European Economic Review 159 (2023) 104561

Contents lists available at ScienceDirect

European Economic Review

journal homepage: www.elsevier.com/locate/eer

Asymmetric heterogeneities and the role of transfers in a public

goods experiment
Ann-Kathrin Koessler a, *, Julia Müller b, Sonja Zitzelsberger c
a
Institute of Environmental Planning, Leibniz University Hannover, Germany
b
School of Business Administration and Economics, Osnabrück University, Germany
c
Department of Economics, University of Kassel, 34117 Kassel, Germany

A R T I C L E I N F O A B S T R A C T

JEL Classification: Previous experimental research has shown that cooperation is especially challenging in situations
H4 involving heterogeneous actors. Here, we investigate the effect of allowing unconditional trans­
C92 fers in a public goods game when actors differ asymmetrically in their endowments and pro­
C72
ductivity levels. Under this setup, full efficiency, in terms of highest maximum group payoff, can
Keywords: only be achieved with the full transfer of resources from the players with high endowment and
Public goods game
low productivity to the players with high productivity but low endowment. We show that, in this
Heterogeneity
setting, the availability of transfers enhances public good provision and cooperation. The
Transfers
Inequality voluntary transfer serves as a powerful cooperative signal, leading to high contribution rates
Cooperation among the recipients. However, if the transfer possibility is given but not used, cooperation
erodes.

1. Introduction

When there are incentives to free-ride, cooperation in social dilemma situations can be challenging to achieve. Research inside and
outside the laboratory has focused intensively on discerning the factors that hamper or foster cooperation. A central finding is that
cooperation tends to be more difficult when agents are heterogeneous. For example, Cherry et al. (2005) compare groups with ho­
mogenous income to groups with heterogeneous income, finding significantly lower cooperation rates in the latter. Fisher et al. (1995)
show similar results for inequality in returns from the public good, as do de Oliveira et al. (2015) for inequality in social preferences.
Previous research has also highlighted that measures designed to foster cooperation tend to work differently and are often less suc­
cessful in heterogeneous groups than in homogenous groups (see Nikiforakis et al. (2012), Reuben and Riedl (2013), and Kölle (2015)
for punishment; Levati et al. (2007) for leadership; and Hackett et al. (1994) and Chan et al. (1999) for communication). Cooperation
can be achieved in homogenous groups if group members agree, implicitly or explicitly, on a common fairness principle or contribution
norm (Fehr and Fischbacher, 2004; Koessler et al., 2021). In heterogeneous groups, however, such an agreement is harder to achieve
because each group member may prefer a different norm and may interpret fairness in a self-serving way (Kesternich et al., 2014;
Kingsley, 2016). Thus, a contribution norm perceived to be fair in a homogenous group is not necessarily agreed upon in a hetero­
geneous group.
Groups and societies outside the laboratory are rarely homogenous and are often characterized by heterogeneities in multiple
dimensions. However, most experimental research to date considers homogenous groups or focuses on heterogeneity in only one

* Corresponding author.
E-mail address: [email protected] (A.-K. Koessler).

https://doi.org/10.1016/j.euroecorev.2023.104561
Received 27 January 2021; Received in revised form 20 July 2023; Accepted 7 August 2023
Available online 8 August 2023
0014-2921/© 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
A.-K. Koessler et al. European Economic Review 159 (2023) 104561

