Sustaining Cooperation in Laboratory Public Goods Experiments: A Selective Survey of the Literature

Abstract

I survey the literature post Ledyard (Handbook of Experimental Economics, ed. by J. Kagel, A. Roth, Chap.2, Princeton, Princeton
University Press, 1995) on three related issues in linear public goods experiments: (1) conditional cooperation; (2) the role of costly monetary
punishments in sustaining cooperation and (3) the sustenance of cooperation via means other than such punishments. Many participants
in laboratory public goods experiments are “conditional cooperators” whose contributions to the public good are positively
correlated with their beliefs about the average group contribution. Conditional cooperators are often able to sustain high
contributions to the public good through costly monetary punishment of free-riders but also by other mechanisms such as expressions
of disapproval, advice giving and assortative matching.

KeywordsPublic goods–Conditional cooperation–Monetary punishments–Non-monetary punishments–Moral suasion–Sorting

Full text

Exp Econ
DOI 10.1007/s10683-010-9257-1
Sustaining cooperation in laboratory public goods
experiments: a selective survey of the literature
Ananish Chaudhuri
Received: 29 June 2009 / Accepted: 14 September 2010
© Economic Science Association 2010
Abstract I survey the literature post Ledyard (Handbook of Experimental Eco-
nomics, ed. by J. Kagel, A. Roth, Chap. 2, Princeton, Princeton University Press,
1995) on three related issues in linear public goods experiments: (1) conditional co-
operation; (2) the role of costly monetary punishments in sustaining cooperation and
(3) the sustenance of cooperation via means other than such punishments. Many par-
ticipants in laboratory public goods experiments are “conditional cooperators” whose
contributions to the public good are positively correlated with their beliefs about the
average group contribution. Conditional cooperators are often able to sustain high
contributions to the public good through costly monetary punishment of free-riders
but also by other mechanisms such as expressions of disapproval, advice giving and
assortative matching.
Keywords Public goods ·Conditional cooperation ·Monetary punishments ·
Non-monetary punishments ·Moral suasion ·Sorting
JEL Classification C71 ·C91 ·C92
1 Introduction
I provide an overview of developments in the experimental literature on linear public
goods games since the survey undertaken by Ledyard (1995).1,2I will structure my
1The dilemma inherent in this game is highlighted by Olson (1965). See Andreoni (1988), Isaac et al.
(1984,1985), Isaac and Walker (1988a,1988b), Kim and Walker (1984) and Marwell and Ames (1979,
1980,1981) for some of the early experimental studies in this area.
2Ledyard (1995, p. 112) provides a succinct description of how a generic linear continuous public goods
game is implemented. Most readers will be well acquainted with this. Here I provide a brief description
A. Chaudhuri ()
Department of Economics, University of Auckland,
655 Owen G Glenn Building Level 6, 12 Grafton Road, Auckland 1142, New Zealand
e-mail: a.chaudhuri@auckland.ac.nz

A. Chaudhuri
survey around three distinct but related ideas. These are: (1) findings regarding the
heterogeneity of social preferences among participants and in particular the concept
of conditional cooperation; (2) the use of costly monetary punishments in sustaining
cooperation in this game and (3) the fact that cooperation can also be sustained via
mechanisms such as expressions of disapproval, advice giving and assortative match-
ing.3
1.1 Where things stood circa 1995
Ledyard (1995) starts by providing an overview of the extant experimental literature
on public goods games till the mid 1990s starting with the influential early work un-
dertaken in the area by Bohm (1972,1983) as well by Robyn Dawes and John Orbell
and their colleagues (for instance Dawes 1980; Dawes et al. 1977,1986;Orbellet
al. 1990), by Gerald Marwell and Ruth Ames (for instance Marwell and Ames 1979,
1980) and by Mark Isaac and James Walker (such as Isaac and Walker 1988a,1988b)
and their colleagues as cited above. The findings of this line of work suggested that:
(1) In one-shot versions of the public goods game, there is much more contribution
than predicted in the Nash equilibrium of the game. Groups of participants on av-
erage contribute between 40% and 60% of the optimal level with wide variations in
individual contributions ranging from 100% contribution by some to 0% by others;
but (2) if the players interact repeatedly over a number of rounds then contributions
often start out at between 40% and 60% of the social optimum and decline steadily
over time as more and more players choose to “free ride.”
Ledyard then goes on to identify a number of factors that enhance cooperation
which include (1) communication, particularly that dealing directly with the problem
at hand; (2) the inclusion of a threshold and/or provision point and (3) the magnitude
of the MPCR; a higher MPCR increases contributions. He also discusses other factors
that might be expected to play a role but have little effect overall such as gender or
mostly as a reminder of terms and concepts that I will use in the rest of the paper. In a one-shot play
of the game, a group of nparticipants is told that each of them has an endowment of ωtokens. Each
participant imust make a contribution decision C(0≤C≤ω) on how many of those tokens she wants
to contribute to a public account. Any remaining tokens are allocated to the private account. Contributions
by all the participants in a group are made simultaneously, without any communication and typically in
whole tokens. In addition to the tokens allocated to the private account, each participant ireceives a fixed
percentage (α) of the total group contribution to the public account, where 0 <α<1<nα.Theterm
αis referred to as the marginal per capita return (MPCR) from the public good. At the end of a round
participants either get to see the individual contributions made by members of the group or the total (and
therefore average) contributions to the public account without learning the identity of the group members.
Each participant’s personal earning is the sum of the tokens kept in the private account plus the return
from the public account. If the game is repeated finitely then this one-shot game is played over a number
of rounds where each successive round proceeds in the same manner, starting with a new endowment of ω
for each participant.
3Even when coverage is restricted to the period after 1995, the literature is still vast. I will focus only
on those papers that look at linear continuous public goods games. I will not be discussing papers in a
number of related areas such as those that analyze public goods with a provision point, prisoner’s dilemma
games or common-pool resource usage games. I will mostly focus on published papers, but I will discuss
a handful of unpublished papers because I thought that these papers were innovative enough to warrant
discussion.

Sustaining cooperation in laboratory public goods experiments
the size of the group; contrary to intuition larger groups are no worse—and may even
be better—at providing the public good than smaller ones. And in an anticipation of
the literature that followed he also presents models of behavior which incorporate
reciprocal motivations and beliefs about others’ contributions.
1.2 What have we learned since then?
At the expense of over-generalization I suggest that advances have been made on two
broad fronts. The first of these is a greater understanding of the fact that there are dis-
tinct types of players in such games who differ in their social preferences and/or their
beliefs about their peers, either one of which may be sufficient to generate behavior
not commensurate with the simple game theoretic prediction of free riding in such
situations. The most notable finding in the area is that many participants behave as
“conditional cooperators”, whose contribution to the public good is positively corre-
lated with their beliefs about the contributions to be made by their group members.
This idea that there may be different types of players certainly did not arise de
novo and was foreshadowed in Ledyard’s concluding section and in other previous
studies. But there had not been a systematic exploration of the motivations behind
these types till that point. At the time Ledyard was writing his survey it was also
not entirely clear as to what factors lay behind the usual pattern of decaying contri-
butions. This was variably attributed to kindness on the part of some and confusion
on the part of others (Andreoni 1995), the “warm glow” of giving (Andreoni 1990),
a combination of learning to play the dominant strategy and strategic play by self-
interested players (Andreoni 1988; Andreoni and Croson 2008) or decision errors of
various types (Palfrey and Prisbrey 1997 or Anderson, Anderson et al. 1998). Led-
yard ends this particular section of his survey by suggesting that more research is
needed to understand the reasons behind contributions decay. We have now come to
realize that the usual decaying pattern of contributions can be better understood by
appealing to heterogeneity in the types of players interacting with one another.
Furthermore, recent experimental research has discovered that conditional cooper-
ators are often willing to engage in punishment of free-riders even when such punish-
ment is personally costly and confers no long-term benefits. They are also often suc-
cessful in sustaining high contributions to the public good even without such costly
punishments via other mechanisms which can broadly be categorized as moral sua-
sion (to borrow a term from Ledyard 1995) and/or assortative matching.4See for
instance Axelrod (1986), Fehr and Fischbacher (2004a,2004b,2005a,2005b), Fehr
and Gächter (2000,2002), and Bowles and Gintis (2002) among others. Experimental
economists have in turn used their understanding of such type heterogeneity to design
better institutions that can help sustain cooperation. This is the part of the literature
that I focus on in the rest of this article.
4For a shorter review of some of the issues discussed here—such as conditional cooperation and the notion
of altruistic punishments—see Gächter and Thöni (2007).

A. Chaudhuri
The second major area of advance has been the development of theoretical models
of behavior in such games based on experimental findings in the area.5These have in
turn prompted further experiments designed to test the implications and applicability
of these models to other games and situations. There are two distinct types of models
that have been developed to deal with such issues: (1) models that focus on distribu-
tional concerns such as those in Fehr and Schmidt (1999) and Bolton and Ockenfels
(2000); and (2) intentions based models that focus on participants’ beliefs about each
others’ actions and a concern for reciprocity. The latter class of models include Rabin
(1993) and Dufwenberg and Kirchsteiger (2004). Models that combine elements of
both include Charness and Rabin (2002), Cox et al. (2007,2008) and Falk and Fis-
chbacher (2006). A recent theoretical paper that does not quite belong to either group
is Ambrus and Pathak (2009).
Models incorporating distributional concerns such as Fehr and Schmidt (1999)as-
sume that players are “inequity averse” in the sense that they experience “guilt” if
they receive a payoff that is higher than others (advantageous inequity) while they
experience “envy” if they receive a payoff that is smaller (disadvantageous inequity).
Utility functions are linear in the difference between these payoffs with disadvan-
tageous inequity leading to higher losses in utility than advantageous inequity. The
Bolton and Ockenfels model also assumes that players care about relative payoff but
allows for non-linearity in the utility function. However, the results presented in Char-
ness and Rabin (2002) among others raise questions about the conclusions or broad
applicability of such inequity aversion based models.
On the other hand, studies like Rabin (1993), Dufwenberg and Kirchsteiger (2004)
and Falk and Fischbacher (2006) develop models that explicitly incorporate first and
higher order beliefs—beliefs of player jabout the strategies to be adopted by player
kas well as beliefs of player jabout the beliefs of player kabout the strategies to be
adopted by player jand so on—directly into the utility function.
A recent theoretical paper by Ambrus and Pathak (2009) extends the earlier
work by Kreps et al. (1982) in explaining cooperation in finitely repeated prisoner’s
dilemma games. These authors assume that there are two kinds of players—purely
self-interested and reciprocal. The actual proportion of each type is common knowl-
edge and hence there is no asymmetric information about types. The main feature of
the model is a reciprocity function whose arguments include past and current con-
tributions of other group members. The authors show that as long as this reciprocity
function obeys certain regularity conditions, then there is a unique sub-game perfect
equilibrium in which contributions drop off over time. The model is able to explain a
number of observed regularities in behavior including the “re-start effect”.
In the interests of parsimony, I refrain from further elaborating on these theoretical
models, which makes this survey less than exhaustive. The rest of this paper is struc-
tured in the following way. Section 2discusses the issue of conditional cooperation.
Section 3examines the role of costly monetary punishments. Section 4looks at how
5These theoretical advances quite possibly deserve a separate survey of their own at this point. The chapter
on other-regarding preferences by Cooper and Kagel (2009) forthcoming in the second volume of The
Handbook of Experimental Economics discusses some of these models.

Sustaining cooperation in laboratory public goods experiments
mechanisms other than monetary punishments can also sustain cooperation among
either sorted or non-sorted groups. Section 5concludes.6
2 Conditional cooperation
In this section I focus on experimental studies that explore the phenomenon of “con-
ditional cooperation”.7If we assume standard preferences then free-riding is the dom-
inant strategy equilibrium in one-shot plays of the game; free-riding is also the sub-
game perfect equilibrium in finitely repeated games and evolutionarily stable. How-
ever alternative models of social preferences that incorporate beliefs and/or inequity
aversion show how conditionally cooperative behavior can emerge in this game.
Rabin (1993), building on the pioneering work of Geanakoplos et al. (1989) de-
velops the concept of “fairness equilibrium” which explicitly incorporates the beliefs
that players hold about others’ actions. Rabin shows that once we allow for reciprocal
motivations on the part of the players, then it is possible to think of the public goods
game as a coordination problem with full contribution being an efficient equilibrium
and full free-riding an inefficient equilibrium with other equilibria in between.
One drawback of Rabin’s model is that it does not apply to sequential move games
as in finitely repeated linear public goods games. Both Dufwenberg and Kirchsteiger
(2004) as well as Falk and Fischbacher (2006) provide a generalization of Rabin’s ap-
proach and develop the concept of “sequential reciprocity” which extends the notion
of conditional cooperation to finitely repeated games. The difference between these
papers lies in the fact that (1) Falk and Fischbacher incorporate both distributional
concerns as well as beliefs about others’ strategies and (2) Dufwenberg and Kirch-
steigher allow for a more sophisticated Bayesian belief updating rule with players
maximizing utility at each node of the game while in Falk and Fischbacher (2006)
only initial beliefs enter into the utility function.
Conditional cooperation can also arise if participants are inequity averse as in
Fehr and Schmidt (1999) or Bolton and Ockenfels (2000). A participant with such
preferences will contribute more if he believes that his peers will contribute more due
to his concern for equity in payoffs.
There are a number of studies that provide evidence of such conditional behavior
using a variety of different means. One of the first direct experimental tests of this
type of conditional cooperation is provided by Fischbacher et al. (2001), henceforth
referred to as FGF.8Given that this is a paradigmatic and widely replicated study,
I will dispense with a detailed discussion.
6Many of the papers discussed combine multiple mechanisms and could fit into more than one section. In
such cases I have included those papers in the section that I deemed most appropriate.
7For another review of the phenomenon of conditional cooperation using both lab and field experiments,
see Gächter (2007). This paper also discusses some policy implications of conditional cooperation in the
areas of tax evasion, tax morale, contributions to charity, etc.
8Probably, the earliest study examining the concept of conditional cooperation is Kelley and Stahelski
(1970) which focuses on the phenomenon in the context of prisoner’s dilemma games. However, Bryan
and Test (1967) can also be construed as a study that examines the phenomenon of conditional cooperation
as embodied in the decision to help only when others are doing so, except Bryan and Test do not elicit any
information regarding beliefs.

