Team Behavior in Public Goods Games with
Ostracism
Stephan Huber
Jochen Modely
Silvio Städterz
May 31, 2014
Abstract
In contrast to many real world situations, economic models of decision making usually refer to individual rather than team decision
makers. Thus a growing body of, mostly experimental, literature is
dedicated to the question of team decision making. Our paper aims
at contributing to this literature. We study individuals and teams in
a public goods game with ostracism - i.e. the possibility to exclude
non cooperators from future bene…ts of the public good.
We …nd, that the ostracism mechanism works in increasing the
contribution to the public good of both individuals and teams. Moreover, we …nd teams earning signi…cantly more than individuals due to
a di¤erence in using the punishment mechanism.
JEL Classi…cation: H41
Keywords: team decision; public good; ostracism; experiment
Department of Economics, University of Regensburg, Germany.
Department of Economics, University of Regensburg, Germany.
z
Department of Economics, University of Regensburg, 93040 Regensburg, Germany,
Phone: +49-941-4632716, E-mail: [email protected] (Corresponding author).
y
We are grateful to the German Federal Ministry of Education and Science (BMBF) for
funding the experiments reported here through the project ECCUITY (Economics of
Climate Change: Distribution, E¢ ciency and Policy under Uncertainty). Moreover, we
thank Lutz Arnold, Wolfgang Buchholz and Andreas Roider as well as the participants of
the 5th Annual Meeting of the French Experimental Economics Association (ASFEE) for
their helpful comments. All remaining errors are, of course, our own.
1
1
Introduction
Economic models of decision making usually refer to individual decision makers. Real world situations, however, are often characterized by team decisions
rather than individual decision makers. In the family, in the workplace, in
the management of …rms and, of course, also in governments, people interact
with others to come to decisions. Thus, a growing body of, mostly experimental, literature is dedicated to questions of team decision making.
The provision of public goods is another key aspect in real life situations.
There are lots of decisions of indiviuals who not only bene…t the decisionmaker himself, but also other people, e.g. reduction in emissions of greenhouse gases or the decision to pay taxes according to law and thus enable the
state to provide e.g. public services. But as is well known from theoretical
literature, contributing to a public good is in some cases not individual rational and thus the provision of public goods fails or is at least inadequate.
This raises the question, which mechanisms are able to increase contributions to public goods. In the case of tax payments legal threat seems to be
an accepted means to do so. In case of greenhouse gas reduction, a solution
still seems to be far. Thus recent scienti…c research is dedicated to exactly
the question how to increase public goods provision.
Our paper aims at contributing to exactly those two strands of literature.
We study individuals and teams contributing to a public good in an experimental setup. Moreover, we focus on a mechanisms, aiming at increasing
contributions to the public good and thus overcoming the free-rider problem.
This mechanism bases on ostracism, i.e. the possibility to exclude non cooperators from future bene…ts of the public good (Maier-Rigaud et al., 2010).
Our setting bases on a standard public goods game with six individuals which
we extend to a setting with six teams, each of them consisting of two players.
Through a survey at the end of the experiment, we analyze the motivation
of indivuals as well as teams to contribute to the public good. Our …ndings indicate that the ostracism mechanism works in the sence that both,
individuals and teams, contribute more to the public good. Although we
…nd no signi…cant di¤erence between individuals´ and teams´ contributions
(neither without nor with ostracism mechanism), teams´s seem to exclude
other teams less than individuals do. Moreover, we …nd that the ostracism
2
mechanism increases overall welfare in the team treatment compared to the
individual treatment.
The structure of the paper is as follows: In section 2 we give a literature overview, in section 3 we provide our experimental setting. Section 4
gives our working hypotheses and section 5 provides results Finally, section
6 concludes.
2
Literature Review
Marschak and Radner (1972) "[...] de…ne a team as an organization the
members of which have only common interests." They are among the …rst
economists to investigate team decision processes. They point out, that it
hinges on the information structure, whether team decisions are di¤erent
or equal to those of individual decision makers. In their theoretical framework they …nd, that teams will never perform worse than individuals due to
shared information. However, they conclude that, especially when information sharing is costly, team members might not share all information with
their teammates. Groves (1973) investigates how a principal can incentivize
multiple agents to communicate information and thus form a real team. One
of the …rst experimental work on team decisions was the study by Michaelsen
et al. (1989). They indeed …nd, that mostly groups in a cognitive ability task
perform better than individuals. Since then, a growing body of experimental literature explores whether team decisions di¤er from those of individuals (for recent overviews, see Charness and Sutter, 2012 or Kugler et al.,
2012). Three lessons seem to prevail reviewing the experimental literature
on team decisions (Charness and Sutter, 2012): First, what Michaelsen et al.