aspect. This is a severe limitation, as multiple dimensions of heterogeneity can have reinforcing or mitigating effects on cooperation
(Kölle et al., 2016; Hauser et al., 2019).1
This paper presents results from an experiment using a public goods game where players differ asymmetrically in their endowments
and productivity levels. Individuals with higher endowments have a lower productivity than individuals with lower endowments. In
situations characterized by such an asymmetric heterogeneity, efficiency gains can be realized through an ex-ante redistribution of
endowment by a central authority or, in the absence of an authority with redistribution powers, by voluntary transfers among the
actors. The voluntary transfers may additionally serve as a signaling tool for cooperation and reciprocity, similar to the effect of a
leader who contributes first (Güth et al., 2007). To empirically assess how a transfer option affects behavior, we introduce a transfer
stage to the game. This stage occurs before the contribution decisions take place and allows the higher-endowed individuals with lower
productivity to transfer (some of) their endowment to the lower-endowed individuals with higher productivity.
Situations in which agents differ in their endowment or productivity are ubiquitous and, in many cases, two heterogeneities
coincide in an asymmetric way. Imagine a generic case of members of a team working on a common project. Each team member has
different amounts of resources that (s)he can invest in the project (e.g., time, money, or equipment), and at the same time, each
member has different productivity levels for a given unit of resource invested. Up to a certain point, transferring resources to the team
member with the highest productivity implies efficiency improvements. Other examples can be found in areas such as climate change
mitigation. Marginal abatement costs are frequently smaller for some individuals, firms, or countries. For example, developing
countries can have lower abatement costs compared to developed countries, whereas the latter may have more available resources to
invest in abatement (OECD, 2009). The forest partnership between Norway and Brazil can serve as an illustrative example. Defor­
estation and forest degradation of the Amazon accounted for approximately 5% of global emissions in 2004 (Boucher et al., 2013). In
2008, Norway promised to pay up to 1 billion USD to Brazil’s Amazon Fund, which is owned and managed by Brazilian bodies,
conditional on reductions in deforestation rates. Additionally, the Memorandum of Understanding foresaw “contributions from
Norway to the Amazon Fund by means of possible initial donations for 2008 and 2009″.2 In fact, Norway made an upfront payment of
100 million NOK in 2007 and 600 million NOK in 2008.3 Brazil achieved an impressive reduction in deforestation over the following
years: while an estimated area of 12,911 km2 was cut in 2008, only 6947 km2 were lost in 2017, with even lower losses in between.
Honoring its pledge, Norway paid a total of around 900 million USD into the Amazon Fund until 2018 (Ministry of Climate and
Environment, 2018). In contrast to most other bilateral funding commitments, trust played a central role in the Norway-Brazil
Agreement. Norway made advanced payments and had no direct supervision or involvement in the Amazon Fund’s governance.
This success story ended with a change in the Brazilian government in 2019 and a subsequent significant reduction of efforts to uphold
environmental protection laws. Consequently, Norway suspended donations (Norway’s International Climate and Forest Initiative,
2020).
Transfers of resources between actors may thus have the potential to uncover substantial efficiency gains in common interests, such
as the mitigation of greenhouse gases. However, when there are incentives to free-ride, it is far from certain that recipients will use the
transfers as intended. Monitoring and punishment are often not possible, which may prevent the occurrence of transfers in the first
place.
This paper aims to provide novel insights in this regard. Employing an abstract decision setting in an economic experiment, we
investigate the effects of introducing transfer possibilities on individual behavior and overall public good provision. Examining this
connection experimentally allows for a clean comparison of behavior with and without transfer possibility. Specifically, we are
interested in (i) whether individuals make use of the transfer possibility, (ii) how the existence and realization of the transfer affect the
cooperative behavior of recipients compared to a baseline where no transfers are possible, and (iii) whether the transfer possibility
increases the overall public good size compared to the baseline. With the help of a second treatment, in which the transfers were
implemented exogenously, we can disentangle the effect of an intentional, voluntary transfer from the pure redistribution effect. The
results show that the possibility of voluntary transfers indeed increases the size of the public good. Most subjects with high endowment
and low productivity transfer their full endowment to the players with low endowment but high productivity, and these players
reciprocate with high cooperation rates. However, if the transfer possibility is not taken advantage of, cooperation rates are lower than
in baseline. The comparison with the exogenous treatment reveals that the redistribution of resources alone does not drive the increase
in cooperation rates but that the deliberate decision to transfer is required for high cooperation rates, underlining the vital role of
intention-based reciprocity.

1
Both studies are discussed in detail in the literature section. In short, Kölle et al. (2016) show theoretically that in case of heterogeneity in
abilities, introducing complementarily ex-ante heterogeneity in wealth can lead to an increase in welfare. Hauser et al. (2019) use experiments and
simulations to compare how cooperation rates vary among different combinations of endowment and productivity heterogeneity.
2
The Memorandum is available at https://www.regjeringen.no/contentassets/2ecbe3693ac04a85bf4d8ddb5d78d858/mou_norway_brazil.16.09.
08.pdf (accessed 09.05.2023)
3
One Norwegian krone (NOK) is approximately 0.1 USD.