A. Chaudhuri
FGF find that 50% of their participants are conditional cooperators. However the
slope of the contributions profile for these participants lies below the 45 degree line.
This implies that while these participants are willing to match the average contribu-
tion expected of others in the group, they do not quite match the group average dollar
for dollar, but exhibit a small amount of “self-serving” bias. FGF go on to argue that
it is the heterogeneity in player types that provides a rationale for why contributions
decay over time. They suggest that any given unsorted group of participants in an
experiment consists of both conditional co-operators and free-riders. Conditional co-
operators with optimistic beliefs regarding the contributions to be made by their peers
will contribute to the public account. But over time as they begin to discover the het-
erogeneity in types and particularly the presence of free-riders they will reduce their
contributions leading to the decaying pattern in contributions.
However, a number of other studies also explore such conditionally cooperative
behavior and were published within close proximity of each other and of FGF. In all
likelihood the first authors to actually use the term “conditional cooperation” were
Sonnemans et al. (1999). They extend the classic “partners” versus “strangers” par-
adigm introduced by Andreoni (1988). Participants play for 36 rounds in groups of
four. Group composition remains constant for a minimum of three and a maximum of
12 rounds. Groups change only gradually, with at most one subject at a time leaving
the group with replacement and the timing of departure is common knowledge. This
design feature is crucial because a subject leaving a group is guaranteed not to inter-
act with his former group members again and so there is no incentive to engage in
strategic behavior in the last period before leaving the group, while strategic behavior
is still possible in other periods. Hence, this design is a combination of a partners and
a “real” strangers treatment. It enables a comparison “between subjects” by looking
at participants who leave a group and those who stay and “within subjects” by ana-
lyzing how the same person behaves in the last period in an old group and the first
period in a new group.
The authors find evidence of strategic behavior in that there is a sharp decline in
contributions in the last period prior to leaving a group but they also find evidence of
conditional behavior in that subjects who expect others to contribute also contributed
themselves.
Keser and van Winden (2000) replicate Andreoni’s (1988) study using the “part-
ners” versus “strangers” paradigm. Their main contribution lies in recognizing that
participants do not behave either as free-riders or as altruists but rather in an inher-
ently conditional manner. Participants use information about the average group con-
tributions as an anchor for their own future contributions. In both treatments, about
80% of the participants behave in a conditional manner; those who are above (be-
low) the average in one round decrease (increase) their contribution in the following
round.9
9Croson (2007) explores different motivations behind the desire to contribute to a public good. These in-
clude theories of commitment, altruism and reciprocity. Croson suggests that participants are primarily
motivated by reciprocal tendencies. Furthermore, when participants are shown the distribution of the con-
tributions made by other members of the group, the data suggests that the participants try to match the
median or the mean contribution in the group, rather than the minimum or the maximum.

Sustaining cooperation in laboratory public goods experiments
Brandts and Schram (2001) use a “contribution function” approach similar to the
one developed by Palfrey and Prisbrey (1997). Here the returns from the private ac-
count and public accounts varies so that for some values of the MPCR the dominant
strategy is to free-ride while for others it is to contribute the entire endowment.10
Participants are expected to switch from the latter to the former at a particular value.
Brandts and Schram argue, contrary to Palfrey and Prisbrey (1997), that decision
errors cannot be the primary driving force behind cooperation and that some partic-
ipants behave in a conditionally cooperative manner while some others seem to be
driven by self-interest. It is the interaction between these two groups that is important
in explaining the temporal pattern of contributions.
It should be noted that FGF’s implementation of the public goods game is different
from the usual approach, as in Sonnemans et al. (1999) or in Keser and van Winden
(2000), in the sense that FGF rely on the “strategy method” (Selten 1967).11 One
way to think about the difference between these two approaches is the following.
Keser and van Winden (2000) study reciprocal behavior by looking at the extent to
which participants’ behavior is conditional on the past behavior of their peers, with
all decisions being payoff relevant. Fischbacher et al., on the other hand, look at
how participants respond to their own prior beliefs about the behavior of their peers.
In the latter case, some decisions will not affect the participants’ payoff. There is
controversy regarding the consistency between the “hot” responses elicited directly
by observing responses to others’ behavior as opposed to the “cold” responses elicited
via hypothetical questions. Brandts and Charness (2009), following up on their earlier
work in Brandts and Charness (2000), analyze a number of studies that compare such
direct versus indirect elicitation of responses and conclude that by and large these
two types of responses are consistent more often than not.
Fischbacher and Gächter (2009,2010) extend the FGF study by asking: when a
particular participant indicates that he would contribute more if his peers in the group
did so too on questionnaires of the type used by FGF, do those responses match actual
behavior when playing the same game for money? They analyze participants’ prefer-
ences using two different experimental treatments. In the P-experiment participants
first play a one-shot public goods game and then fill out a questionnaire stating how
much they will contribute to the public good conditional on the average contribution
of the other group members. In the C-experiment participants play 10 rounds of a
linear public goods game with random re-matching. At the end of each round partic-
ipants are asked to estimate the other group members’ average contribution. In half
of the sessions participants play the P-experiment and followed by the C-experiment
(P-C treatment). This sequence is reversed in the remaining sessions (C-P treatment).
The authors find that 55% of participants are conditional cooperators while 23%
are free-riders. The results show a positive and stable correlation between the be-
liefs and contributions for conditional cooperators in both the C- and P-experiments.
Participants who are classified as conditional cooperators using the questionnaire ap-
proach in the P-experiment behave in the same way when asked to play for 10 rounds
10Brandts and Schram use the term “marginal rate of transformation” which is the inverse of the MPCR.
11The strategy method, introduced by Selten (1967) collects data by asking participants to respond to
hypothetical questions. Not all of the participant responses may be payoff relevant. This has the advantage
that it allows the experimenter to collect large quantities of data.

A. Chaudhuri
in the subsequent C-experiment. The distributions of beliefs in the P-C or the C-P
treatments are not significantly different. Hence eliciting participants’ beliefs after
they have participated in the public goods game does not affect their preferences.
Fischbacher and Gächter go on to argue that it is the so-called “self-serving bias”
in conditional cooperation that leads to contributions decay. Even if an entire group of
participants consists of conditional cooperators, as long as each conditional coopera-
tor tries to contribute a little less than others, this would lead to a fall in contributions
over time. Using simulations that rely on elicited beliefs, actual contribution patterns
and the belief updating rule discovered from the data (beliefs of subject jin period t
are based on a convex combination of j’s beliefs in period t−1 and the contribution
of other group members in period t−1) they show that such “imperfect” conditional
cooperation is at the heart of the decay in contributions over time.12
Like Fischbacher and Gächter (2009,2010), Kurzban and Houser (2005) also look
at the robustness of conditional cooperation using a more elaborate experimental de-
sign. They begin by first classifying participants into types according to their behav-
ior in a linear public goods game. The authors then go on to observe if participants
remain true to type in a different public goods game.
There are four session with 84 participants who take part in a sequence of one-
shot linear public goods games in randomly formed groups of four. Each game starts
with players making a decision regarding how to allocate their endowments between
a private account and the public account. Each contribution decision is followed by a
number of rounds each of which proceeds as follows: first, one player in each group is
provided with the current aggregate contribution to the public account and is afforded
an opportunity to change his allocation to the two accounts. Then the next player is
given the same opportunity and so on. Payoffs to participants in each game are de-
termined by the final allocation of tokens between the private and public accounts at
the point where the game ends. The exact number of rounds that succeed each con-
tribution decision is unknown to the participants. They only know that following the
initial contribution decision each participant will get at least one chance to change her
mind and that there is a 4% probability that the game will end after each subsequent
decision.
Each experimental session contains at least seven such games. This multiple elic-
itation of contribution responses is designed to see if there is attenuation in the ten-
dency to behave in a conditional manner over time. Participants are classified as un-
conditional cooperators, free-riders and conditional cooperators on the basis of a plot
12I should point out that while all groups seem to contain some free riders, albeit a minority, their presence
is not necessary for contributions to decay. For instance a self-serving conditional cooperator with opti-
mistic beliefs who expects group members to contribute 8 tokens might contribute (8 −ε) tokens while
a pessimistic conditional cooperator who expects group members to contribute 2 tokens might contribute
(2 −ε) tokens but to the former the latter will appear as a free-rider leading to an eventual decay in con-
tributions. In fact a recent paper by Chaudhuri and Paichayonvijit (2010b) suggest that the usual pattern
of decay is caused by participants essentially realizing that while many of them behave in a conditional
manner, the distribution of initial beliefs is very different. Those who have optimistic beliefs contribute
more and those with pessimistic beliefs contribute less. Over time the optimists reduce their contribution
while the pessimists actually increase their contributions but the contribution increases coming from the
latter are too small to offset the reductions from the former leading to contributions decaying over time.
This explanation is at odds with the one put forth by Fischbacher and Gächter (2009,2010)orbyAmbrus
and Pathak (2009).

Sustaining cooperation in laboratory public goods experiments
of the participant’s contributions against the average contribution to the public ac-
count he observed before making his own contribution, along the lines of FGF. 63%
are conditional cooperators, 20% are free-riders and 13% are unconditional coopera-
tors.
The authors find that these classifications are stable by having the participants take
part in three additional games and find that those classified as free-riders contribute
less on average than their peers, cooperators more and conditional cooperators about
the same as their group members. Furthermore groups that consist of more coopera-
tors generate higher contributions on average.
Burlando and Guala (2005) also test for the robustness of conditional cooperation
by having participants take part in four different tasks: (1) the “Strategy Method” used
by FGF; (2) the “Decomposed Game Technique” used by Offerman et al. (1996);13
(3) contributions to the public account in a repeated linear public goods game played
for 20 rounds and (4) a questionnaire.14 Overall, the outcomes of the four classi-
fication tasks are consistent. 35% are classified as conditional cooperators, 18% as
unconditional cooperators, 32% as free-riders and the remaining 15% could not be
classified.
Two studies replicate the FGF results with a similar experimental design but a
more heterogeneous participant pool. Kocher et al. (2008) recruit 36 participants,
in groups of three, from three different locations—North Carolina, USA, Innsbruck,
Austria and Tokyo, Japan. The majority of the participants act as conditional coopera-
tors except that there are more of them among the US participants (81%) than among
those in Austria (44%) or Japan (42%). Hermann and Thöni (2008) recruit 160 par-
ticipants at four separate universities spread across Russia, two in rural areas and two
in small cities. They find that overall 56% of participants behave as conditional co-
operators, and only 6% as free riders. The distribution of preferences does not differ
significantly across rural or urban backgrounds and socio-economic conditions do
not seem to have an impact on the preference for conditional cooperation. Brandts et
al. (2004) also undertake a cross-cultural study with participants from Japan, Nether-
lands, Spain and USA, except, rather than using the FGF procedure they use the
13The Decomposed Game technique (Griesinger and Livingston 1973;Liebrand1984) has been widely
used by psychologists and economists in order to measure social values and attitudes towards cooperation.
Participants are asked to make 24 choices between pairs of allocations. Each participant knows that she
has been paired anonymously with another participant and the pairings remain unchanged for the entire
duration of the session. Each allocation consists of a number of tokens paid to the player concerned and
another sum paid to the other player. The token amounts can be positive or negative. A typical choice
may involve, e.g., a combination A=(75,−130)vs. B=(39,−145), where one must choose between
gaining either 75 or 39 tokens, with related losses on the other’s part of either 130 or 145 tokens. There
is no feedback concerning the other player’s choices. The final payoff is obtained by combining the 24
choices of each participant with those of the other player. Players are then classified into one of five pos-
sible categories—altruistic, cooperative, individualistic, competitive and aggressive—depending on their
choices in this allocation task. Offerman et al. (1996) use this decomposed game technique to classify par-
ticipants into separate categories. In their study 61% of participants are classified as individualistic while
27% are deemed cooperative with relatively few people in the other categories. The authors then go on to
investigate the behavior of the different types in a step-level public goods game.
14The questionnaire included questions such as (1) “What were you trying to do in the experiment (in other
words: what were your goals or objectives)?” (2) “Did you achieve your objectives?” (3) “What were the
other members of your group trying to do (what were their objectives)?” (4) “What was the scope of this
experiment (in other words, what were the experimenters trying to discover)?”