(1989) already found, teams seem to be more congitively sophisticated. In
the beauty contest game, teams state signi…cantly lower numbers than individuals, the bigger the team the smaller stated numbers (Kocher and Sutter,
2005 and Sutter, 2005). From a game-theoretic viewpoint, teams act more
rational in a sence that they come closer to the individual rational gametheoretic prediction. Other studies …nd teams to make less irrational errors
in experimental games (see e.g. Charness et al., 2010 or Fahr and Irlenbusch,
2011). Second, teams can help to overcome problems of self-control leading
3
to higher motivation and thus higher productivity (see Falk and Ichino, 2006
or, though non-experimental, Bandiera et al., 20101 ). Third, teams may act
more sel…shly and thus decrease social welfare. This result was found in trust
games (Kugler et al., 2007), centipede games (Bornstein et al., 2004) as well
as prisoner´s dilemma games (Charness and Rustichini, 2007). The result
of teams behaving more sel…shly and rational in a game theoretic sence was
also documented by Bornstein and Yaniv (1998) who …nd in an ultimatum
game that teams o¤ered less than individuals. On the acceptance stage of
the game, however, teams were willing to accept less than individuals. This
result seems to be in line with a recent study by Auerswald et al. (2013), who
found that teams in a public goods game with punishment mechanisms punish signi…cantly less than individuals. From an individual rational viewpoint,
rejecting any positive o¤er in an ultimatum game is irrational. If, however,
responders reject a postive o¤er, they might do so to punish the proposer for
a perceived unfair o¤er. Thus, the …ndings of Bornstein and Yaniv (1998)
and Auerswald et al. (2013) on team behavior seem to coincide. The paper
by Auerswald et al. (2013) is probably most related to our work. Moreover,
it is one of the few studies …nding that teams behave more socially than
individuals in a sence that they contribute more to the public good than
individuals do. Another study …nding that teams behave less sel…sh and
more other-regarding is the one by Cason and Mui (1997). They conduct
a dictator game with teams of two players and …nd that the team decision
is mostly driven by the more other-regarding team mate. The …ndings of
Cason and Mui (1997) are, however, challenged by a study of Luhan et al.
(2009) who …nd in an experimental setting with 3-player-teams that the most
sel…sh team mate has the strongest in‡uence. They provide several possible
reasons for their …ndings to di¤er from Cason and Mui (1997), among them
a possible team-size e¤ect (two vs. three team mates) and a communication e¤ect. In the Cason and Mui (1997) study, the team members decided
in face-to-face interaction on their common decision whereas in the Luhan
et al. (2009) study, team members communicated in a computer chat not
meeting each other. Kocher and Sutter (2007) conclude in their team giftexchange study that the decision making process within the team is crucial
1
The study of Bandiera et al. (2010) relates to social ties at the workplace and how they
improve workers´ productivity.
4
to the teams decision and the outcome of the game. This also contributes to
the question why the two studies of Cason and Mui (1997) and Luhan et al.
(2009), although investigating basically the same thing, came up with such
di¤erent results.
Since Samuelson´s (1954) work, we know the problems concerning the private provision of public goods. In a standard public good game (Ledyard,
1995) an individual rational decision maker would not contribute anything
to a public good when its marginal per capita return is smaller than 1, and
would instead try to free-ride on others´ contributions. Economic theorists
have spend a lot of research on mechanisms to overcome the free-rider problem (see e.g. La¤ont, 1987 or, more recently, Falkinger et al., 2000 for an
overview). However, experimental research on the topic has shown, that the
problem of free-riding might be less severe as supposed by theory. There
are at least ambiguous results concerning the contribution of individuals in
public goods games. Much seems to hinge on whether the experiment is one
shot or whether it is repeated several periods. Moreover, marginal per capita
return seems to play an improtant role. Ledyard (1995) provides an overview
of early experimental work on public goods provision. He identi…es several
factors that a¤ects cooperation in public goods experiments. Repetition and
learning e¤ects seem to have a strong impact. Contribution rates decrease,
when the public goods game is played several periods. Increased marginal per
capita return and communication possiblities are among the stronger e¤ects
increasing cooperation. Participants fairness concerns also seem to have an
impact on contribution rates. Theoretical mechanisms to overcome the freerider problem are as well investigated in laboratory experiments. Falkinger
et al. (2000) provide an overview over laboratory experiments testing various
mechanisms in the lab. Moreover, they provide evidence, that the mechanism
proposed by Falkinger (1996) does well to overcome the free-rider problem in
an experimental setup. The Falkinger mechanism is a combination of reward
and punishment, depending on whether the player contributes more or less
than average to the public good. Mechanisms of punishment of non cooperators have intensively been studied. Chaudhuri (2011) provide a recent
literature survey on how to sustain cooperation in public goods games. He
intensively deals with the literature on costly punishment mechanisms, i.e.
