classmate. I am not going to tell you which classmate. Also, s/he does not know you are playing against him/her. Consequently, s/he also does not know what choices you made between yourself and him/her.
Every chip is worth a present. I do also own another 3 chips myself. I am going to give you these 3 poker chips. You have to decide whether you keep all of the poker chips additionally given to you, or you may decide to give 1, 2, or all of them to your random classmate. In addition, you also have the choice to take 1-5 chips of the initial stack of your classmate and distribute them to yourself. When you have made your decision, you divide the chips according to your preference on the desk: your chips on your end of the table, and the other chips at the other end of the table (those belonging to your random classmate). Remember, you are free to make any decision you like.
At the end of the day, you can collect your reward at my desk. Your reward depends on the decisions you made earlier with respect to the amount of poker chips you have distributed yourself. Also, your random classmate receives his/her reward depending on the choices you made.
At the end of the game, I want you to be silent on the game and your decisions towards your classmates when you reenter your classroom!