A. Chaudhuri
contributions function approach pioneered in Brandts and Schram (2001). They re-
port little differences in contributions across these locations and a recurring finding
of conditionally cooperative behavior.
Given the plurality of conditional cooperators in these studies, one interest-
ing question to ask is what happens when conditional cooperators learn about the
presence of other conditional cooperators in the group. This is what Chaudhuri
and Paichayontvijit (2006) examine. They implement a between subjects treatment,
where in the control treatment participants fill out a conditional cooperation question-
naire similar to FGF and for some participants it is the response on this questionnaire
that is relevant. This is followed by three other treatments, where participants are
provided progressively more detailed information about the presence of conditional
cooperators in the group. The authors find that 62% of participants are conditional
cooperators. But more interestingly, there is an increase in contributions when par-
ticipants are provided information about the presence of conditional cooperators and
this increase is most pronounced for the conditional cooperators themselves.15
2.1 Concluding remarks
The evidence presented in this section suggests that the many participants in linear
public goods games are conditional cooperators whose contributions to the public
good are positively correlated either with their ex ante beliefs about the contributions
to be made by their peers or to the actual contributions made by the same. The behav-
ior of conditional cooperators deviates substantially from the game-theoretic predic-
tion of free-riding in this context. Furthermore, the evidence suggests that conditional
cooperation is a stable preference type that is quite prevalent among participants and
the phenomenon is robust across cultures and the mode of elicitation of responses.
3 Sustaining cooperation by punishing free-riders
Fehr and Gächter (2000) study the efficacy of costly punishments using both “part-
ners” and “strangers” protocol. Participants play for 20 rounds in groups of four—the
first 10 rounds without any punishment possibility and then the next 10 rounds with
punishment.16 In each round a participant has an endowment of 20 tokens and the
MPCR is 0.4. This study has been widely replicated and the experimental design is
15For the sake of completeness I should point out another issue that has been raised in this context. This
has to do with the extent to which conditional cooperation is not a stable preference type but merely
a form of conformism. The tendency to copy the most prevalent behavior among a group is a robust
phenomenon, often because non-conformism can be psychologically painful. See for instance Asch (1951,
1955)andMoscovici(1985) among others. Carpenter (2004) and Bardsley and Sausgrüber (2005) suggest
that conformism may partially explain conditionally cooperative tendencies. But the evidence presented in
this section shows that there is clearly more to conditional cooperation than a mere desire to conform.
16Ya magis h i ( 1986,1988) and Ostrom et al. (1992) also look at the role of punishments in sustaining
cooperation in social dilemmas. While these papers pre-date Fehr and Gächter, as mentioned above, I am
going to focus on the period since 1995 and therefore I will forego a detailed discussion of these earlier
studies.

Sustaining cooperation in laboratory public goods experiments
Tab l e 1 Punishment levels and
associated costs for the
punishing participant in Fehr
and Gächter (2000)
Source: Fehr and Gächter (2000)
Punishment 0 1 2 3 4 5 6 7 8 9 10
Points
Costof 0124691216202530
Punishment
well-known. I will provide a brief summary only, primarily because I need to refer to
aspects of this study when I discuss subsequent papers in this area below.
The second set of ten rounds has two stages in each round. In the first stage, partic-
ipants play a standard linear public goods game. This is followed by a second stage
where players, having seen the contributions of others (but without learning their
identities) can choose to punish the other group members. Each punishment point
reduces the punished participant’s payoff by 10%, so that if a participant receives 10
or more punishment points then this participant’s earnings fall to zero. The punish-
ment is costly to the punisher as well. Table 1shows the cost associated with each
punishment point in terms of tokens.17
Figure 1provides an overview of average contributions in the two treatments with-
out and with punishments. Across all rounds and the two treatments the average con-
tribution to the public good is 19% without punishment and 58% with punishments.
The average contribution in the last round without punishments is 10% but with pun-
ishments the average last round contribution is 62%. Fehr and Gächter (2000)also
find that punishments are primarily aimed at those who contribute less than the group
average in any round and the further below the group average is the participant’s
contribution, the greater is the magnitude of the punishment handed out to this par-
ticipant.18,19
17Studies implementing a punishment mechanism fall into one of two categories. Either they use a “within
subjects” design where the same participant takes part in two treatments, an experimental treatment with a
punishment option and control treatment without punishment. Others use a “between subjects” treatment
where some participants take part in treatments with punishment while others take part in a control treat-
ment without. In what follows, every time I say that this is a “between subjects” treatment, it will imply
that besides the experimental treatments, one set of participants always take part in a control treatment,
where no punishment opportunity is available.
18Fehr and Schmidt’s (1999) model of inequity aversion can justify the use of costly punishments. Given
that free-riders are better off in payoff terms compared to cooperators, the latter may well be willing to
sacrifice a further part of their payoffs to punish the free-riders especially if that reduces the inequity of
payoffs between the two.
19Fehr and Gächter (2002) extend the results obtained by Fehr and Gächter (2000) using a similar within
subjects design except with random re-matching in all sessions. In one treatment participants play 6 rounds
without a punishment option followed by another 6 rounds with the punishment option. This sequence
is reversed in the second treatment. Here each token of punishment reduces the earnings of the punished
participant by 3 tokens. Since participants are randomly re-matched at the beginning of each round theories
of reciprocity or costly signalling cannot explain cooperation and punishment in this environment. The
authors suggest that punishment in this context is “altruistic” in the sense that this punishment is costly to
the punisher but does not generate any immediate benefits for the current group members since groups are
randomly re-constituted at the end of each round. The results show that 84% of participants punish at least
once. Punishments appear to be triggered by negative emotions given that 74% of punishment is imposed
by participants who contribute higher than the group average on those who contribute lower than the group
average.

A. Chaudhuri
Fig. 1 Average contributions without and with punishments in Fehr and Gächter (2000). Source: Fehr and
Gächter (2000)
While the papers by Fehr and Gächter show that costly punishments can indeed
raise contributions to levels above those attainable in the absence of such punish-
ments, in those studies the participants do not get to choose whether punishments
are available or not. Gürerk et al. (2006) analyze contribution behavior in a public
goods game where participants can choose to be in either a sanctioning environment
(i.e. one which allows participants to punish their group members) or a sanction-free
environment.
Each round of this experiment consists of multiple stages. In stage 1, participants
have an opportunity to choose to be in either a sanctioning or a sanction-free insti-
tution. In stage 2 participants participate in a linear public goods game. The round
ends here for participants who choose to be in the sanction-free institution. Partici-
pants who choose to be in the sanctioning institution continue to stage 3 where they
can allocate either positive or negative sanction points to other members. A positive
sanction (actually a reward) awards the recipient 1 token and costs the sender 1 token.
A negative sanction (a punishment) costs the recipient 3 tokens and costs the sender
1 token. Participants receive feedback on earnings at the end of the round. The exper-
iment consists of 30 rounds and participants are randomly re-matched at the end of
each round. This is a between subjects design and once participants choose to be in a
particular institution they do not learn the results of the other institutions.
In the first round of the game, a majority (63%) of the participants choose to
be in the sanction free institution (SFI) rather than the sanctioning institution (SI).
Participants who do choose to be in the SI, however, contribute on average 64% to
the public good which is significantly higher than the 37% contributed by those who
choose to join the SFI. Over time the SI becomes the predominant institution and
eventually close to 100% of participants chooses the SI. By round 10 of the 30 round
interactions, contributions in the SI increases to 90% and continues to go up from

Sustaining cooperation in laboratory public goods experiments
Fig. 2 Evolution of subjects’ choice of institutions and contributions over time from Gürerk et al. (2006).
Source: Gürerk et al. (2006)
there. In contrast, contributions in SFI decrease to zero. Averaged across all rounds,
contributions in the SI are 91%, significantly higher than the 14% in the SFI.
Moreover, not only do more and more participants migrate from the SFI to the SI
over time, migrating participants engage in high levels of cooperation and also very
quickly adopt the prevailing SI norm of punishing low contributors. Over time as the
amount of free-riding falls away to zero and the need for punishment diminishes,
the difference in payoffs between high contributors who do not punish free-riders
and high contributors who do punish becomes smaller, suggesting that “selection
pressures” against strong reciprocators become weaker over time.
Figure 2shows the evolution of subjects’ choice of institutions and contributions
over the 30 periods of interaction. The average contributions in both institutions over
time are measured as the percentage of the endowment contributed to the public
good.20
20Rockenbach and Milinski (2006) extend this line of investigation by analyzing the interaction between
punishment and reputation formation. In their main experimental treatment (called PUN&IR for punish-
ment and indirect reciprocity), participants are put into groups of eight and interact for 20 rounds. In each
round there are four stages. In the first stage participants decide whether to allow for costly punishments
or not following the public goods game. In the second stage, they play the public goods game. In the third
stage, those who choose to allow for sanctions get to mete out and/or receive punishment points. Finally,
in the last stage, each participant plays a gift exchange game along the lines of Berg et al. (1995), both as
a sender and a receiver. But prior to sending money the sender learns about the receiver’s “reputation” by
receiving information about the receiver’s contributions in the public goods game in the previous stage and
about the amounts sent by the receiver in prior plays of the gift exchange game. There is a control treatment
(called PUN) where participants get to choose whether to allow punishments or not following the public
goods game but there is no gift exchange game after that. Their main conclusion is that both contributions
and efficiency are higher in the treatment that allows for both punishment and indirect reciprocity. Effi-
ciency in this treatment averages more than 90% of the social optimum and is close to 100% during the

A. Chaudhuri
Gächter et al. (2008) examine, using a between subjects design, whether the du-
ration of interaction affects the efficacy of punishments by looking at two different
punishment treatments, one which lasts ten rounds (treatment P10) while the other
lasts fifty rounds (treatment P50). There are also two control treatments without any
punishment opportunities, one lasting 10 rounds (treatment N10) and the other lasting
50 rounds (treatment N50). Here each punishment point costs the punisher one token
but reduces the punished participant’s payoff by three tokens. What is particularly
striking in this study is that per period contributions in the P50 treatment are 25%
higher than those in the P10 treatment and 50% higher than those in the N50 treat-
ment. Average net earnings are significantly higher in the P50 treatment compared to
the N50 or P10 treatments. Finally, towards the later stages of the P50 treatment, co-
operation seems to become stabilized without much actual punishment being meted
out which results in punishment costs becoming negligible resulting in higher earn-
ings in this treatment.
Walker and Halloran (2004) look at the efficacy of rewards and/or sanctions in
a sequence of one-shot games. The main innovation of this study is that both the
reward and the sanction can either be certain or uncertain.Inthecertain treatments,
each dollar of punishment (reward) reduces (increases) the recipient’s payoff by two
dollars. In the uncertain treatments, with each dollar of punishment (reward), there
is a 50% chance that the punished subject’s payoff will be reduced (increased) by
4 dollars and a 50% chance that the payoff will remain unchanged. There are five
treatments including certain and uncertain punishment, certain and uncertain reward
and a control treatment with neither.
However neither rewards nor sanctions—whether certain or uncertain—have any
significant impact on contributions compared to the control treatment, suggesting that
repeated interactions and the consequent dynamics have an important role to play in
sustaining cooperation with punishments. A one-off threat of punishment may not be
as effective.
Sefton et al. (2007) explore the relative efficacy of punishments and rewards as
well using a repeated public goods game and a between subjects design. They look at
treatments that allow for only rewards or only punishments or both. Here each dollar
given up in punishment (or reward) reduces (increases) the recipient’s payoff by the
same amount.
While the introduction of an opportunity to punish or reward group members has
a salutary effect on cooperation in that contributions in those treatments are signifi-
cantly higher than those in the control treatment, the effect of sanctions is much less
pronounced compared to say the Fehr and Gächter (2000) results. In fact the treat-
ment that generates the highest contribution allows for both rewards and sanctions.
However, the relatively low cost-effectiveness of the punishment mechanism imple-
mented in this study—it costs $1 to punish another participant by $1—may have
something to do with its relative lack of success. As I argue in the next section, in
order to have an impact on contributions, the cost-effectiveness of punishments must
be sufficiently high.
last 10 rounds of interaction. In contrast when there is only punishment and no possibility to establish a
reputation, efficiency averages only about 75% and these differences are significant at conventional levels.

Sustaining cooperation in laboratory public goods experiments
3.1 On the cost-effectiveness of costly punishments
Nikiforakis and Normann (2008) suggest that the ability of costly punishments to
sustain high contributions to the public good depends crucially on the effectiveness
of that punishment, i.e., the factor by which each punishment point reduces the recip-
ient’s payoff.
Nikiforakis and Normann look at four different experimental treatments where
each punishment point meted out costs the punisher one experimental currency unit
(ECU). But in treatment 1, each point reduces the punishment recipient’s income by
one unit, in treatment 2 by two units, in treatment 3 by three units and finally, in treat-
ment 4 by four units. There is also a control treatment where no punishment mecha-
nism is available. Participants are placed into groups of 4 and play for 10 rounds in
a “partners” protocol. In order to prevent reputation formation, given fixed groups,
participant identification numbers are changed randomly from one round to next.
The authors find that there is a monotonic relation between the effectiveness of the
punishment and mean contributions; as the effectiveness goes up, so does the mean
contribution. Average contributions range from 9% in the control treatment to 33%
in treatment 1, 57% in treatment 2, 87% in treatment 3 and 90% in treatment 4. How-
ever the effectiveness of the punishment matters, in that it is only in the two “high”
punishment treatments, where each punishment point costs the recipient either three
or four currency units that contributions actually increase over time as in the original
Fehr and Gächter (2000) study. Contributions in the other less effective punishment
treatments show the familiar pattern of decay.
Furthermore, higher contributions sustained on the basis of punishments do not
necessarily translate into higher efficiency. Compared to the control treatment, it is
only in treatment 4 (where the punishment inflicts the maximum penalty on the re-
cipient) that average earnings are consistently higher. So it appears that the mere ex-
istence of punishment may not always be sufficient to enhance cooperation. In order
for punishment to truly make a difference, it must inflict a penalty that is substantially
higher than the cost of meting out that punishment.
Figure 3summarizes that the role of punishment effectiveness on efficiency. It
is only when the punishment has maximum effectiveness, depicted by the line with
stars labeled “4”, that it leads to significant gains in efficiency compared to the control
treatment.
Fig. 3 Cumulative relative
earnings across treatments in
Nikiforakis and Normann
(2008). Source: Nikiforakis and
Normann (2008)