5
the punisher will himself be hurt by his punishment decision as he will loose
money by punishing another player. Therefore, the question arises, whether
such kind of punishment is e¤ective in a sence, that it yields higher cumulative earnings of the group. Indeed, research on this topic indicates, that only
low-cost punishment options with high impact have a positive in‡uence on
the groups overall performance (Chaudhuri, 2011). Thus, some recent experimental studies deal with other instruments trying to sustain cooperation in
public goods games. Page et al. (2013) conduct a public goods experiment
in which one person´s punishment is another person´s reward. They …nd,
that such a punishment mechanism leads to increased contributions and increased e¢ ciency in the sense that average earnings of the group members
increase. Moreover, they provide evidence that so called "perverse" punishment2 decreases when implementing a punishment and reward mechanism.
Studying non monetary punishment mechanisms, Cinyabuguma et al. (2005)
and Maier-Rigaud et al. (2010) conduct public goods experiments with expulsion possibilities or ostracism, respectevely. Both studies …nd a signicant
increase in average contributions when the punishment mechanism was implemented.
3
Experimental Design
The experiments were conducted between November 2013 and January 2014
at the University of Regensburg. In total 288 participants were recruited,
most of them students from di¤erent departments. The experiment was programmed and conducted with the software z-Tree (Fischbacher, 2007). Table
1 provides an overview of di¤erent designs and treatments of the experiments.
2
"Perverse" (or "anti-social") punishment refers to the fact, that group members who contribute much to the public good are nevertheless punished by others´ (Chaudhuri, 2011).
6
Table 1
Experimental Design
D e sig n
1
2
Tre a tm e nt o rd e r
N o . o f S e ssio n s
N o . o f S u b je c ts
N o . o f Te a m s
P e rio d 1 - 1 0
P e rio d 1 1 -2 0
In d iv id u a l
N o O stra c ism
O stra c ism
8
48
-
Te a m s
N o O stra c ism
O stra c ism
8
96
48
In d iv id u a l
O stra c ism
N o O stra c ism
8
48
-
Te a m s
O stra c ism
N o O stra c ism
8
96
48
To ta l
288
Concerning the individual treatment and the ostracism mechanism we refer to Maier-Rigaud et al. (2010). In total 96 participants took part in the
experiment for individuals. This experiment consists of 2 treatments, each
of them played for 10 periods. In the …rst ten periods, the participants palyed a standard linear public goods game ("No Ostracism"), in the second
part they again played for ten periods a public goods game with ostracism
("Ostracism"). In order to control for order e¤ects, we randomly assigned 48
subjects to play the experiment in the order described above. The remaining 48 subjects played the treatments in exactly reverse order. Before each
treatment, the participants received instructions 3 .
For the experiments, subjects were randomly assigned to groups of 6 individuals each. They stayed in the same group for the whole experiment.
We thus de…ne a session as one group of six individuals playing in total 20
periods (10 No Ostracism and 10 Ostracism or vice versa). In every period,
they had to decide on how much of an initial period endowment of 10 token to contribute to a public good4 . Participant i´s payo¤ in each period
was calculated according to the following function, commonly known for all
particiants:
6
X
=
10
g
+
0:4
gj ,
(1)
i
i
j=1
with gi representing the contribution to the public good. Thus, where from
a social perspective full contribution would be optimal, from an individual
perspective it is clearly not. Investing nothing in the public good and freeride on other participants´ contributions is the individual payo¤-maximizing
3
4
See Appendix for instructions.
100 token were worth 2.50 e.
7
strategy (Maier-Rigaud et al., 2010). After each period, subjects were informed about the contributions of other members of their group and about
their earnings in this period. Each member of the group was assigned a randomized number (from 1 to 6) in each period, which was common knowledge
for all participants. Thus, we can abstract from any reputation e¤ects. After
information about other participants´ contributions in the "No Ostracism"
treatment, the game proceeded with the following period. In the "Ostracism"
treatment, the participants (individuals or teams) had the chance to give a
vote to one of the other participants. In case 50 per cent or more of the votes
were given to one participant, this participant was excluded from the game
in the following periods. This means, the participant only received ten token
in each of the upcoming periods without being allowed to contribute to the
public good and receive bene…ts of the public good.
192 participants took part in the experiment for teams. Before the experiment each participant was randomly assigned to a team partner. Thus,
in total 96 teams played this experiment. Each team was separated in one
room with one computer. Thus the team members could directly communicate face-to-face and then had to type in their respective team decision5 .
Apart from team building, the experiment for teams was the same as for
individuals. Concerning the earnings, each team member got exactly the
amount the team had earned in the experiment, thus the incentive structure
for a team member was the same as for an individual player.