2
A.-K. Koessler et al. European Economic Review 159 (2023) 104561

2. Literature

Research that shares our interest in multiple, particularly asymmetric heterogeneities in public good settings is the paper by Hauser
et al. (2019). In this study, the authors consider reciprocity among unequal individuals on the basis of evolutionary simulations and
experimental data from an online two-player public goods game.4 Five settings are compared: (1) full equality, (2) endowment
inequality, (3) productivity inequality, (4) symmetric inequality with the more productive player also having a higher endowment, and
(5) asymmetric inequality with the more productive player having a lower endowment. Their results show that when players are
equally productive but their endowments differ (2), relative contributions are significantly lower than in the standard situation with
full equality (1). When the productivity of players differs, no statistically significant difference is found between the setting with
productivity inequality (3) and symmetric inequality (4). When, however, endowment and productivity differ asymmetrically (5),
relative contributions are the lowest overall. Players with high endowment but low productivity drive this result; their relative
contributions are significantly lower than under full equality (1) and symmetric inequality (4). In summary, situations involving
asymmetric inequality in endowment and productivity pose a particular challenge for cooperation.
In a theoretical paper, Kölle et al. (2016) show that in groups where individuals are inequality averse and differ in their pro­
ductivity, introducing ex-ante symmetric inequality in wealth can lead to an increase in welfare.5 The higher ex-ante wealth induces
the high-productivity individuals to contribute more, from which the low-wealth-low-productivity individuals benefit, leading to
ex-post equality. This finding hints at the potential that transfers may have in our setting, transforming asymmetric inequality into a
symmetric inequality situation and thus inducing cooperation.
Moving to general insights on heterogeneities in public good games, there are four different ways in which heterogeneity can be
implemented in productivity: (1) changes in the marginal per capita return (MPCR) of selected individuals (e.g., Croson and Marks,
2001; Nikiforakis et al., 2012; Reuben and Riedl, 2009, 2013), (2) changes in the price for the private good for selected individuals (e.
g., Fisher et al., 1995), (3) different MPCRs for contributions from selected individuals for all group members (e.g., Tan, 2008;
Noussair and Tan, 2011), and (4) multiplication of the contributions of selected individuals with a factor larger than 1 (henceforth
called the productivity parameter) (e.g., Kölle, 2015; Hauser et al., 2019).6 While the two former and the two latter approaches are
theoretically equivalent, they differ in certain practical ways. For example, all approaches reduce the individual costs of contribution
for the more productive individuals, but the latter additionally include a positive externality of a player’s contribution to all group
members. Increasing the contribution level of the highly productive type increases payoffs for all group members. This distinction does
not play a role with standard preferences, but it does in the case of social preferences. This difference, however, has not received much
attention in the literature.7 Empirically, most papers find a negligible effect of (singular) heterogeneity in productivity on average
contributions (Fisher et al., 1995; Hauser et al., 2019).
While there is only a limited literature on productivity, many papers have investigated the effects of endowment heterogeneity.
Substantial asymmetries in endowments mostly have a negative effect on contributions to the public good. Higher-endowed players
give less in relative terms than lower-endowed players (Zelmer, 2003; Cherry et al., 2005; Buckley and Croson, 2006; Hargreaves Heap
et al., 2016). However, a handful of studies also find no change in contribution levels (Hofmeyr et al., 2007; Levati et al., 2007) or even
an increase among the higher-endowed players (van Dijk and Wilke, 1994). Fung and Au (2014) show experimentally that the negative
effect of endowment inequality on cooperation is mediated by its skewness distribution. With increasing inequality, the decrease in
cooperation rates is less severe in groups with a highly skewed distribution (one high-endowed individual and two low-endowed
individuals) compared to a more symmetric setup (one high-, one medium-, and one low-endowed individual). The authors argue
that the self-efficacy of the high-endowed individual increases with the skewness of the distribution, which in turn motivates higher
contributions by the high-endowed individual. Synthesizing the insights from previous studies, heterogeneity in endowments seems to
generally hinder cooperation.
To the best of our knowledge, no paper has tested the effect of unconditional ex-ante transfers (i.e., transfers from one player to
another before contribution decisions are made) within groups in a public goods setting. Blanco et al. (2018) studied a scenario in
which outsiders, who benefit from the public good but cannot contribute to it, can make ex-ante transfers to insiders who benefit from
and can contribute to the public good. They find that such transfers do not significantly increase cooperation rates compared to a
baseline without transfers. With the possibility of transfers, insiders condition their contributions mainly on the received transfer and
no longer on their co-players’ contributions, but each token transferred elicits only around 1/3 tokens of contribution. A potential