A. Chaudhuri
Egas and Riedl (2008) also report that the only punishment treatment, which suc-
ceeds in sustaining cooperation over time is the low cost-high impact treatment where
each punishment point costs the punisher one token but reduces the recipient’s pay-
off by three tokens. There are two innovative features of this study. The first one is
the large sample size with 846 participants. The second is the non-standard pool of
participants with any Dutch-speaking person eligible to participate, making the par-
ticipants much more representative of the population as a whole. Participants were
recruited via newspaper advertisements and the experiments were conducted over the
Internet. Another interesting finding is that while contributions to the public good are
only weakly influenced by the age of the participant (given the non-student partic-
ipant pool in this study, the average age is 35 years with a range of 12–80 years),
older males are significantly more likely to punish, controlling for the punished par-
ticipant’s level of contribution.21
Carpenter (2007a) approaches the issue of punishment effectiveness by asking
what happens to monitoring of free riders and punishment as the group size becomes
bigger.22 On the one hand, as groups grow, it becomes harder for each individual to
monitor others and as a result free riding might become more prevalent as punishment
becomes less of a deterrent. On the other hand, in larger groups there are more people
monitoring each free-rider, so that free-riders might actually receive more punishment
in total compared to smaller groups.
Carpenter looks at both the effect of the group size,aswellasthemonitoring frac-
tion, which refers to the fraction of the group each agent can monitor. There are two
group sizes, 5 and 10, two values of the MPCR, 0.375 and 0.75 and four possible
types of monitoring: (1) no monitoring;(2)full monitoring;(3)half monitoring and
(4) single monitoring.Intheno monitoring treatment agents see what others con-
tributed but cannot punish them. In full monitoring, agents can punish any and all
other group members as long as they have the resources to do so. In half monitoring,
agents can punish that half of the group which is located closest to them on a circle
and finally in single monitoring they can punish only one other group member.
The parameters of the punishment technology are the same as that in Fehr and
Gächter (2000). For either value of the MPCR, contributions in the 5 person groups
with either full of half monitoring start at about 55% and remain stable around that
mark for the duration of the session. For the 10 person groups, with either full or
half monitoring, contributions start at the same level but then show a clearly increas-
ing profile reaching almost 100% contribution in the last three rounds of interaction.
21Nikiforakis (2010) reports that the efficacy of punishments also depends on the nature of the feedback
provided to participants. In all prior studies looking at punishments, participants get feedback about in-
dividual contributions at the end of each round. These studies report that the availability of punishment
opportunities leads to an increase in contributions. Nikiforakis shows that if participants are shown infor-
mation about individual earnings at the end of each round, then this leads to significantly lower contri-
butions and lower efficiency compared to a treatment where participants get information about individual
contributions.
22Isaac and Walker (1988b) and Isaac et al. (1994) examine patterns of contributions in linear public
goods games with groups of 5, 10, 40 and 100 players. Comparing groups of 5 with groups of 10, they
find that contributions are not significantly different with a MPCR of 0.75 while larger groups achieve
higher contributions with a lower MPCR of 0.375. When they look at groups of 40 and 100, they find that
contributions are actually higher compared to groups of 5 or 10 and this is independent of the MPCR.

Sustaining cooperation in laboratory public goods experiments
In both five and ten person groups, especially with MPCR =0.75, there is a clear
bifurcation with contributions showing an increasing profile with either full or half
monitoring on the one hand and a decaying pattern with single or no monitoring on
the other.
3.2 Do punishments obey the law of demand?
Anderson and Putterman (2006) explore the price responsiveness of the demand for
punishment by systematically varying the cost of each punishment point. In each
round, participants are divided into groups of three using a “perfect stranger” protocol
and take part in a public goods game for 5 rounds with the cost of a punishment point
changing from one round to the next. There are three treatments. In each treatment
there are five different costs for buying a dollar’s worth of punishment. (For instance,
in one treatment the cost of a dollar of punishment can take of one of five values—0,
30, 60, 90 or 120 cents.) In each treatment, a participant faced each one of the five
costs in random order over the five rounds.
What they find is that, across all treatments, the law of demand does hold for
punishments. Even after controlling for a number of variables including whether a
particular participant contributed more or less than the group average, the amount
of punishment to this participant decreases when the price of punishment goes up.
Assuming rational self-interest as the primary motivating factor, the act of punish-
ment seems to be non-rational, seeing as it yields no strategic benefits especially with
random re-matching of participants at the end of rounds. But there clearly seems to
be a fundamentally rational element of price-responsiveness built into the decision to
punish.
In Carpenter (2007b) participants play in groups of four using a “strangers” proto-
col for fifteen rounds. This study also has five possible per unit cost of punishment—
0.25, 0.5, 1, 2 and 4. So if the cost is 0.25 (4), for instance, that implies it costs 1/4of
a token (4 tokens) to buy one token of punishment. Participants face these costs either
in ascending or descending order and each value stays unchanged for three consecu-
tive periods. Participants are aware of this sequence. Another major design difference
is that each participant can punish only one other member of her current group.
While Anderson and Putterman (2006) suggest that the demand for punishments
may be elastic, Carpenter concludes that, after controlling for issues such as how
far below the group average a particular participant’s contribution is, the demand for
punishment is inelastic: a 10% increase in the price of punishment leads to about
8% reduction in the quantity demanded. Furthermore, the quantity demanded is not
responsive to changes in income, at least not significantly so.23
23I have chosen to discuss the Anderson and Putterman (2006)andCarpenter(2007b) papers in a different
section than Nikiforakis and Normann (2008)orEgasandRiedl(2008) for the following reason. In the
two latter papers the cost of buying each punishment point is always constant; what differs is the cost
inflicted on the recipient which can vary from a low to a high value. In the two papers in this section,
each punishment point meted out imposes the same cost on the recipient but the cost of buying each of
those punishment points varies. The net effect is of course the same which is to vary the cost-effectiveness
of punishments. But it is possible that participants may perceive these two mechanisms differently which
might lead to differences in behavior.

A. Chaudhuri
3.3 The possibility of “perverse” punishments
The above discussion suggests that costly monetary punishments of free-riders can
sustain high levels of contribution to the public good. Nikiforakis (2008), however,
sounds a note of caution about the salutary effects of punishments. He shows that
if one allows the possibility of counter-punishments by punished free-riders, coop-
erators are less willing to punish. He finds that punishments are often “perverse” or
“anti-social” in nature in the sense that those who free-ride often punish those who co-
operate and such punishments are driven partly by strategic considerations and partly
by a desire to avenge the punishments meted out by others. In what follows I will
stick to the terms “anti-social” punishment to indicate punishment of high contribu-
tors and “pro-social” punishment to indicate the more usual punishment of free-riders
by cooperators.
Nikiforakis looks at two different punishment mechanisms. The punishment treat-
ment is identical to that implemented in Fehr and Gächter (2000) discussed above
with the same punishment costs as shown in Table 1. The other treatment, which is
the primary focus of this study, adds a third counter-punishment stage following the
second punishment stage to each round. At the beginning of this third stage each par-
ticipant is informed about the number of punishment points assigned to him by his
group members and is given an opportunity to assign counter-punishment points to
those participants in turn. The punishment costs are the same as in stage 2 (shown in
Table 1). These costs work cumulatively. A participant’s final earnings in a round in
this third treatment is his earnings from the public goods game in stage 1 minus all
the income reductions caused by the punishment points assigned to this participant
by others and those assigned by the participant to his peers over the two stages of
punishment and counter-punishment.
A crucial feature of this study is that only those participants who are actually
punished in stage 2 are allowed to engage in counter-punishment and they can only
punish those who punished them in the first place; moreover, a participant must have
a positive payoff in order to carry out any counter-punishment and these must be
carried out in the stage immediately following the punishment.
This is a within subjects design with both fixed groups and random re-matching.
Participants interact for 20 rounds in two blocks which are counter-balanced. In one
of those blocks participants play the standard public goods game while in the other
block they have the opportunity to either engage in punishment or both punishment
and counter-punishment.
Nikiforakis finds that individuals in the punishment treatment are significantly
more likely to contribute to the public account compared to those in the counter-
punishment treatment, where contributions show the familiar pattern of decay. The
counter-punishment treatment also leads to lower average earnings compared to both
the punishment treatment and the control treatment.
In looking at why the counter-punishment treatment does worse than the punish-
ment treatment, the author finds that participants in this treatment engage in substan-
tial amounts of “anti-social” punishment which can be attributed to one of two factors
or a combination of those. One is the anticipation by some free-riders of the forthcom-
ing punishment from cooperators and their willingness to retaliate those sanctions.

Sustaining cooperation in laboratory public goods experiments
Here participants use counter-punishments strategically to signal that future sanc-
tions will not be tolerated and this is especially true in fixed groups, which affords
scope for such signaling. The second factor is the desire to avenge sanctions meted
out to them in previous periods. In fact, participants in the counter-punishment treat-
ment are 15% less likely to punish free-riding compared to the punishment treatment
mostly because cooperators anticipate that this might, in turn, lead to “anti-social”
punishment and wish to avoid the same.
However, Cinyabuguma et al. (2005) suggest that the specific mechanism for im-
plementing counter-punishments might make a big difference to their eventual im-
pact. In the Cinyabuguma et al. study, participants do not learn the identity of those
who punished them. This makes it impossible to engage in targeted revenge. Instead
each participant is told the pattern of punishing high, average and low contributors
in the group in the first stage and then that participant can decide who to counter-
punish, i.e. whether to engage in counter-punishment of “pro-social” or “anti-social”
punishers.
Contrary to Nikiforakis, who found that contributions and earnings in the
treatment with counter-punishment were lower than the punishment treatment,
Cinyabuguma et al. find that the availability of counter-punishment does not reduce
either contributions or earnings to levels lower than those achieved in the punish-
ment only treatment; at least not significantly so. It is conceivable that this is due
mostly to the design differences between the two studies. The opportunity to engage
in counter-punishment then seems to have very different impact depending on its
exact implementation.
Ertan et al. (2009) allow their participants to vote on who should be punished—
those who contribute less than, equal to or greater than the group average contribu-
tion. The authors find that there are no groups where a majority voted to allow pun-
ishment of participants who contribute higher than the group average contribution.
This ruled out the possibility of “anti-social” punishments. In initial votes there is a
tendency to vote for no punishment. However over time, there is a clear evolution to-
wards allowing punishment of low contributors while still prohibiting the punishment
of high contributors.
Hermann et al. (2008) provide even more compelling evidence of such “anti-
social” punishments in an ambitious cross-cultural experiment that compares the
behavior of under-graduate students across 16 different locations using a between
subjects protocol.24 The punishment parameters are the same as in Fehr and Gächter
(2000). The incidence of “anti-social” punishments is the lowest among the partici-
pants from western industrialized nations where the bulk of the previous experimental
data comes from. This, in turn, suggests that the evidence in favor of the cooperation
enhancing role of punishments that comes from prior experiments run in western
societies may actually be over-estimating the efficacy of such punishments and in
24These locations are: Athens, Bonn, Boston, Chengdu, Copenhagen, Dnipropetrovs’k, Istanbul, Mel-
bourne, Minsk, Muscat, Nottingham, Riyadh, Samara, Seoul, St. Gallen and Zurich. Henrich et al. (2010)
point out that most people are not WEIRD meaning Western, Educated, Industrialized, Rich and Demo-
cratic. Yet the participants in the vast majority of studies including those cited in this article come from
that somewhat unusual pool. Hence, the extent to which these results are applicable to those who are not
WEIRD is open to debate.

A. Chaudhuri
Fig. 4 Mean punishment expenditures across different locations in Hermann et al. (2008). Source: Her-
mann et al. (2008)
other societies the presence of “anti-social” punishments may actually have a large
detrimental role.
Different participant pools reacted very differently to the punishment received. In
only 11 out 16 societies, those punished in one round for contributing less than the
group average increased their contribution in the next round and the extent of the
mean estimated increase per punishment point received varies considerably. Thus,
punishment did not have an equally strong disciplinary effect on free riders in all
participant pools in increasing their cooperation and in some societies punishments
did not increase cooperation at all. Figure 4summarizes the nature of pro-social and
anti-social punishments across these diverse societies.
The authors suggest that “anti-social” punishments are more prevalent in societies
which are characterized by (1) a lack of strong social norms of civic cooperation as
expressed in people’s attitudes towards tax evasion, abuse of the welfare system or
dodging fare on public transport and (2) weak law enforcement.25,26
25The rule of law indicator is based on a number of variables that measure “the extent to which agents have
confidence in and abide by the rules of society, and in particular the quality of contract enforcement, the
police, and the courts, as well as the likelihood of crime and violence”. (Hermann et al. 2008, pp. 1366.)
26Gächter and Hermann (2010) demonstrate the existence of anti-social punishments using participants
from both urban and rural areas of Russia as well as participants who are “young” (less than 30 years of
age) and “mature” (older than 30). The urban participants come from Kursk and Zheleznogorsk, located
in the so-called “Central Black Earth Zone” about 400 miles south of Moscow. The rural participants are
located in the rural areas around Kursk and in Ust-Kinel. The urban young participants are mostly univer-
sity students. Participants play two one-shot public goods games, one with punishments and one without,
in groups of three. Each punishment point meted out costs the punisher one token in earnings while it

Sustaining cooperation in laboratory public goods experiments
3.4 Counter-punishments: revenge or cooperative norm enforcement?
The above suggests that counter-punishments can either be “pro-social” or “anti-
social”. “Pro-social” counter-punishments can take two forms. They could be aimed
at (1) those who engage in “anti-social” punishments in the first stage and punish high
contributors or (2) those who fail to punish free-riders. As such “pro-social” counter-
punishments should enhance cooperation by deterring free-riding or “anti-social” first
stage punishments and can play a powerful cooperative norm enforcement role. “Anti-
social” counter-punishments, on the other hand, are more like acts of revenge. They
are usually carried out by free riders and aimed at cooperators.
An innovative study by Denant-Boemont et al. (2007) addresses these two aspects
of counter-punishment; its role (1) in exacting revenge on those who have punished
one before or (2) in enforcing cooperative norms via punishment of free riders as well
as of those who engage in “anti-social” punishments.
This study uses a between subjects design with groups of four participants and
a “partners” protocol. The principal innovation here is implemented at the third,
counter-punishment stage. In the full information (FI) treatment, players are not only
informed about the contributions of others but also of how each individual sanctioned
each other individual. In the revenge only (RO) treatment, each individual is informed
of the source and quantity of the sanctions directed towards him in a treatment that is
analogous to Nikiforakis (2008) and allows the authors to focus on the “anti-social”
aspect of counter-punishments. In the no revenge (NR) treatment, no individual is
informed about who sanctioned him personally and by how much. Rather, all indi-
viduals are informed of the source and the quantity of the sanctions directed toward
each player other than one’s own self. This treatment is similar to that implemented
by Cinyabuguma et al. (2005) and allows for isolating the “norm enforcement” effect
of counter-punishments. There is also a six stage full information treatment but with
five stages of unrestricted sanctioning following the initial contribution decision in
the first stage.
The availability of sanction enforcing “pro-social” punishments coupled with a
restriction on revenge seeking is welfare improving in that earnings in the no revenge
treatment, at 81% of the social optimum, are significantly higher than either the full
information treatment (53% of the optimum) or the revenge only treatment (36% of
the optimum). In the full information treatment, which allows for both “anti-social”
punishments as well as “norm enforcing” punishments, increments in contributions
costs the receiver three tokens in earnings. Gächter and Hermann find that there is significant punish-
ment of cooperators across all subject pools, with people punishing both those who contributed the same
amount as the punisher as well as those who contributed significantly more than the punisher. Rural resi-
dents contribute more than urban ones and mature participants contribute more than young ones. Gächter
and Hermann (2009) compare anti-social punishments among university students located in Belgorod and
Yekaterinburg (both in Russia) and those located in Zurich and St. Gallen (in Switzerland). Like Hermann
et al. (2008), Gächter and Hermann (2009) show that there may be a cultural component to the desire
to impose anti-social punishments in that such punishments are far more pronounced among the Russian
participants than among the Swiss participants. Gächter and Hermann (2009) also provide a brief overview
of various mechanisms that enhance cooperation.