After the experiments (individuals and teams) we asked each participant
questions concerning some socio-demogra…c factors and some questions regarding their social attitude in general and their strategy they tried to follow
in the experiment. Apart from some linguistic alignments, questionnaires
were the same for indivual and team players.
5
See Luhan et al. (2009) for a discussion of advantages and disadvantages of face-to-face
communication as opposed to indirect forms of communication (e.g. via computer chat)
in experimental settings.
8
4
Working Hypotheses
Following from the literature on teams vs. individual behavior, we set up the
following working hypotheses:
H1: The ostracism mechanism enhances the contributions to the public
good in the individual as well as the team treatment.
As Maier-Rigaud et al. (2010) have shown for individuals, the ostracism
mechanism works in the way, that it increases contributions to the public
good. We see no reason why this should be di¤erent for teams, so we expect
the mechanism to work for teams, too.
H2: Teams´ contributions to the public good are lower than individuals´
contributions.
As is well known, there is a clear game theoretic prediction concerning the
standard public goods game. From an individual perspective contributing
nothing and free-riding on others´ contributions would clearly be optimal.
As we know from the literature, that teams act more in line with game
theoretic predictions, we expect teams´ to contribute less and try to freeride
more on other teams´ contributions.
H3: Teams punish more sophisticated.
This hypothesis needs some more explanation. In our setup, excluding a
group member of the group will not only be harmful to the excluded, but
also to all other group members. This is due to the fact, that an excluded
group member cannot contribute to the public good on the furhter course
of the game. Even if this member would only have contributed one single
token, this would have enhanced the earnings of all other members by 0.4
token. Thus from a rational perspektive it is clearly not optimal to exclude
group members. However, casting a vote might be fruitful in case it does
not lead to exclusion. In that case the cast vote could serve as a signal for
non-cooperators, that their behavior is not accepted in the group and that
they should enhance contributions in order not to be excluded. So if casting
votes does not lead to exclusion, it might be earnings enhancing for the
whole group. According to the literature, teams seem to be more cognitively
sophisticated. Thus we expect them to use the punishment mechnism in a
9
more sophisticated way than individuals.
H4: In the treatment without ostracism, teams earn less than individuals,
in the ostracism treatment teams earn more than individuals.
If teams are more in line with game theoretic predictions and if H1 turns
out to be true, than we should see more earnings in the individual treatment as individuals contribute more to the public good. In the ostracism
treatment, however, two mechanisms are at work: …rst, earnings depend on
contributions. Second, earnings also depend on exclusions. If teams use
the punishment mechanism more sophisticated, this should enhance their
contributions to the public good and thus could compensate the di¤erence
in earnings compared to individuals. Moreover, if teams exclude less, this
would lead to another enhancement in earnings compared to individuals and
thus in the ostracism treatment we expect teams to earn more.
5
Results
Figure 1 provides an overview over average contributions of individuals and
teams for the course of the game. Gross average contributions in the "Ostracism" treatment are calculated based on all participants, i.e. excluded
and remaining. Net average contributions are based only on non-excluded
players. Similar to Maier-Rigaud et al. (2010), we …nd a heavy decline in
contributions of both, individuals and teams for the "No Ostracism" treatment after some periods. Initial contributions are around 6 token for both,
individuals and teams and then fall to below 4 token. The treatment "Ostracism" shows higher average contributions for both, individuals and teams.
Nevertheless, contributions drop also sharply in the course of the game, but
only do so in the last two periods. Thus …gure 1 indicates, that the mechanism seems to work concerning overall contributions, and even in the last
period, contributions with ostracism are higher than without.
10
10
10
Average cont ribut ion
6
8
8
Average cont ribut ion
6
4
4
2
2
1 2 3 4 5 6 7 8 9 1011121314151617181920
period
1 2 3 4 5 6 7 8 9 1011121314151617181920
period
F ig . 1 . A ve ra g e c o ntrib u tio n s p e r p e rio d fo r in d iv id u a ls (le ft) a n d te a m s (rig ht). S o lid lin e : tre a tm e nt " N o O stra c ism " ,
d o tte d lin e : g ro ss c o ntrib u tio n s tre a tm e nt " O stra c ism " , d a sh e d lin e : n e t c o ntrib u tio n s tre a tm e nt " O stra c ism " .
A permutation test for two independent samples (Siegal and Castellan,
1988) of average group contributions of individuals and teams coborates this
…nding. We calculate the average contributions of groups from period 1 to 10
and period 11 to 20 (individuals and teams) of the treatment "No Ostracism"
and compare it with the treatment "Ostracism" (net and gross averages).
The average of periods 1 to 10 as well as from periods 11 to 20 of treatment
"Ostracism" is signi…cantly higher than of treatment "No Ostracism" for
both, individuals and teams. Thus, our …rst working hypothesis is found to
be true.