4
In their experiment, the public goods game was repeated for at least 20 rounds. After that, the game continued with a probability of 50%.
5
Kölle et al.’s (2016) finding that introducing symmetry into the inequalities can be beneficial differs here from that of Hauser et al. (2019), who
found no significant difference in average contributions between the case of singular inequality in productivity and symmetric inequality. The
research approach and specifications in both studies differed. Future research may want to investigate what caused this difference in results.
Important for our study is the challenge asymmetric inequalities pose for cooperation.
6
In the standard public goods game, the productivity parameter is 1 and usually not reported.
7
Fellner et al. (2011), for example, introduced different productivity parameters but compared their results to those of Fisher et al. (1995), who
focused on differences in MPCRs. A study by Kölle (2015) explicitly tested the differences between different productivity parameters and different
MPCRs, with making the more productive player preferential (i.e., the effective MPCR of the more productive player is larger than one). Thus, the
dominant strategy of the preferential player is to contribute fully to the public good. Kölle found that heterogeneous MPCRs have no effect on
voluntary contributions and even a detrimental effect when punishment is possible. Heterogeneous productivity parameters, in contrast, had a
positive and stabilizing effect on contributions, both with and without punishment.

3
A.-K. Koessler et al. European Economic Review 159 (2023) 104561

Table 1
Overview of treatments.
Treatment Endowment Productivity Transfer Number of groups

Baseline eA> eB pA < pB No 40

Transfer eA> eB pA < pB Yes 40
25
Exogenous eA − t
<
eB + t pA < pB Exogenous 39
>

Note: 12 – 30 subjects participated in each session with a fixed group size of three subjects, resulting in a total of 357 subjects.
25
In Exogenous one group less was conducted, due to lower show up in one session. This means that for one group in Transfer no counter group
exists in Exogenous. Given that the main results presented in the following do not differ regardless of whether we include or exclude this group, we
present the results on the basis of the full dataset.

reason for this feeble response lies in the fact that transfers were given to the entire group of insiders (i.e., they were split equally). This
constellation may have impeded the development of a stronger reciprocal relationship between transfer senders and recipients
compared to the case of a one-to-one transfer.
The remaining literature focuses primarily on (enforceable) contracts, conditional on a pre-specified strategy or contribution level
(e.g., Andreoni and Varian, 1999; Charness et al., 2007; Kimbrough and Sheremeta, 2013) or post-play side payments (e.g., Bruttel and
Güth, 2018; Bruttel et al., 2022; Chatziathanasiou et al., 2021, 20228) in different game settings. A common finding is that side
payments are frequently used. Possible explanations for this are that subjects use side payments to express their cooperative intentions
or to offset payoff inequalities. Side payments can induce stable cooperation if they move the cooperation equilibrium to the interior of
the Nash Equilibria region. But even if they fail to do so, cooperation rates are often still higher when side payments are offered. This
suggests that side payments generally have the potential to help players coordinate on cooperation.
Lastly, the sequential order in the transfer treatment in our experiment also has some similarities to public goods games with
leadership, which suggests that the leader’s decisions positively impact the other players’ consequent decisions (e.g., Güth et al.,
2007). Contributions tend to be higher when one group member, the leader, contributes first, compared to a baseline with simulta­
neous contributions. Levati et al. (2007) confirm a positive effect of leadership on contributions in groups with endowment hetero­
geneity if the distribution of endowment is known.