A. Chaudhuri
caused by the norm enforcing sanctions are not enough to fully offset the negative
effect exerted on contributions by anti-social counter-punishments.27
3.5 Concluding remarks
The evidence on costly punishments suggests that providing participants with the
opportunity to engage in such punishment of group members can usually help sustain
high levels of contributions. But this finding is subject to at least three caveats.
First, there is the problem that the punishment itself creates a second-level public
good; those who are willing to mete out costly punishment must not only punish free-
riders but also those non-punishers who might contribute but do not punish free riders
and hence free ride on others’ punishment and so on. This requires the creation of
“meta-norms” of punishment in the words of Axelrod (1986). However, the existence
of punishment opportunities and of conditional cooperators who are willing to use
such punishment does seem to reduce free riding. And at times, given a long enough
time horizon, as in Gächter et al. (2008), the threat of punishment might be enough
to sustain cooperation without the punishment actually having to be carried out.28
The second problem has to do with the issue of “anti-social” punishments as dis-
cussed in Sect. 3.3. The presence of such “anti-social” punishments may result not
only in no increase in cooperation but also seriously reduce efficiency compared to
control treatments with no punishment. The cooperation enhancing effect of punish-
ments seems more prominent in particular participant pools, especially in western
industrialized societies. Also, the exact nature of counter-punishment matters. If par-
ticipants are only allowed to engage in “pro-social” punishments but not in targeted
revenge, then this could be welfare improving. However, if we allow for both “anti-
social” and “norm enforcing” punishments then the net effect is detrimental to coop-
eration because the increase in cooperation caused by norm enforcing sanctions does
not fully offset the contribution reducing effect of anti-social punishments.
The third problem is that the efficiency implications of costly punishments are
not clear cut. By and large, across the majority of studies cited above, efficiency is
actually lower in treatments with punishment compared to control treatments with-
out punishment. The ability of punishments to enhance efficiency seems to depend
27The authors also find that adding more stages of unrestricted punishment as in the six stage full infor-
mation treatment is unambiguously welfare reducing. This reduction arises from two sources: a reduction
in contributions and much higher levels of sanctioning. In fact the average punishment points awarded in
this treatment is 2.7 times those in the next highest treatment.
28Henrich and Boyd (2001) develop a theoretical model to demonstrate that “conformist transmission”
(where agents use the popularity of a choice as an indirect measure of its worth) can address this problem
of higher order punishments and stabilize cooperation. Suppose being punished is sufficiently costly so
that co-operators have higher payoffs than defectors. Then a second-order free rider who cooperates but
does not punish free-riders will achieve a higher payoff because they avoid the punishment cost. But if
defections do not pay, then such defections will occur rarely and by mistake; so over time such defections
become less frequent and as we ascend higher orders of punishing the difference in payoffs between
punishers and second-order free-riders will start to approach zero. Thus conformist transmission, no matter
how weak, will at some stage (t) be able to overcome payoff biased imitation in the form of free-riding and
stabilize punishment. Once this happens, by backward induction, payoffs will favor strategies that punish
at the (t−1)th stage, which, in turn, favors punishment in the (t−2)th stage and so on till cooperation is
stabilized in the very first stage.

Sustaining cooperation in laboratory public goods experiments
crucially on two things: (1) the cost-effectiveness of the punishment; as shown by
Nikiforakis and Normann (2008) and Egas and Riedl (2008) it is only when the pun-
ishment is low cost and high impact that it also leads to an increase in efficiency over
and above any increase in contributions. (2) The other factor that makes a difference
is the time horizon. As Gächter et al. (2008) show, if the time horizon is sufficiently
long, then even low impact punishments can generate efficiency gains that are not
present over a shorter time frame.
These caveats highlight the practical difficulties of implementing such costly de-
centralized peer punishment. Furthermore, Guala (2010) points out that ethnographic
evidence from tribal societies or the historical evidence on common pool resource
usage does not provide a lot of support for either the use or the efficacy of such costly
monetary punishments. Therefore, mechanisms that rely less on monetary punish-
ments and more on other factors might be easier to adopt. This is what I explore
next.
4 Sustaining cooperation without monetary punishments
Axelrod (1986) suggests that a social norm is essentially an implicit rule that mem-
bers of society feel compelled to adhere to. One way of creating and sustaining such a
norm is via internalization, where a norm becomes so entrenched in a society that vi-
olating it causes psychological discomfort. Below, I explore different ways in which
such internalization may be achieved. I begin by exploring punitive measures that
are non-monetary in nature such as social ostracism or exclusion and then go on to
discuss other non-punitive mechanisms that lead to successful cooperation.
4.1 Even non-monetary punishments can sustain cooperation
Mascletetal.(2003) is a pioneering study demonstrating that non-monetary punish-
ments such as expressions of disapproval can enhance cooperation. This study will
be familiar to most readers and hence I will avoid a lengthy discussion. The authors
compare the efficacy of non-monetary punishments with that of monetary punish-
ments. The latter is similar to those in Fehr and Gächter (2000). In the non-monetary
punishment treatment, participants are given the opportunity of expressing approval
or disapproval of the actions of other group members but these do not affect monetary
payoffs. As with the punishments each participant can assign between zero and ten
points to another participant where zero indicates no disapproval and ten indicates
maximum disapproval.
The authors find that both monetary and non-monetary sanctions initially increase
contributions by a similar amount but that over time, monetary sanctions are more
effective and lead to higher contributions than non-monetary sanctions. Furthermore,
and not surprisingly, they find that non-monetary sanctions are more effective in the
“partners” treatment as opposed to the “strangers” treatment. The authors also find
that the average earnings of participants are higher with either monetary or non-

A. Chaudhuri
monetary punishments compared to the baseline situation where no sanctions are
available.29
Cinyabuguma et al. (2006) look at punishment from a different perspective by
allowing group members the opportunity to expel free-riders by majority vote. This is
a between subjects study with each session consisting of 16 participants all belonging
to the same group, in a “partners” protocol. Participants play for 15 rounds with
two stages in each round. In each round after learning individual contributions, each
participant is given an opportunity to vote to remove an individual from the group.
A group member will be removed from the group if half or more voted to expel that
person and such exclusion is irreversible. Group members who are expelled play a
similar public goods game except with half the endowment in each round compared
to the original group and therefore will potentially earn less.
The authors find that while there are few actual expulsions, a majority of partic-
ipants voted to expel another participant at least once and typically the ones getting
expulsion votes or actually being expelled are either the lowest contributor or the
second lowest contributor. When a participant receives an expulsion vote in a given
round, even if he is not actually expelled, he responds by increasing his contribution
in the next period, as in Masclet et al. (2003). Both the average contribution to the
public good and the overall earnings are higher in the expulsion treatment than in the
control treatment.
4.2 Sustaining cooperation via non-punitive mechanisms
The studies that look at non-punitive measures can be broadly classified into (1) those
that attempt to foster cooperation among participants without making any attempt to
sort them and (2) those that try to form sorted groups on the basis of similarity of
behavior or preferences.
4.2.1 Sustaining cooperation in non-sorted groups
One obvious mechanism to promote cooperation is to allow for communication
among participants. Ever since Dawes et al. (1977) and Isaac and Walker (1988a)
we have known that communication can improve cooperation. Bochet et al. (2006)
extend this literature by directly comparing the relative efficacy of communication
vis-à-vis punishment in sustaining cooperation using a between subjects design.
The authors look at three types of communication; (1) face-to-face, (2) chat-room
and (3) numerical cheap talk. In the treatments with face-to-face communication,
each participant has the opportunity to talk to the other three group members for five
minutes before the game starts. In the treatments with chat room communication,
each participant can exchange verbal messages with group members via a computer
chat room. In the numerical cheap talk treatment, each participant has the option of
29Noussair and Tucker (2005) extend the work done by Masclet et al. (2003) by looking at whether the
availability of both monetary and non-monetary sanctions can generate higher welfare than either type on
its own. They find that over time contributions in the non-monetary punishment treatment falls consistently
as the non-monetary sanctions appear to lose effectiveness when not backed up by a monetary sanction.

Sustaining cooperation in laboratory public goods experiments
typing in a number to indicate his/her potential contribution to the public account. No
other form of communication is allowed.
The authors also look at three more treatments where each particular communi-
cation strategy is combined with the opportunity to punish one’s group members. It
costs 0.25 token to punish another participant by 1 token. There is also a punishment-
only treatment which only allows for punishment of group members but no opportu-
nity to communicate.
The main insight of this study is that once face-to-face communication is allowed
contributions jump up to 96% of the maximum which is significantly higher than
those in the control and the punishment-only treatments. Given the already high con-
tributions in the face-to-face condition, allowing punishments on top of that leads to
only a small increase in contributions to 97%. The chat room communication with
punishment does almost as well as the face-to-face treatment and gets average contri-
butions of around 96% while the chat room communication without punishment does
not do as well with contributions averaging 81%. However, both versions of the chat
room treatment do better than either the control or the punishment-only treatment.
Unlike the other communication treatments, the numerical cheap talk treatment with
or without punishments does not exhibit either higher contributions or higher earn-
ings as compared to either the control or the punishment-only treatments.30
Chaudhuri et al. (2006) investigate a different type of communication scheme by
allowing participants to pass advice. The focus here is on the evolution of cooperative
norms using an inter-generational approach. Participants in one generation leave ad-
vice for the succeeding generation via free form messages. Such advice can be private
knowledge (advice left by one player in generation tis given only to her immediate
successor in generation t+1), public knowledge (advice left by players of generation
tis made available to all members of generation t+1) or common knowledge (where
the advice is not only public but also read aloud by the experimenter). Contributions
in these advice treatments are compared to those in a control treatment with no op-
portunity to leave advice. Participants play in groups of 5 for 10 rounds. However
each participant in generation tis connected to another participant in the immedi-
ately succeeding generation t+1 and each participant in generation tearns a second
payment which is equal to 50% of the earnings of her generation t+1 successor.
The authors find that average contributions in the common knowledge of advice
treatment are significantly higher than the other treatments including the control treat-
ment. Common knowledge of advice generates a process of social learning that leads
to high contributions and less free riding. In later generations of the common knowl-
edge treatment contributions and efficiency are greater than 90% of the social op-
timum and the modal contribution to the public account is the entire token endow-
ment. This behavior is sustained by advice that is generally exhortative, suggesting
high contributions, which in turn creates optimistic beliefs among participants about
others’ contributions.
The authors also collect data on beliefs using an incentive-compatible mechanism.
Average post advice beliefs are the highest in the common knowledge treatment.
30For two other studies that also explore the role of communication in fostering cooperation in public
goods game see Brosig et al. (2003) and Bochet and Putterman (2009).