Concerning contributions to the public good, the …ndings of our study are
depicted in …gure 2. It shows average contributions to the public good over all
groups and all periods for the treatments "No ostracism" and "Ostracism".
As depicted, in the "No Ostracism" treatment, contributions of individuals
are slightly higher than of teams. However, performing an Permutation test
for two independent samples (Siegal and Castellan, 1988) shows no statistical di¤erence in the average contributions. Thus for the "No Ostracism"
treatment, we can reject that teams contribute less than individuals (H2). In
treatment "Ostracism" the situation is slightly di¤erent. As we investigate
group average contributions, we have to distinguish between net average con11
tributions and gross average contributions. The net values are calculated on
the basis of only the remaining active (i.e. non ostracized) group members.
Gross values are calculated based on all group members with ostracized members counting as zero contributions. As depicted in …gure 2, for net average
contributions the situation is similar to the "No Ostracism" treatment. Individuals contributions are slightly higher than teams contributions. However,
again performing the permutation test for two independent samples shows
no signi…cant di¤erence in average contributions. Looking on gross contributions, the results are slightly di¤erent. Here, teams average contributions are
slightly higher than individuals, although not statistically signi…cant, too.
F ig . 2 . G ro u p ave ra g e c o ntrib u tio n s o f in d iv id u a ls (g re y ) a n d te a m s (b la ck ).
Thus we can reject H2 also for the treatment "Ostracism". However, the
di¤erence in the relations of teams and individuals contributions depending
on net or gross values needs some more discussion. Obviously, this di¤erence
points to some di¤erence in the use of the ostracism mechanism of teams and
individuals. As the drop in average contributions is higher for individuals
than for teams, it seems that individuals ostracize more than teams.
To investigate whether teams and individuals show di¤erent punishment
behaviors, we …rst looked for di¤erences in vote casting. To do this, we calculated the share of cast votes (actual cast votes divided by group members)
of all groups over all periods where ostracism was played. We thus received
16 values of share of cast votes for individuals and 16 values for teams. The
mean of all groups of individuals is 0.29, for teams it is 0.24. A permutation
test for two independent samples, however, cannot reject the null hypothesis
that the mean values for individuals and teams are di¤erent. However, if we
12
look at …gure 3, there seem to be di¤erences in the timing of cast votes.
F ig . 3 . A ve ra g e c a st vo te s in g ro u p s o f in d iv id u a ls (so lid lin e ) a n d te a m s (d a sh e d lin e ).
Individuals cast more votes compared to teams in the beginning of the
game, whereas the share of cast votes is higher for teams at the middle and
the end of the game, respectevely. If we do the above mentioned permutation
test periodwise, we …nd that individuals cast signi…cantly for votes in periods
1 and 2 and 11 and 12, respectevely. The null hypotheses of share of cast
votes of individuals to be smaller than that of teams could be rejected for
those periods. In periods 4 and 9 and 16 and 19, teams cast signi…cantly
more votes. Here the permutation test rejects the null hypotheses of teams´
share of cast votes to be smaller than individuals´.
As already mentioned, the worthiness of a cast vote depends on whether
it really leads to exclusion or only serves as a signalling device for low contributors. One could argue, that casting votes earlier, as individuals do, is
worth it because low contributors are then motivated to contribute more at
an earlier stage. From this viewpoint, teams might cast votes too late. On
the other hand, if casting votes leads to exclusion, casting votes earlier is
clearly harmful for the whole group. Thus, we have a look at the actual
exclusions over the course of the game. Figure 4 provides a graph of average
cumulative exclusions in individuals´ and teams´ groups.
13
F ig . 4 . A ve ra g e c u m u la tive e x c lu sio n s o f g ro u p s o f in d iv id u a ls (so lid lin e ) a n d te a m s (d a sh e d lin e ).
Average cumulative exclusions are higher for teams than for individuals
in the course of the game, especially in the …rst periods individuals exclude
more than teams. Thus the …ndings that individuals cast more votes in
the …rst periods also translates into exclusions. As we know, that those
exclusions are costly for the whole group, our hypothesis that teams punish
more sophisticated seems to turn out to be true.
In a further step, we want to check whether the di¤erent use of the punishment mechanism translates into di¤erent earnings of individuals and teams.
As we found no di¤erence in average contributions of teams and individuals neither in treatment "No ostracism" nor in treatment "Ostracism" - but we
…nd di¤erences in exclusions, we expect to …nd di¤erences in earnings, too.
F ig . 5 . G ro u p ave ra g e e a rn in g s o f in d iv id u a ls (g re y ) a n d te a m s (b la ck ).
14
Figure 5 shows average earnings of the group members for individuals
and teams. Teams earn slightly more in the whole experiment. However,
performing a permutation test for two independent samples shows no statistically signi…cant di¤erence. The same holds true for the treatment "No
Ostracism", where individuals earn slightly more, although also not statistically signi…cant. In the treatment "Ostracism", however, earnings di¤er
signi…cantly. In this part of the game, teams earn more than individuals.