3. Experimental design and procedures

Our experiment is based on a repeated public goods game where players are asymmetrically heterogeneous in endowment and
productivity. Three treatment conditions are implemented: a baseline, a transfer treatment, and a treatment with an exogenous
transfer. The details are outlined in the following sections.
During the experiment, we used neutral language and took great care to avoid framing effects. The experiment instructions that we
handed to the participants can be found in the online appendix.9

3.1. The game

We utilize a linear public goods game with groups of three players. One player receives a high endowment of eA = 30 tokens (type-A
player) and two players receive a low endowment of eB = 15 tokens (type-B players). All players simultaneously decide how much of
their endowment they want to invest in the public good. Contributions by type-B players cB are more productive than contributions by
type-A players, denoted with cA . This heterogeneity in productivity is implemented by doubling their investment in the public good.
The productivity parameter, with which investments into the public good are multiplied, is pA = 1 for type-A players and pB = 2 for
type-B players. This reduces the cost of contributing to the public good for type-B players and increases everyone’s payoff from the
public good. The MPCR a is fixed at 0.4 for all players.10
Payoffs are given by
( )
∑2 ∑
2
πA = eA − tj − cA + a pA cA + pB cBj
j=1 j=1

and
( )
∑
2
πBj = eB + tj − cBj + a pA cA + pB cBj .
j=1

8
Chatziathanasiou et al. study redistributive transfers in a polit-economic context—in their battle of the sexes game, high-ranked players can send
post-play transfers to low-ranked players to “reward” their compliance with an established hierarchical order. Thus, redistributive transfers can, via
indirect reciprocity, be used as a means to avoid wealth-consuming disputes (in future interactions).
9
https://osf.io/9n7d3/?view_only=208c2ea1e96c42b795ea0df507e294d9
10
In short, contributions by type A-players have a multiplier of 1.2 and contributions by type-B players a multiplier of 2.4.

4
A.-K. Koessler et al. European Economic Review 159 (2023) 104561

The transfer payments are denoted by tj ∈ {0,15,30} and we elaborate on these transfers in detail in the next section. The subindex j
identifies the two type-B players. At the end of each round, players are informed about all contributions and payoffs in their group.

3.2. Experimental treatments

The treatment variations are implemented by an antecedent transfer stage, which takes place in some treatments before the
contribution decision. Three treatment conditions exist: Baseline, Transfer, and Exogenous. Table 1 provides an overview.
Baseline is the above-described game without an antecedent transfer stage and transfer options, i.e. tj = 0, ∀ j. In the treatment
condition Transfer, type-A players can decide before play whether they want to make a transfer. Specifically, they are given the
following six options: (1) no transfer, (2) transfer of 30 tokens to the first type-B player, (3) transfer of 30 tokens to the second type-B
player, (4) 15 tokens to each type-B player, (5) 15 tokens to the first type-B player, or (6) 15 tokens to the second type-B player. Thus,
the available options encompass the purely selfish and the socially optimal decision (zero and full transfer) as well as a middle ground,
in which type-A players retain half of their endowment. After the transfer decision is made by type-A players, type-B players are
informed of the choice of type-A players (i.e., the realized transfers).
Due to the resource re-allocation and differences in productivity of the two player types, the maximum amount of possible public
good provision changes between Baseline and Transfer. To account for this, we implemented the additional treatment condition
Exogeneous. Here the transfer patterns in Transfer are exogenously imposed. Each group in Transfer has a counter group in Exogeneous
where the same transfers per round are implemented exogenously. Subjects in Exogeneous are informed about everyone’s round
endowment at the beginning of each round, but subjects are unaware that their endowment may have been modified due to an
exogenous transfer. With these three treatment conditions, it is possible to isolate the effect of transfers from the effect of increased
potential gains from the public good.

3.3. Implementation of the game in the experiment

The experiment consists of 10 consecutive rounds. The total payoff in the game is the sum of all the rounds’ payoffs. The data
presented in the manuscript come from sessions conducted using the partner-matching protocol (i.e., groups were fixed and players
were identified by an ID), where matching groups remained the same across all ten rounds of the game. Thus, a build-up of reputation
was possible. To test the importance of reputational effects in transfer decisions, we ran additional sessions using the stranger-
matching protocol (details can be found in the online appendix).