A. Chaudhuri
Given the strong positive correlation between beliefs and contributions, it is not sur-
prising that this treatment generated high contributions.
Rege and Telle (2004) examine the impact of social norms via indirect social ap-
proval and framing on cooperation. The authors look at the effect of two factors:
(1) social approval and (2) associative framing.
The first no-approval treatment is run using a double-blind protocol, thereby mak-
ing social approval or disapproval impossible. In the social approval treatment this
anonymity is removed where each participant’s identity and contribution to the public
good are revealed to the group members. In an associative treatment the participants
are referred to as a “community” with the intention of creating social and internal-
ized norms for cooperation, while in a non-associative treatment the instructions are
written in more abstract language. This generates four different conditions—(1) no
approval/non-associative; (2) no approval/ associative; (3) approval/non-associative
and finally (4) approval/associative.
The authors find that the treatments have a significant impact on contributions.
Average contributions increase from 34% in the no-approval/non-associative treat-
ment to 55% in the no-approval/associative treatment, to 68% in the approval/non-
associative treatment to 77% in the approval/associative treatment.31
4.2.2 Cooperation in sorted groups
Chaudhuri (2009) points out, in many of the things we do in life we actually choose
the people we wish to interact with as when we join religious or social groups, osten-
sibly because these people have preferences similar to ours. Sustaining cooperation
in such sorted groups might prove to be less of a challenge. Below I discuss papers
where such sorting is either (1) exogenous (undertaken by the experimenter on the
basis of a pre-determined rule which may or may not be known to the participants)
or (2) endogenous (allowing participants to form groups or leave groups on their own
accord).
4.2.2.1 Exogenously sorted groups Gunnthorsdottir et al. (2007) investigate this is-
sue by sorting cooperative contributors in the same group. Each session includes 12
31Seely et al. (2005), look at the role of a “credible assignment”—essentially a non-binding public
announcement—in a linear public goods game. Except they look at a game which will be terminated
after a certain number of periods with a given probability. In essence Seely et al.’s endeavor amounts to
a test of the folk theorem in an infinitely repeated prisoner’s dilemma. Participants are assigned to differ-
ent treatments where they are asked to adopt various strategies vis-à-vis their contributions to the public
good. The strategy that succeeds in fostering the most cooperation involves using a “grim trigger”. Here
participants are asked to make the socially optimal contribution to the public account as long as the other
group members do so but contribute nothing in all subsequent periods following a deviation from this
norm. But given that this is not a finitely repeated game and hence the strategic considerations are quite
different, I will avoid a more elaborate discussion of this study. Chaudhuri and Paichayontvijit (2010a)
compare the efficacy of such public announcements with that of costly punishments in a finitely repeated
linear public goods game. They find that contributions in the initial rounds are higher in the treatment
with an announcement compared to the control and punishment treatments. However contributions decay
much faster in the treatment with an announcement whereas contributions increase in the treatment with
punishments over time. Payoffs are higher in the announcement treatment in the initial rounds but decrease
over time whereas payoffs increase in the punishment treatment over time.

Sustaining cooperation in laboratory public goods experiments
participants. Participants are grouped into four where they play 10 rounds of a public
goods experiment with three possible values of the MPCR: 0.3, 0.5 and 0.75. The au-
thors look at two different grouping rules: (1) participants are randomly re-matched
into different groups at the end of each round (Random treatment) and (2) partici-
pants are sorted into groups depending on their contribution at the end of each round
(Sorted treatment). The four participants who contribute the most to the public ac-
count are placed into one group; the fifth to eighth highest contributors are placed
into another group; and the four lowest contributors are placed in the third group.
Hence the grouping is dependent on the contribution in the current round. To avoid
strategic behavior participants are not informed about how the groups are formed.
For a given value of the MPCR, contributions among the sorted groups are always
greater than among randomly formed groups. Also the decay in contributions is much
slower among sorted groups compared to the randomly formed groups with little or
no decay in the two sorted treatments with MPCR =0.5 and MPCR =0.75.
The authors define “free-riders” as those who contribute 30% or less of their en-
dowment to the public account. The rest are defined as “cooperators”. Within each
MPCR, by round 4 at the latest, contributions by cooperators in the sorted treatment
exceed contributions by cooperators in the random treatment. Since the sorted treat-
ment reduces the number of interactions between cooperators and free-riders, the au-
thors conclude that higher contributions by the cooperators in this treatment are due
primarily to the more efficacious nature of their prior interactions and the exposure
to a history of cooperative interactions. On the other hand, the decay in contribu-
tion in the random re-matching treatment is due almost entirely to the reduction in
contribution by the cooperators who experience much greater interaction with free
riders.
In Gächter and Thöni (2005) participants first take part in a “ranking experiment”
which consists of playing a one-shot linear public goods game with an MPCR of
0.6 in randomly formed groups of three. Participants did not receive any information
about the contribution of other group members or their earnings at this point. Fol-
lowing this, participants take part in the main experiment which consists of playing a
ten-period repeated linear public goods game.
For the main experiment, the three highest contributors in the ranking experiment
are put together in one group, the next three in the second group and so on till the
three lowest contributors who form the last group. Participants get to know how these
groups are formed and are also informed how much their new group members con-
tributed in the ranking experiment.
There is also a control treatment, where the groups are formed randomly and has
nothing to do with what the participants contributed in the ranking experiment. The
authors also combine the two grouping protocols with the opportunity to punish group
members. This then gives rise to four separate conditions: (1) Sorted no punishment;
(2) Random no punishment;(3)Sorted punishment and (4) Random punishment.
Sorting people into groups based on their performance in the ranking exercise led
to a substantial increase in contributions. Even without any punishment opportuni-
ties, the top third of contributors in the sorted groups contribute significantly more
than the most cooperative third in the randomly formed groups with average contribu-
tions of 70% of the social optimum among sorted groups and only 48% of the social

A. Chaudhuri
optimum among random groups. Not only that, the three highest contributors in the
sorted groups achieved the same level of contribution as the most cooperative third
of randomly formed groups even when the latter had a punishment option at their
disposal. The availability of a punishment opportunity does not make a difference in
sorted groups since the three highest contributors in the treatment with sorting but no
punishment manage to sustain the same level of cooperation as those in the treatment
with sorting and punishment.
de Oleveira et al. (2009) also engage in exogenous sorting but in their study some
participants are explicitly informed about the type of the other group members while
others are not. So the focus is on the role of information regarding the type of group
members. Participants first play a one-shot public goods game where they are cate-
gorized either as “Conditional Cooperators” or as “Selfish” using the same approach
as in FGF. Then, on a different day they take part in a linear public goods game re-
peated for 15 rounds. Participants are placed into groups of three, where the groups
can be (1) homogeneous, consisting of either all conditional cooperators or all self-
ish players or (2) heterogeneous, consisting of two players of one type and one of
the other. In the Known Distribution treatment, participants are explicitly told about
the composition of the group prior to starting the experiment, while in the Unknown
Distribution treatment they are not given this information. In both treatments, the
participant knows his own type. The composition of the groups remains unchanged
for the duration of the session.
There are two important insights coming out of this study. First, not surprisingly
contributions in groups with three conditional cooperators are significantly higher
than in those with two conditional cooperators or one. But more importantly, contri-
butions in groups with three conditional cooperators are higher when the distribution
is known as opposed to when it is unknown. This suggests that the mere presence
of conditional cooperators (which can conceivably be inferred from the contribution
patterns) is not enough, conditional cooperators need to know that there are no self-
ish types in their group for them to sustain cooperation. This latter finding echoes the
results reported by Chaudhuri and Paichayontvijit (2006) that conditional coopera-
tors cooperate more when they are made aware of the presence of other conditional
cooperators.32
4.2.2.2 Endogenous sorting of participants Page et al.’s (2005) approach is similar
to Gächter and Thöni (2005) except that in Page et al. participants can choose who
they want in a group with them. There are 16 participants in each one of four sessions.
At periodic intervals during a session, each participant is shown a list, without other
identifying information and in a random order, of each of the other 15 participants’
average contribution to the public account till that point. Participants are then given
the opportunity to express a preference among possible future partners by ranking
them. The four individuals with the lowest rank are then put together in the same
32Burlando and Guala (2005) also form sorted groups of four based on their classification of participants
as described in Sect. 2above. These sorted groups then play a linear public goods game for 10 rounds a
week later. The striking finding of this study is that in the second session with sorted groups—the groups
consisting of all unconditional cooperators or all conditional cooperators achieved almost full cooperation
for the entire duration of the session.

Sustaining cooperation in laboratory public goods experiments
group. Here the group size is always equal to four, except participants get to choose
which group they wish to belong to.
After new groups are formed, participants resume play without information about
whom they have been grouped with, a matter on which only indirect inference can
be made by observing one’s three partners’ contributions. The authors also look sep-
arately at a punishment treatment and a combined treatment with regrouping and
punishment.
Regrouping leads to significant increases in contribution to the public good com-
pared to the control treatment. Moreover, average contributions in the regrouping
treatment are about the same as in the punishment treatment (about 70% of the social
optimum on average). Thus the participants’ abilities to influence with whom they
are grouped has a demonstrable positive effect on cooperation and efficiency in this
study.33
The last two studies discussed in this section adopt a more complex mechanism
where both the size and the composition of the group are determined endogenously. In
Ehrhart and Keser (1999) there are five sessions with 18 participants in each playing
for 30 rounds. Participants in a session are divided into two populations of 9 each.
No participant knows who the other 8 members of the population are. In the first
round of a session the 9 members of a population are randomly assigned to a three
3-person groups. From the second round onwards, each round consists of two stages.
In the first stage, participants make contribution decisions as usual. Participants here
get to see the contributions of all members of the population. In the second stage,
each participant gets to decide whether to continue with the same group or whether
to leave the group at a cost. It is possible for a participant to leave and form a one
member group during the particular round.
The returns to the public good are set in a way that the socially efficient outcome
would be to form a “grand coalition” with all 9 members of the population belonging
to the same group. However, Ehrhart and Keser find that this grand coalition is seldom
achieved. Because the returns are higher with increasing group size, groups with
higher contributions tend to grow in size. But as they do, the extent of free riding
within the group increases as well preventing the groups from reaching the optimal
size. Participants who make high contributions are more likely to leave a larger group
and form a smaller one (or even a “single”) while free-riders are more likely to join
larger groups in order to reap the economies of scale.
Charness and Yang (2007) undertake a more elaborate investigation where partic-
ipants are not only free to leave their current groups as in Ehrhart and Keser (1999)
but they can also vote to expel group members as in Cinyabuguma et al. (2006).
33Ones and Putterman (2007) also study the impact of costly punishments in a situation where groups are
sorted according to their levels of cooperation and find that groups consisting of participants with similar
cooperative tendencies outperform randomly composed groups. They extend the findings of Page et al.
(2005) and Gächter and Thöni (2005) by testing whether cooperative preferences are stable over time and
whether differences in group outcomes can be predicted by knowing the types of participants who compose
those groups. Like Gunnthorsdottir et al. (2007), Ones and Putterman (2007) find that early contributions
can serve as a significant predictor of contributions in later periods. Moreover the combination of own
type measures coupled with measures of experiences of interacting with other groups members can explain
substantial parts of the variation in later contributions by participants.

A. Chaudhuri
However, here the expulsion vote is less punitive because expelled members are free
to join other groups or remain as “singles” with the same endowment. Thus expul-
sion need not imply a reduction in payoff. Beyond this, there are also opportunities
for mergers among groups as a whole. There are two experimental treatments and a
control treatment, each consisting of two blocks of 15 rounds each.
The main focus of interest in this study is their treatment 2, where 9 people in a
“society” are placed into three groups of 3 participants each and play a public goods
game for 3 rounds. The social value of an allocation to the public account depends
on the group size and as in Ehrhart and Keser’s study, the greatest group returns
are achieved by forming a “grand coalition” where all 9 members of society belong
to the same group. After the first three periods, participants learn about the average
contribution of each other individual in their society (by identification number only)
for those three periods. At that point participants can choose to either exit the group
or vote to expel other group members. Groups are allowed to merge as well. After the
end of the first segment of 15 rounds, groups are re-formed and play a second set of
15 rounds which proceed along similar lines.
Clearly endogenous group formation enhances contributions to the public good in
comparison to exogenously formed groups as in the control treatment. While the con-
tribution rate in exogenously formed groups in a control treatment steadily decline to
around 25% of the social optimum, in endogenously formed groups the rate increases
to above 95% in the later periods.
Contrary to Ehrhart and Keser (1999), the most commonly occurring group com-
position in this treatment is in fact the grand coalition of with all 9 members of so-
ciety belonging to the same group followed by 8-1 and 7-2 splits respectively and
these larger groups tend to be quite stable over time. The authors also find that par-
ticipants are less likely to exclude another group member the higher that member’s
contribution vis-à-vis the group average and individuals/groups are more likely to
merge with another group, when that latter group is larger and achieves higher aver-
age contributions vis-à-vis the contributions in the former group. Given the ability to
sort cooperators in this treatment, profit-maximizing participants find that it pays to
cooperate given that they manage to belong to groups where others also contribute.34
4.3 Concluding remarks
The evidence presented in this section suggests that in the presence of conditional
cooperators, contributions to the public good can be sustained by means other than
34See Ahn et al. (2009) for a study looking at endogenous group formation in a congestible, rather than
pure, public goods game. In this game, the payoff function is such that the contributions become increas-
ingly more expensive the higher the contribution level. Payoff is also decreasing in increasing group size.
Kosfeld et al. (2009) also examine the issue of endogenous formation of institutions using theory and ex-
periments. Prior to taking part in a public goods game, participants get to decide whether to join a group
that will allow for punishment of free-riders or a group where no such punishment is possible. They find
that subjects frequently implement an organization with punishment and like Charness and Yang (2007)
the majority of these consist of the “grand coalition” of all four group members. However, the experimen-
tal implementation of their theoretical model raises questions because in the actual experiment those who
join the groups with sanction are constrained to contribute their endowment to the public account while the
possible punishment of non-contributors is a major factor behind the formation of the sanctioning groups
in the first place. This renders their conclusions somewhat difficult to interpret.