Thus our expectations from analyses concerning exclusions are ful…lled, due
to earlier exclusions, individuals earn signi…cantly less than teams. Therefore, our working hypothesis 4 turns out to be partly true. Earnings in the
treatment "Ostracism" are indeed higher for teams, but this does not follow
from higher contributions, but from the pure fact, that teams exclude less
in earlier rounds and thus more group members stay in the game and can
contribute to the public good.
6
Conclusions
We provide evidence from a public goods game with ostracism mechanism,
showing that individuals and teams behave di¤erently. In line with existing
literature on team behavior concerning better cognitive ability of teams, we
…nd them to act more sophisticated in using the punishment mechanism.
However, in contrast to the most experimental literature on team behavior,
we do not …nd teams to act more sel…shly than individuals. Teams do not
contribute signi…cantly less to the public good in our experiment and hence
we also …nd no signi…cant shortfall in overall earnings of teams compared to
individuals in the standard public goods game. In fact, we …nd that teams
earn even more especially in the game with ostracism. This follows from the
fact, that teams exclude other group members later in the game and thus
show more patience with low contributors.
15
References
Auerswald, H., C. Schmidt, M. Thum, and G. Torsvik (2013). Teams punish
less. CESifo Working Paper 4406.
Bandiera, O., I. Barankay, and I. Rasul (2010). Social incentives in the
workplace. The Review of Economic Studies 77 (2), 417–458.
Bornstein, G., T. Kugler, and A. Ziegelmeyer (2004). Individual and group
decisions in the centipede game: Are groups more "rational" players?
Journal of Experimental Social Psychology 40 (5), 599–605.
Bornstein, G. and I. Yaniv (1998). Individual and group behavior in the
ultimatum game: Are groups more "rational" players? Experimental Economics 1 (1), 101–108.
Cason, T. N. and V.-L. Mui (1997). A laboratory study of group polarisation
in the team dictator game. The Economic Journal 107 (444), 1465–1483.
Charness, G., E. Karni, and D. Levin (2010). On the conjunction fallacy in
probability judgment: New experimental evidence regarding linda. Games
and Economic Behavior 68 (2), 551–556.
Charness, G. and M. Sutter (2012). Groups make better self-interested decisions. The Journal of Economic Perspectives 26 (3), 157–176.
Charness, G., L. R. and A. Rustichini (2007). Individual behavior and group
membership. American Economic Review 97(4), 1340–1352.
Chaudhuri, A. (2011). Sustaining cooperation in laboratory public goods
experiments: a selective survey of the literature. Experimental Economics 14 (1), 47–83.
Cinyabuguma, M., T. Page, and L. Putterman (2005). Cooperation under
the threat of expulsion in a public goods experiment. Journal of Public
Economics 89 (8), 1421–1435.
Fahr, R. and B. Irlenbusch (2011). Who follows the crowd - groups or individuals? Journal of Economic Behavior & Organization 80 (1), 200–209.
16
Falk, A. and A. Ichino (2006). Clean evidence on peer e¤ects. Journal of
Labor Economics 24 (1), 39–57.
Falkinger, J. (1996). E¢ cient private provision of public goods by rewarding
deviations from average. Journal of Public Economics 62 (3), 413–422.
Falkinger, J., E. Fehr, S. Gächter, and R. Winter-Ebmer (2000). A simple mechanism for the e¢ cient provision of public goods: Experimental
evidence. The American Economic Review 90 (1), 247–264.
Fischbacher, U. (2007). z-tree: Zurich toolbox for ready-made economic
experiments. Experimental Economics 10(2), 171–178.
Groves, T. (1973). Incentives in teams. Econometrica: Journal of the Econometric Society, 617–631.
Kocher, M. G. and M. Sutter (2005). The decision maker matters: Individual versus group behaviour in experimental beauty-contest games. The
Economic Journal 115 (500), 200–223.
Kocher, M. G. and M. Sutter (2007). Individual versus group behavior and
the role of the decision making procedure in gift-exchange experiments.
Empirica 34 (1), 63–88.
Kugler, T., G. Bornstein, M. G. Kocher, and M. Sutter (2007). Trust between
individuals and groups: Groups are less trusting than individuals but just
as trustworthy. Journal of Economic psychology 28 (6), 646–657.
Kugler, T., E. E. Kausel, and M. G. Kocher (2012). Are groups more rational
than individuals? a review of interactive decision making in groups. CESifo
Working Paper 3701.
La¤ont, J.-J. (1987). Incentives and the allocation of public goods. Handbook
of public economics 2, 537–569.
Ledyard, J. O. (1995). Public goods: A survey of experimental research.