3.4. Procedures and participants

For the experimental sessions, we used the software z-Tree (Fischbacher, 2007) and the online recruiting software hroot (Bock
et al., 2014). Upon entering the laboratory, each participant was randomly assigned a seat and, accordingly, their role in the exper­
iment. They then answered a short questionnaire to elicit risk, trust, and inequality preferences. Afterwards, instructions were handed
out and read aloud to ensure common knowledge. Only after the subjects had answered all control questions and no queries remained
was the experiment allowed to begin. After the game, a short questionnaire was administered, and subjects were paid according to
their earnings in the game. A typical session lasted one hour, and subjects earned, on average, 17 euros. The exchange rate was 1 token
to 0.05 EUR.
The experimental data were collected in two waves. In the first wave, which took place from November 2019 to January 2020, 186
observations were collected at the experimental laboratory of the University of Magdeburg.11 During the second wave, in September
2022, an additional 171 observations were collected at the experimental laboratory of the University of Hamburg. All regression
models include fixed effects for the second data collection.12
In total, our data encompasses 353 observations. Table A1 in the appendix provides an overview of subjects’ characteristics by
treatment. Participants are, on average, 25.2 years old, and 66% are female.

4. Behavioral predictions

Under the assumption that subjects are fully rational and only interested in maximizing their own monetary payoff, the dominant
strategy for each player—and thus the subgame perfect Nash equilibrium—is zero contributions in all treatments, meaning no transfers
will be made. Thus, the asymmetric setup and possibility of transfers do not alter the standard prediction for public goods games of zero
contributions.
Of course, abundant evidence by now has shown that human behavior is influenced by more than purely monetary preferences.
Contributions to public goods are commonly positive (Chaudhuri, 2011), motivated by cooperation norms and conditional cooperation
(Fischbacher et al., 2001). Moreover, in heterogenous settings, a considerable number of cooperative acts are found, particularly when

11
The supplementary data gleaned from the sessions conducted using the stranger-matching protocol were also collected during the first wave. The
respective analysis is presented in the online appendix.
12
The pre-registration of the experiment can be found here: https://osf.io/ts29r/?view_only=6c12b5b41ada4b898bd5a79d45985f76. Please note
that the presented work deviates from the original study plan, a second data collection took place, and the hypotheses have changed.

5
A.-K. Koessler et al. European Economic Review 159 (2023) 104561

Fig. 1. Average relative group contribution and public good size over time by treatment. The left panel shows the average group contribution in
percentage, averaged over groups and measured as contributions relative to endowments. The right panel shows the average public good size in
tokens, measured as contributed tokens multiplied by the respective productivity factor and averaged over groups.

the inequality stems from productvity or symmetric heterogeneity in productiivity and endowment (Fisher et al., 1995; Hauser et al.,
2019). People’s dislike for inequality motivates redistributional actions (Kölle et al., 2016; Bolton and Ockenfels, 2000).
In our experiment, players face an asymmetric heterogeneity in productivity and endowment. Players have the option to (i)
transform the setting into one with symmetric heterogeneity where previous research has shown cooperation can be more easily
established and (ii) obtain substantial efficiency gains in the case of cooperation. However, to materialize the efficiency gains and
potentially reach the social optimum, the higher-endowed type-A players need to trust and transfer their (entire) endowment, while
the type-B players need to reciprocate the trusting action by contributing (all) available funds.
Inequality aversion and the outlook of efficiency gains can motivate type-A players to realize a transfer. The transfer dynamic
thereby resembles features of a trust game (Berg et al., 1995; McCabe et al., 1998), where one player (sender) is given the opportunity
to transfer any part of her/his endowment to a second player (receiver), which is then multiplied by a factor > 1, after which the
receiving player can decide if and on how much to return from this amount.13 Experimental work shows that senders commonly make
a transfer and receivers return in response significant parts of the transfer. Berg et al. (1995) state that the following conditions are
necessary for a trust relationship to evolve: (1) The trustor takes a risk when trusting the trustee. (2) The trustee’s decision benefits the
trustor at a cost to the trustee. (3) Both players are better off with this transaction compared to a situation without the initial trusting
action of the trustor. (4) The trustor must believe that the trustee will interpret her action as a trusting one and that (s)he will
reciprocate accordingly, otherwise the trustor will not take the risk of the trusting action. And (5) the trustee must indeed interpret the
trustor’s play as a trusting action, otherwise the trustee will not be willing to give up some part of her profit to reciprocate (McCabe
et al., 2003).
In our Transfer treatment, the three prerequisites for a trust relationship are met. Substantial efficiency gains can be achieved to the
benefit of both players with transfers and corresponding contributions to the public good game. Nevertheless, the unconditional
transfer bears a substantial risk for type-A players and is thus likely interpreted by the type-B players as a trusting action. Building on
empirical and theoretical evidence of the role of reciprocity in behavior (Rabin, 1993; Falk and Fischbacher, 2000; Dufwenberg and
Kirchsteiger, 2004; Fehr et al., 1997; Fehr and Gächter, 2000), type-B players are likely to reciprocate in the form of higher public good
contributions. Furthermore, higher transfer amounts are likely to be positively correlated with higher public good contributions by
type-B players as they convey a stronger willingness to trust. However, decisions not to transfer despite the potential efficiency gains
may be interpreted as distrust and in turn be reciprocated with low contributions. To summarize, we predict that voluntary transfers are
made (H1). On the side of the receiving type-B players, we predict that contribution levels relative to their available endowment (after a
potential transfer) are higher when a voluntary transfer was received as compared to the relative contributions levels of type-B players in
Baseline (H2a) and lower as in Baseline when a voluntary transfer was available but not made (H2b).
Previous work has highlighted the importance of intentions for reciprocal behavior (e.g., Blount, 1995; Charness, 2004). A study by