Sustaining cooperation in laboratory public goods experiments
monetary punishments. These may include non-monetary punitive measures such as
expressions of disapproval or social exclusion. They could also include other inter-
ventions such as different types of communication including advice giving from one
generation to next and assortative matching of like-minded participants. While in
some cases such assortative matching is achieved exogenously with the experimenter
sorting participants into groups based on similarities in their behavior or preferences,
in some cases, participants left to themselves can form cooperative groups endoge-
nously and can sustain cooperation via either expulsion of free-riders or via leaving
less cooperative groups for more cooperative ones.
5 Conclusion: where to from here?
We now have a fairly clear picture about the preference heterogeneity among partic-
ipants and the preponderance of conditional cooperators. We also have a good idea
of how we can go about creating institutions—particularly those relying on costly
punishments—that exploit such conditional aspects of behavior to sustain coopera-
tion. I think that the value added by yet another paper exploring these issues is going
to be limited.
What then would be fruitful avenues of exploration? One obvious way forward is
to apply the lessons learned to “field” settings in designing institutions that deal with
social dilemmas. There is already a fairly robust literature in the area. Nevertheless,
the evidence cited by Frey and Meier (2004), Gächter (2007), Ostrom (1990) and
Ostrom et al. (1994) and Ostrom’s being awarded the 2009 Nobel memorial prize
in economics suggests a renewed and wider interest in policy implications of these
findings.35
I think that the phenomenon of contributions decay might lead to further work in
the area in terms of both experiments and theory. Existing theoretical models in this
area all assume complete information regarding player types. It is possible that one
potential area of advance would involve assuming asymmetric information regarding
types. There are two ways to think about such incomplete information; one is to
assume heterogeneity in types (such as conditional cooperators and free riders) and
the other is to assume heterogeneity in prior beliefs among conditional cooperators
and then look for sequential equilibria in such games. Of course, it is hard to predict
how tractable these models might be and how much additional light they might shed
on the controversies at hand.
It is also clear that there will continue to be substantial contributions to this lit-
erature from a neuroeconomic perspective as in de Quervain et al. (2004), Fehr and
Camerer (2007), Knoch et al. (2010) and Spitzer et al. (2007) especially in terms of
understanding the motives behind altruistic punishments and norm compliance.
35The possible applications are numerous including charitable contributions (Andreoni 2006; Andreoni
and Petrie 2004; List and Lucking-Reiley 2002; Martin and Randal 2008; Vesterlund 2003), tax compliance
(Andreoni et al. 1998;FreyandTorgler2007), managing natural resources (Cardenas et al. 2002; Carpenter
and Seki 2009;Ostrom1990;Ostrometal.1994), labor relations (Bewley 1999,2005), legal enforcement
(Bohnet et al. 2001; Kahan 2005) and many others, possibly unexplored as of yet.

A. Chaudhuri
Finally, this line of work is expanding upon traditional socio-biological theories
of human cooperation with their emphasis on individual selection such as kin se-
lection (Hamilton 1964), reciprocal altruism (Trivers 1971) or costly signaling (Za-
havi and Zahavi 1997). In fact, the emerging literature on “strong reciprocity” (see
Gintis et al. 2005 for a broad overview) argues that the presence of homo recip-
rocans—conditional cooperators who are willing to punish free riders even if such
punishment is costly to the punishers—may be the primary driving force behind
sustaining cooperation in a variety of social settings. These insights seem to pro-
vide new evidence in favor of group (or multi-level) selection as well as gene-
culture co-evolution (Boyd and Richerson 1985; Cavalli-Sforza and Feldman 1981;
Sober and Wilson 1998) as opposed to selection at the level of the individual. The
findings of this line of work, especially those carried out among small-scale tribal
societies, as in Henrich et al. (2004)orEffersonetal.(2010), will likely provide
valuable clues to human evolutionary processes.
Acknowledgements I am deeply grateful to Jordi Brandts for his advice over multiple revisions. I am
also indebted to Indira Basu, Jeff Carpenter, Gary Charness, Ernst Fehr, Simon Gächter, Martin Kocher,
Nikos Nikiforakis, Louis Putterman, Andrew Schotter, Barry Sopher, Steve Tucker and four anonymous
referees for extensive feedback. I thank Meg Paichayontvijit, Geoff Brooke, Parapin Prak and Lisa Meehan
for research assistance. I gratefully acknowledge the financial assistance provided by the University of
Auckland Research Council. I thank the University of Auckland Business School for granting me research
and study leave which allowed me to work on the paper as well as Marie-Claire Villeval, Izabela Jelovac
and Phillippe Polome of GATE (CNRS) at the University of Lyon and the Department of Economics at
Rutgers University for their hospitality during my visits to those places while working on this survey.
I dedicate this article to my parents-in-law Nandita and Samir Basu for looking after the kids and thereby
giving me time to carry out my research for this article. I am responsible for all errors in the paper, though
Ishannita (5) and Ananrita (2) should share some of the blame.
References
Ahn, T., Isaac, R. M., & Salmon, T. (2009). Coming and going: experiments on endogenous group sizes
for excludable public goods. Journal of Public Economics,93(1–2), 336–351.
Ambrus, A., & Pathak, P. (2009). Cooperation over finite horizons: a theory and experiments. Working
Paper, Department of Economics, Harvard University.
Anderson, C., & Putterman, L. (2006). Do non-strategic sanctions obey the law of demand? The demand
for punishment in the voluntary contribution mechanism. Games and Economic Behavior,51(1),
1–24.
Anderson, S., Goeree, J., & Holt, C. (1998). A theoretical analysis of altruism and decision error in public
goods games. Journal of Public Economics,70, 297–323.
Andreoni, J. (1988). Why free ride? Strategies and learning in public goods experiments. Journal of Public
Economics,37, 291–304.
Andreoni, J. (1990). Impure altruism and donations to public goods: a theory of warm-glow giving. Eco-
nomic Journal,100, 464–477.
Andreoni, J. (1995). Cooperation in public goods experiments: kindness or confusion? American Economic
Review,85(4), 891–904.
Andreoni, J. (2006). Philanthropy. In: S.-C. Kolm & J. Mercier Ythier (Eds.), Handbook of giving, reci-
procity and altruism (pp. 1201–1269). Amsterdam: North Holland.
Andreoni, J., & Croson, R. (2008). Partners versus strangers: random rematching in public goods experi-
ments. In C. Plott & V. L. Smith (Eds.), Handbook of experimental economics results (pp. 776–783).
Amsterdam: North-Holland.
Andreoni, J., & Petrie, R. (2004). Public goods experiments without confidentiality: a glimpse into fund-
raising. Journal of Public Economics,88, 1605–1623.

Sustaining cooperation in laboratory public goods experiments
Andreoni, J., Erard, B., & Feinstein, J. (1998). Tax compliance. Journal of Economic Literature,XXXVI,
818–860.
Asch, S. (1951). Effects of group pressure upon the modification and distortion of judgment. In H. Guet-
zkow (Ed.), Groups, leadership and men. Pittsburgh: Carnegie Press.
Asch, S. (1955). Opinions and social pressure. Scientific American,193, 31–35.
Axelrod, R. (1986). An evolutionary approach to norms. American Political Science Review,80(4), 1095–
1111.
Bardsley, N., & Sausgrüber, R. (2005). Conformity and reciprocity in public good provision. Journal of
Economic Psychology,26, 664–681.
Berg, J., Dickhaut, J., & McCabe, K. (1995). Trust, reciprocity and social history. Games and Economic
Behaviour,10, 122–142.
Bewley, T. (1999). Why don’t wages fall in a recession? Cambridge: Harvard University Press.
Bewley, T. (2005). Fairness, reciprocity and wage rigidity. In H. Gintis, S. Bowles, R. Boyd, & E. Fehr
(Eds.), Moral sentiments and material interests: the foundations of cooperation in economic life (pp.
303–338). Cambridge: MIT Press.
Bochet, O., & Putterman, L. (2009). Not just babble: Opening the black box of communication in a vol-
untary contribution experiment. European Economic Review,53(3), 309–326.
Bochet, O., Page, T., & Putterman, L. (2006). Communication and punishment in voluntary contribution
experiments. Journal of Economic Behavior & Organization,60(1), 11–26.
Bohm, P. (1972). Estimating demand for public goods: an experiment. European Economic Review,3,
111–130.
Bohm, P. (1983). Revealing demand for an actual public good. Journal of Public Economics,24, 135–151.
Bohnet, I., Frey, B., & Huck, S. (2001). More order with less law: on contract enforcement, trust, and
crowding. American Political Science Review,95, 131–144.
Bolton, G., & Ockenfels, A. (2000). ERC—a theory of equity, reciprocity and competition. American
Economic Review,90, 166–193.
Bowles, S., & Gintis, H. (2002). Homo reciprocans. Nature,415, 125–128.
Boyd, R., & Richerson, P. (1985). Culture and the evolutionary process. Chicago: Chicago University
Press.
Brandts, J., & Charness, G. (2000). Hot vs. cold: sequential responses and preference stability in experi-
mental games. Experimental Economics,2(3), 227–238.
Brandts, J., & Charness, G. (2009). The strategy versus the direct response method: a survey of experimen-
tal comparisons. Working paper, Department of Economics, University of California—Santa Barbara
and Department of Business Economics, U. Autonoma de Barcelona.
Brandts, J., & Schram, A. (2001). Cooperation and noise in public goods experiments: applying the con-
tributions function approach. Journal of Public Economics,79, 399–427.
Brandts, J., Saijo, T., & Schram, A. (2004). How universal is behavior? A four country comparison of
spite, cooperation and errors in voluntary contribution mechanisms. Public Choice,119, 381–424.
Brosig, J., Ockenfels, A., & Weimann, J. (2003). The effect of communication media on cooperation.
German Economic Review,4(2), 217–241.
Bryan, J., & Test, M. (1967). Models and helping: naturalistic studies in aiding behavior. Journal of Per-
sonality and Social Psychology,6, 400–407.
Burlando, R., & Guala, F. (2005). Heterogeneous agents in public goods experiments. Experimental Eco-
nomics,8(1), 35–54.
Cardenas, J., Stranlund, J., & Willis, C. (2002). Economic inequality and burden-sharing in the provision
of local environmental quality. Ecological Economics,40, 379–395.
Carpenter, J. (2004). When in Rome: conformity and the provision of public goods. Journal of Socio-
Economics,33(4), 395–408.
Carpenter, J. (2007a). Punishing free-riders: how group size affects mutual monitoring and the provision
of public goods. Games and Economic Behavior,60(1), 31–51.
Carpenter, J. (2007b). The demand for punishment. Journal of Economic Behavior and Organization,
62(4), 522–542.
Carpenter, J., & Seki, E. (2009, forthcoming). Do social preferences increase productiv-
ity? Field experimental evidence from fishermen in Toyama Bay. Economic Inquiry,
doi:10.1111/j.1465-7295.2009.00268.x.
Cavalli-Sforza, L., & Feldman, M. (1981). Cultural transmission and evolution: a quantitative approach.
Princeton: Princeton University Press.
Charness, G., & Rabin, M. (2002). Understanding social preferences with simple tests. Quarterly Journal
of Economics,117, 817–869.

A. Chaudhuri
Charness, G., & Yang, C. (2007). Endogenous group formation and public goods provision: exclusion, exit,
mergers and redemption. Working paper, Department of Economics, University of California-Santa
Barbara. Available. at SSRN: http://ssrn.com/abstract=932251.
Chaudhuri, A. (2009). Experiments in economics: playing fair with money. London: Routledge.
Chaudhuri, A., & Paichayontvijit, T. (2006). Conditional cooperation and voluntary contributions to a
public good. Economics Bulletin,3(8), 1–14.
Chaudhuri, A., & Paichayontvijit, T. (2010a, forthcoming). Recommended play versus costly punishments
in a laboratory voluntary contributions mechanism. In K. Ghosh Dastidar, H. Mukhopadhyay, & U.
Sinha (Eds.), Dimensions of economic theory and policy: essays for Anjan Mukherji. New Delhi:
Oxford University Press.
Chaudhuri, A., & Paichayonvijit, T. (2010b). Does strategic play explain the decay in contributions in
laboratory public goods games? Working paper, Department of Economics, University of Auckland.
Chaudhuri, A., Graziano, S., & Maitra, P. (2006). Social learning and norms in a public goods experiment
with intergenerational advice. Review of Economic Studies,73(2), 357–380.
Cinyabuguma, M., Page, T., & Putterman, L. (2005). Can second order punishment deter perverse punish-
ment? Experimental Economics,9, 265–279.
Cinyabuguma, M., Page, T., & Putterman, L. (2006). Cooperation under the threat of expulsion in a public
goods experiment. Journal of Public Economics,89, 1421–1435.
Cooper, D., & Kagel, J. (2009, forthcoming). Other regarding preferences: a survey of experimental results
In J. Kagel & A. Roth (Eds.), The handbook of experimental economics (Vol. 2). Princeton: Princeton
University Press.
Cox, J., Friedman, D., & Gjerstad, S. (2007). A tractable model of reciprocity and fairness. Games and
Economic Behavior,59, 17–45.
Cox, J., Friedman, D., & Sadiraj, V. (2008). Revealed altruism. Econometrica,76(1), 31–69.
Croson, R. (2007). Theories of commitment, altruism and reciprocity: Evidence from linear public goods
games. Economic Inquiry,45(2), 199–216.
Dawes, R. (1980). Social dilemmas. Annual Review of Psychology,31, 169–193.
Dawes, R., McTavish, J., & Shaklee, H. (1977). Behavior, communication and assumptions about other
people’s behavior in a common dilemma situation. Journal of Personality and Social Psychology,35,
1–11.
Dawes, R., Orbell, J., Simmons, R., & van de Kragt, A. (1986). Organizing groups for collective action.
American Political Science Review,8, 1171–1185.
de Oleveira, A., Croson, R., & Eckel, C. (2009). One bad apple: uncertainty and heterogeneity in public
good provision. Working paper, Department of Resource Economics, University of Massachusetts-
Amherst. .
de Quervain, D., Fischbacher, U., Treyer, V., Schallhammer, M., Schnyder, U., Buck, A., & Fehr, E.
(2004). The neural basis of altruistic punishments. Science,305, 1254–1258.
Denant-Boemont, L., Masclet, D., & Noussair, C. (2007). Punishment, counterpunishment and sanction
enforcement in a social dilemma experiment. Economic Theory,33, 145–167.
Dufwenberg, M., & Kirchsteiger, G. (2004). A theory of sequential reciprocity. Games and Economic
Behavior,47, 268–298.
Efferson, C. H. Bernhard, Fischbacher, U., & Fehr, E. (2010). The ultimate origins of human prosocial be-
havior: an empirical test. Working paper, Institute for Empirical Research in Economics, University
of Zurich.
Egas, M., & Riedl, A. (2008). The economics of altruistic punishment and the maintenance of cooperation.
Proceedings of the Royal Society B: Biological Sciences,275(1637), 871–878.
Ehrhart, K.-M., & Keser, C. (1999). Cooperation and mobility: on the run. Working paper, CIRANO and
University of Karlsruhe.
Ertan, A., Page, T., & Putterman, L. (2009). Who to punish? Individual decisions and majority rule in
mitigating the free rider problem. European Economic Review,53(5), 495–511.
Falk, A., & Fischbacher, U. (2006). A theory of reciprocity. Games and Economic Behavior,54, 293–315.
Fehr, E., & Camerer, C. (2007). Social neuroeconomics: the neural circuitry of social preferences. Trends
in Cognitive Science,11(10), 419–427.
Fehr, E., & Fischbacher, U. (2004a). Third party punishment and social norms. Evolution and Human
Behavior,25, 63–87.
Fehr, E., & Fischbacher, U. (2004b). Social norms and human cooperation. Trends in Cognitive Sciences,
8(4), 185–190.
Fehr, E., & Fischbacher, U. (2005a). Altruists with green beards. Analyse & Kritik,27(1), 73–84.