Handbook of Experimental Economics, 111–194.
Luhan, W. J., M. G. Kocher, and M. Sutter (2009). Group polarization in the
team dictator game reconsidered. Experimental Economics 12 (1), 26–41.
17
Maier-Rigaud, F. P., P. Martinsson, and G. Sta¢ ero (2010). Ostracism and
the provision of a public good: experimental evidence. Journal of Economic
Behavior & Organization 73 (3), 387–395.
Marschak, J. and R. Radner (1972). Economic theory of teams, 1972. New
Haven, Yale University Press.
Michaelsen, L. K., W. E. Watson, and R. H. Black (1989). A realistic test
of individual versus group consensus decision making. Journal of Applied
Psychology 74 (5), 834.
Page, T., L. Putterman, and B. Garcia (2013). Voluntary contributions with
redistribution: The e¤ect of costly sanctions when one person’s punishment
is another’s reward. Journal of Economic Behavior & Organization 95, 34–
48.
Samuelson, P. A. (1954). The pure theory of public expenditure. The review
of economics and statistics 36 (4), 387–389.
Sutter, M. (2005). Are four heads better than two? an experimental beautycontest game with teams of di¤erent size. Economics Letters 88 (1), 41–46.
18
Appendix
Experimental Instructions for individuals (english)6
INSTRUCTIONS
GENERAL INSTRUCTIONS
Today you will participate in two di¤erent experiments. Decisions in experiment I don’t have an in‡uence on experiment II. The instructions for
experiment I follow on this page. The instructions for experiment II will be
distributed at the end of experiment I. Every participant receives exactly the
same instructions.
If you read the instructions for both experiments carefully, you will be able
to earn a considerable amount of money. The amount you earn today will
be paid out in cash at the end of experiment II. You receive the pro…t out of
the two experiments and additionally a show up fee of 4 e.
In both experiments earnings are denominated in Taler. Taler are converted
into e at the following exchange rate: 100 Taler = 2.50 e. Your decisions in
the experiments as well as the amount you earn will remain anonymous.
During the experiment all communication with other participants is strongly
prohibited. If there are any questions, please raise your hand. We will answer all questions individually. It is very important that you follow that rule
because otherwise you will be excluded from both experiments and receive
no payment.
6
As the experiments were conducted at the University of Regensburg, original intructions
are in german language. In the following, we provide an english translation. German
instructions available on request.
19
EXPERIMENT I
In this experiment you will be randomly paired into groups of 6 people. Your
group therefore consists of you and 5 other participants. You play 10 rounds
in that group composition.
At the beginning of each of the 10 rounds all participants receive 10 Taler
each. It is your task to divide these 10 Taler between two projects (project
A and project B). You only can choose integer numbers between 0 and 10.
The income from project A corresponds exactly to the amount of Taler that
you invest in that project.
The income from project B corresponds to 40% of the amount of Taler that
are invested in your group on aggregate into project B.
Your income in each round is therefore calculated as follows:
Your income in round t
=
Your investment in project A in round t
+
0.4
(sum of all investments by group members in project B in round t)
Your total income corresponds to the sum of all income received in all rounds.
This sum will be paid out to you in e at the end of experiment II together
with the total income earned in experiment II and the show up fee of 4 e.
20
Afterwards you receive information about the individual investments of the
other group members in project B.
Please note that the participant numbers are remixed each round,
so that a participant will not be listed under the same number each
round.
Please raise your hand if you have any remaining questions.
21
EXPERIMENT II
Experiment II is a variation of Experiment I. You will remain in the same
group of 6 people in which you participated in experiment I, so your 5 group
members are identical to those in experiment I. Again you play 10 rounds in
the same group composition.
At the beginning of each of the 10 rounds all participants receive 10 Taler
each. It is your task to divide these 10 Taler between two projects (project
A and project B). You only can choose integer numbers between 0 and 10.
The income from project A corresponds exactly to the amount of Taler that
you invest in that project.
The income from project B corresponds to 40% of the amount of Taler that
are invested in your group on aggregate into project B
Your income in each round is therefore calculated as follows:
Your income in round t
=
Your investment in project A in round t
+
0.4
(sum of all investments by group members in project B in round t)
Afterwards you receive information about the individual investments of the
other group members in project B.
Now you have the option to vote for a participant in your group in order
to expel him. You can vote by clicking on the respective …eld next to the
participant. Please note that the participant numbers are remixed
each round, so that a participant will not be listed under the same
number each round.
Please note as well that you can neither vote for yourself nor for someone
that already was expelled (the program will ask you to choose again).
22
Expulsion from the group is based on the following rules:
Actual group size
Minimum number of votes
6
3
5
3
4
2
3
2
2
1
For example, if with an actual group size of 5 participants one of these participants receives 3 or more votes, he is expelled from the group. Expulsion
from the group implies that the expelled participant continues to
receive his 10 Taler per round, but he neither has to make an investment decision nor will he receive any income out of project B.