13
As an anonymous reviewer correctly noted, the efficiency gain in our case occurs on the way back (i.e., when type-B players contribute to the
public good game which is in regard to the implementation of the multiplication parameter closer to gift exchange games, the trust game’s sibling).
Since trust and reciprocity are also key enablers of efficiency gains (see e.g., Fehr et al., 1998; Fehr and Gächter, 2000), but the gift exchange game is
commonly used in the literature to study behavior in employment relationships, we base our predictions on the more generic reference to trust
games.

6
A.-K. Koessler et al. European Economic Review 159 (2023) 104561

Table 2
Probability of receiving a transfer in t in transfer.
Pr(Receiving transfer in t)

Received transfer in t-1 − 0.435***

(0.101)
Relative contribution in t-1 0.001
(0.001)
Received transfer in t-1 * 0.010***
Relative contribution in t-1 (0.001)
Round − 0.0098***
(0.004)
Constant 0.326***
(0.084)

Observations 720

Note: Estimation is based on a random effects linear probability model. The spec­
ification includes fixed effects for the second wave of data collection. Standard
errors clustered at the group level are in parentheses. Level of significance: * p <
0.10, ** p < 0.05, *** p < 0.01.

Cox (2004) experimentally examines intention-based reciprocity and inspired the set-up of our Exogenous treatment condition. Cox
compared receiver behavior in a standard trust game (treatment A) and a modified trust game, in which the transfer amounts were
determined exogenously (treatment C). The exogenously determined amounts in treatment C, as in our experiment, were the amounts
players had chosen in the endogenous set-up of treatment A. Cox found that with low amounts transferred by the sender, returned
amounts in treatment C exceeded those in treatment A, albeit on a low level. The reverse was true for high amounts. While returned
amounts were large in treatment A, most receivers returned nothing or only a small amount in treatment C. The intentional transfer
was thus a pre-condition for reciprocal behavior. Consequently, we predict that transfers are only reciprocated in terms of higher relative
contribution levels to the public good when the transfer is voluntary and not exogenous (H3).
Building on the above predictions, we hypothesize that overall average group contributions and public good size will reach higher
levels when a voluntary transfer is possible.

5. Results

At first, to assess the average treatment effect, average group contributions and the size of the generated public good are analyzed
for each treatment condition. Thereafter, we examine the realized transfers and type-specific decision-making of the players. At the
end, average profits and earnings are compared.