Sustaining cooperation in laboratory public goods experiments
Fehr, E., & Fischbacher, U. (2005b). Human altruism: proximate patterns and evolutionary origins. Analyse
& Kritik,27(1), 6–47.
Fehr, E., & Gächter, S. (2000). Cooperation and punishment in public goods experiments. American Eco-
nomic Review,90(4), 980–994.
Fehr, E., & Gächter, S. (2002). Altruistic punishment in humans. Nature,415, 137–140.
Fehr, E., & Schmidt, K. (1999). A theory of fairness, competition and cooperation. Quarterly Journal of
Economics,114, 817–868.
Fischbacher, U., & Gächter, S. (2009). On the behavioral validity of the strategy method in public good
experiments. Discussion Paper No. 2009-25, Centre for Decision Research and Experimental Eco-
nomics, University of Nottingham.
Fischbacher, U., & Gächter, S. (2010). Social preferences, beliefs, and the dynamics of free riding in public
good experiments. American Economic Review,100(1), 541–556.
Fischbacher, U., Gächter, S., & Fehr, E. (2001). Are people conditionally cooperative? Evidence from a
public goods experiment. Economics Letters,71(3), 397–404.
Frey, B., & Meier, S. (2004). Social comparisons and pro-social behavior. Testing ‘conditional coopera-
tion’ in a field experiment. American Economic Review,94(5), 1717–1722.
Frey, B., & Torgler, B. (2007). Tax morale and conditional cooperation. Journal of Comparative Eco-
nomics,35, 136–159.
Gächter, S. (2007). Conditional cooperation. Behavioral regularities from the lab and the field and their
policy implications. In B. Frey & A. Stutzer (Eds.), Economics and psychology. A promising new
cross-disciplinary field.CESifo Seminar Series. Cambridge: MIT Press.
Gächter, S., & Hermann, B. (2009). Reciprocity, culture and human cooperation: previous insights and a
new cross cultural experiment. Philosophical Transactions of the Royal Society B,364, 791–806.
Gächter, S., & Herrmann, B. (2010, forthcoming). The limits of self-governance when cooperators
get punished: experimental evidence from urban and rural Russia. European Economic Review.
doi:10.1016/j.euroecorev.2010.04.003.
Gächter, S., & Thöni, C. (2005). Social learning and voluntary cooperation among like-minded people.
Journal of the European Economic Association,3(2–3), 303–314.
Gächter, S., & Thöni, C. (2007). Rationality and commitment in voluntary cooperation: insights from ex-
perimental economics. In P. Fabienne & H. B. Schmidt (Eds.), Rationality and Commitment. Oxford:
Oxford University Press.
Gächter, S., Renner, E., & Sefton, M. (2008). The long run benefits of punishment. Science,322, 1510.
Geanakoplos, J., Pearce, D., & Stacchetti, E. (1989). Psychological games and sequential rationality.
Games and Economic Behavior,1, 60–79.
Gintis, H., Bowles, S., Boyd, R., & Fehr, E. (Eds.) (2005). Moral sentiments and material interests.Cam-
bridge: MIT Press.
Griesinger, D., & Livingston, J. (1973). Toward a model of interpersonal motivation in experimental
games. Behavioral Science,18, 173–188.
Guala, F. (2010). Reciprocity: weak or strong? what punishment experiments do (and do not)
demonstrate. IDEAS Working Paper 2010-23, University of Milan. http://ideas.repec.org/p/mil/
wpdepa/2010-23.html.
Gunnthorsdottir, A., Houser, D., & McCabe, K. (2007). Disposition, history and contributions in public
goods experiments. Journal of Economic Behavior & Organization,62(2), 304–315.
Gürerk, Ö., Irlenbusch, B., & Rockenbach, B. (2006). The competitive advantage of sanctioning institu-
tions. Science,312, 108–111.
Hamilton, W. (1964). The genetical evolution of social behavior. Journal of Theoretical Biology,37, 1–52.
Henrich, J., & Boyd, R. (2001). Why people punish defectors: weak conformist transmission can stabilize
costly enforcement of norms in cooperative dilemmas. Journal of Theoretical Biology,208, 79–89.
Henrich, J., Heine, S., & Norenzayan, A. (2010). Most people are not WEIRD. Nature,466, 29.
Henrich, J., Boyd, R., Bowles, S., Camerer, C., Fehr, E., & Gintis, H. (Eds.) (2004). Foundations of hu-
man sociality: economic experiments and ethnographic evidence in fifteen small-scale societies.New
York: Oxford University Press.
Hermann, B., & Thöni, C. (2008). Measuring conditional cooperation: a replication study in Russia. Ex-
perimental Economics,12(1), 87–92.
Hermann, B., Thöni, C., & Gächter, S. (2008). Antisocial punishments across societies. Science,319,
1362–1367.
Isaac, R., & Walker, J. (1988a). Communication and free riding behavior: the voluntary contributions
mechanism. Economic Inquiry,26(4), 585–608.

A. Chaudhuri
Isaac, R., & Walker, J. (1988b). Group size effects in public goods provision: the voluntary contributions
mechanism. Quarterly Journal of Economics,103, 179–199.
Isaac, R., Walker, J., & Thomas, S. (1984). Divergent evidence on free riding: an experimental examination
of possible explanations. Public Choice,43, 113–149.
Isaac, R., McCue, K., & Plott, C. (1985). Public goods provision in an experimental environment. Journal
of Public Economics,26, 51–74.
Isaac, R., Walker, J., & Williams, A. (1994). Group size and the voluntary provision of public goods.
Journal of Public Economics,54, 1–36.
Kahan, D. (2005). The logic of reciprocity: trust, collective action, and law. In H. Gintis, S. Bowles, R.
Boyd, & E. Fehr (Eds.), Moral sentiments and material interests: the foundations of cooperation in
economic life (pp. 339–378). Cambridge: MIT Press.
Kelley, H., & Stahelski, A. (1970). Social interaction basis of cooperators’ and competitors’ beliefs about
others. Journal of Personality and Social Psychology,16, 66–91.
Keser, C., & van Winden, F. (2000). Conditional cooperation and voluntary contributions to public goods.
Scandinavian Journal of Economics,102(1), 23–39.
Kim, O., & Walker, M. (1984). The free rider problem: experimental evidence. Public Choice,43, 3–24.
Knoch, D., Giannoti, L., Baumgartner, T., & Fehr, E. (2010). A neural marker of costly punishment be-
havior. Psychological Science,21(3), 337–342.
Kocher, M., Cherry, T., Kroll, S., Netzer, R., & Sutter, M. (2008). Conditional cooperation on three conti-
nents. Economics Letters,101(3), 175–178.
Kosfeld, M., Okada, A., & Riedl, A. (2009). Institution formation in public goods games. American Eco-
nomic Review,99(4), 1335–1355.
Kreps, D., Milgrom, P., Roberts, J., & Wilson, R. (1982). Rational cooperation in the finitely repeated
prisoners’ dilemma. Journal of Economic Theory,27, 245–252.
Kurzban, R., & Houser, D. (2005). An experimental investigation of cooperative types in human groups:
a complement to evolutionary theory and simulations. Proceedings of the National Academy of Sci-
ences,102(5), 1803–1807.
Ledyard, O. (1995). Public goods: some experimental results. In J. Kagel & A. Roth (Eds.), Handbook of
experimental economics. Princeton: Princeton University Press (Chap. 2).
Liebrand, W. (1984). The effects of social motives, communication and group size on behavior in an n-
person multi stage mixed motive game. European Journal of Social Psychology,14, 239–264.
List, J., & Lucking-Reiley, D. (2002). The effects of seed money and refunds on charitable giving: experi-
mental evidence from a university capital campaign. Journal of Political Economy,110(1), 215–233.
Martin, R., & Randal, J. (2008). How is donation behavior affected by the donations of others? Journal of
Economic Behavior and Organization,67(1), 228–238.
Marwell, G., & Ames, R. (1979). Experiments on provision of public goods I: resources, interest, group
size, and the free riding problem. American Journal of Sociology,84(6), 1335–1360.
Marwell, G., & Ames, R. (1980). Experiments on provision of public goods II: provision point, stake,
experience, and the free riding problem. American Journal of Sociology,85(4), 926–937.
Marwell, G., & Ames, R. (1981). Economists free ride, does anyone else? Journal of Public Economics,
15, 295–310.
Masclet, D., Noussair, C., Villeval, M., & Tucker, S. (2003). Monetary and nonmonetary punishment in
the voluntary contributions mechanism. American Economic Review,93(1), 366–380.
Moscovici, S. (1985). Social influence and conformity. In G. Lindzey& E. Aronson (Eds.), The Handbook
of Social Psychology. New York: Random House.
Nikiforakis, N. (2008). Punishment and counter-punishment in public good games: can we really govern
ourselves? Journal of Public Economics,92, 91–112.
Nikiforakis, N. (2010). Feedback, punishment and cooperation in public goods experiments. Games and
Economic Behavior,68, 689–702.
Nikiforakis, N., & Normann, H. (2008). A comparative statics analysis of punishment in public good
experiments. Experimental Economics,11, 358–369.
Noussair, C., & Tucker, S. (2005). Combining monetary and social sanctions to promote cooperation.
Economic Inquiry,43(3), 649–660.
Offerman, T., Sonnemans, J., & Schram, A. (1996). Value orientations, expectations and voluntary contri-
butions in public goods. Economic Journal,106(437), 817–845.
Olson, M. (1965). The logic of collective action: public goods and the theory of groups. Cambridge:
Harvard University Press.

Sustaining cooperation in laboratory public goods experiments
Ones, U., & Putterman, L. (2007). The ecology of collective action: a public goods and sanctions ex-
periment with controlled group formation. Journal of Economic Behavior & Organization,62(4),
495–521.
Orbell, J., Dawes, R., & van de Kragt, A. (1990). The limits of multilateral promising. Ethics,100, 616–
627.
Ostrom, E. (1990). Governing the commons: the evolution of institutions for collective action. Cambridge:
Cambridge University Press.
Ostrom, E., Walker, J., & Gardner, R. (1992). Covenants with and without a sword: self-governance is
possible. American Political Science Review,86(2), 404–417.
Ostrom, E., Gardner, R., & Walker, J. (Eds.) (1994). Rules, games, and common pool resources.Ann
Arbor: University of Michigan Press.
Page, T., Putterman, L., & Unel, B. (2005). Voluntary association in public goods experiments: reciprocity,
mimicry, and efficiency. Economic Journal,115, 1032–1053.
Palfrey, T., & Prisbrey, J. (1997). Anomalous behavior in public goods experiments: how much and why?
American Economic Review,87, 829–846.
Rabin, M. (1993). Incorporating fairness into game theory and economics. American Economic Review,
80(5), 1281–1302.
Rege, M., & Telle, K. (2004). The impact of social approval and framing on cooperation in public good
situations. Journal of Public Economics,88(7–8), 1625–1644.
Rockenbach, B., & Milinski, M. (2006). The efficient interaction of indirect reciprocity and costly punish-
ment. Nature,444, 718–723.
Seely, B., Van Huyck, J., & Battalio, R. (2005). Credible assignments can improve efficiency in laboratory
public goods games. Journal of Public Economics,89, 1437–1455.
Sefton, M., Shupp, R., & Walker, J. (2007). The effect of rewards and sanctions in provision of public
goods. Economic Inquiry,45(4), 671–690.
Selten, R. (1967). Die Strategiemethode zur Erforschung des eingeschränkt rationalen Verhaltens im Rah-
men eines Oligopolexperiments. In H. Sauermann (Ed.), Beiträge zur experimentellen Wirtschafts-
forschung (Vol. I, pp. 136–168). Tübingen: Mohr.
Sober, E., & Wilson, D. (1998). Unto others: the evolution and psychology of unselfish behavior.Cam-
bridge: Harvard University Press.
Sonnemans, J., Schram, A., & Offerman, T. (1999). Strategic behavior in public good games: when partners
drift apart. Economics Letters,62(1), 35–41.
Spitzer, M., Fischbacher, U., Herrnberger, B., Grön, G., & Fehr, E. (2007). The neural signature of social
norm compliance. Neuron,56, 185–196.
Trivers, R. (1971). The evolution of reciprocal altruism. Quarterly Review of Biology,46, 36–57.
Vesterlund, L. (2003). The informational value of sequential fundraising. Journal of Public Economics,
87, 627–657.
Walker, J., & Halloran, M. (2004). Rewards and sanctions and the provision of public goods in one-shot
settings. Experimental Economics,7(3), 235–247.
Yamagishi, T. (1986). The provision of a sanctioning system as a public good. Journal of Personality and
Social Psychology,51(1), 110–116.
Yamagishi, T. (1988). The provision of a sanctioning system in the United States and Japan. Social Psy-
chology Quarterly,51(3), 265–271.
Zahavi, A., & Zahavi, A. (1997). The handicap principle: a missing piece of Darwin’s puzzle.NewYork:
Oxford University of Press.