After the votes are cast, the amount of votes that each individual participant
received will be displayed in the following screen.
The amount you are paid is the sum of the total income from experiment I
and the sum of all incomes of each round of this experiment (= total income
experiment II). The amount converted from Taler into e is paid out to you
23
in addition to the show up fee of 4 e without other participants knowing the
amount or the decisions you made during both experiments.
Please raise your hand if you have any remaining questions.
24
Experimental Instructions for teams (english)
INSTRUCTIONS
GENERAL INSTRUCTIONS
Today you will participate in two di¤erent experiments. Decisions in experiment I don’t have an in‡uence on experiment II. The instructions for
experiment I follow on this page. The instructions for experiment II will be
distributed at the end of experiment I. Every participant receives exactly the
same instructions.
At the beginning of the experiment another participant will be randomly
assigned to you. You will make decisions as a team. The composition of the
teams remains unchanged for the whole experiment process.
If you read the instructions for both experiments carefully, you will be able
to earn a considerable amount of money. The amount you earn today will be
paid out in cash at the end of experiment II. You receive exactly the pro…t
the team you belong to earned in the two experiments and additionally a
show up fee of 4 e.
In both experiments earnings are denominated in Taler. Taler are converted
into e at the following exchange rate: 100 Taler = 2.50 e. Your decisions in
the experiments as well as the amount you earn will remain anonymous.
During the experiment only communication with your team partner is allowed. All communication with other participants is strongly prohibited. If
there are any questions, please raise your hand. We will answer all questions
individually. It is very important that you follow that rule because otherwise
you will be excluded from both experiments and receive no payment.
25
EXPERIMENT I
In this experiment your team and 5 additional teams, which all consist of two
participants, form a group. You play 10 rounds in that group constellation.
At the beginning of each of the 10 rounds all teams receive 10 Taler each.
It is your task to make a decision jointly with the person assigned to you to
divide these 10 Taler between two projects (project A and project B). You
only can choose integer numbers between 0 and 10.
The income from project A corresponds exactly to the amount of Taler that
you invest in that project.
The income from project B corresponds to 40% of the amount of Taler that
are invested in your group on aggregate into project B.
The income of your team in each round is therefore calculated as follows:
Your income in round t
=
Your investment in project A in round t
+
0.4
(sum of all investments by group members in project B in round t)
The total income of your team corresponds to the sum of all income received
in all rounds. This sum will be paid out to you and the person assigned to
you in e at the end of experiment II together with the total income earned
in experiment II. Additionally everybody receives a show up fee of 4 e.
26
Afterwards you receive information about the investments of the other group
members in project B.
Please note that the numbers are remixed each round, so that a
team will not be listed under the same number each round.
Please raise your hand if you have any remaining questions.
27
EXPERIMENT II
Experiment II is a variation of Experiment I. The composition of the group
is identical to that of experiment I. Again you play 10 rounds in the same
group constellation.
At the beginning of each of the 10 rounds all teams receive each 10 Taler.
It is your task to make a decision jointly with the person assigned to you to
divide these 10 Taler between two projects (project A and project B). You
only can choose integer numbers between 0 and 10.
The income from project A corresponds exactly to the amount of Taler that
you invest in that project.
The income from project B corresponds to 40% of the amount of Taler that
are invested in your group on aggregate into project B.
The income of your team in each round is therefore calculated as follows:
Your income in round t
=
Your investment in project A in round t
+
0.4
(sum of all investments by group members in project B in round t)
Afterwards you receive information about the individual investments of the
other group members in project B.
Now you have the option to vote for a team in your group in order to expel
it. You can vote by clicking on the respective …eld next to the team. Please
note that the team numbers are remixed each round, so that a
team will not be listed under the same number each round.
Please note as well that you can neither vote for your team nor for a team
that already was expelled (the program will ask you to choose again).
28
Expulsion from the group is based on the following rules:
Actual group size
Minimum number of votes
6
3
5
3
4
2
3
2
2
1
For example, if with an actual group size of 5 teams one of these teams
receives 3 or more votes, it is expelled from the group. Expulsion from
the group implies that the expelled team continues to receive its 10
Taler per round, but it neither has to make an investment decision
nor will it receive any income out of project B. After the votes are
cast, the amount of votes that each team received will be displayed in the
following screen.
The amount your team earned is the sum of the total income from experiment
I and the sum of all incomes of each round of this experiment (=total income
experiment II). The amount converted from Taler into e is paid out to you
and the person assigned to you in addition to the show up fee of 4 e without
29
other participants knowing the amount or the decisions you made during
both experiments.
Please raise your hand if you have any remaining questions.
30