5.1. Average group contributions and size of the public good

The left panel of Fig. 1 displays the average group contributions over the ten rounds for each treatment condition, measured relative
to the endowment. In the Baseline condition, contribution rates start relatively high, at 65%, before settling at an average of 58% over
the full ten rounds. In the Transfer treatment condition, the initial contribution rate is 71%, and contribution rates are, except for the
last round, stable at this level (average of 70% over the ten rounds). Meanwhile, the contribution rates in the Exogenous treatment
condition, where transfers are realized exogenously, start at 64% in the first round and average 60% over the full ten rounds. When
probing for differences in the average group contributions using a Mann-Whitney Wilcoxon test (MWW), a statistically significant
difference is found between the Transfer and Baseline condition (z = − 2.084, p = 0.037) but not between Exogenous and Baseline (z =
− 0.441, p = 0.663) or Transfer and Exogenous (z = 1.497, p = 0.136).1415 A random effects GLS estimation taking into account the time
trends (see Table A2 in the appendix) confirms this difference between Transfer and Baseline, albeit in a weaker form (p = 0.060).
Result 1a: Average group contribution rates are higher when voluntary transfers are possible as compared to when no transfer option is
given.
When examining the size of the generated public goods instead, the Transfer condition outperforms the other two conditions. The
public good size that is reached in the Transfer groups is, on average, 83 tokens, which is significantly larger than the size of 54 tokens
in the Baseline groups (MWW, z = − 3.758, p = 0.000) and 67 tokens in the Exogenous groups (z = 2.203, p = 0.027).16 The right panel
of Fig. 1 illustrates the resulting public good sizes for all treatments over the ten rounds, while in the appendix demonstrates that these
results hold when time trends are accounted for.
Result 1b: The average size of the public good is significantly larger when voluntary transfers are possible as compared to when no transfer

14
The unit of observation is the group. All tests are two-sided unless stated otherwise.The results presented in the manuscript on basis of Mann-
Whitney-Wilcoxon tests remain when correcting for multiple hypotheses testing, see supplementary analysis in the online appendix.
15
When the last round is excluded from the analysis, we get the following: Transfer vs. Baseline: z = − 2.051, p = 0.040 and Transfer vs. Exogenous:
z = 1.737, p = 0.083.
16
Public good size in Exogenous vs. Baseline: z = − 1.844, p = 0.065.

7
A.-K. Koessler et al. European Economic Review 159 (2023) 104561

Fig. 2. Average marginal effect of having received a transfer in t-1 on the likelihood of receiving a transfer in t for different contribution levels in t-1
with 95% confidence intervals.

Fig. 3. Average relative contributions by player type and treatment condition. The graph depicts the average contributions of different player types
as percentages of their respective endowment after transfers by treatment condition. Type B (Transf) refers to type-B players in rounds when a
transfer was made, and Type B (No transf) refers to type-B players in rounds when no transfer was made.

option is given or when transfers are exogenously imposed.

5.2. Transfer rates

The observed differences in public good sizes suggest that resources within groups are allocated and used more efficiently in the
Transfer condition. In the next step, we analyze the respective transfer rates and corresponding reactions of type-B players. In the
Transfer condition, type-A players frequently and extensively use the transfer option. In the first round, 68% of type-A players transfer
all of their available endowment and 23% transfer half. Over the course of the interaction rounds, the composition of transfers
remained relatively stable (see Fig. A1 in the appendix). Overall, 63% of the transfers are equal to the full endowments and 15% to half
of the endowments. Out of 40 type-A players, only two never made a transfer, while 13 transferred the maximum amount each time.
This provides support for Hypothesis 1.
Result 2: When voluntary transfers are possible, type-A players make use of it and frequently transfer the maximum possible amount.
When type-B players receive a transfer, they contribute in 92% of the cases at least the entrusted (i.e., the transferred amount), and

8
A.-K. Koessler et al. European Economic Review 159 (2023) 104561

in 78% of the cases they contribute everything. The low defection rate is essential for stable transfer patterns. Table 2 shows the
estimation results for the probability of receiving a transfer, while Fig. 2 illustrates the model’s interaction term, demonstrating the
average marginal effect of having received a transfer in t-1 on the likelihood of receiving a transfer in t for different contribution levels
in t-1. As can be seen, for type-B players who received a transfer in the previous round, compared to those who have not, the likelihood
of receiving another transfer in the current round significantly drops from their previously realized contributions levels.
Result 3: The transfer-receiving type-B players almost always contribute at least the transferred amount to the public good. A significant
percentage of type-B players reciprocate the transfer by contributing their entire endowment to the public good.
Given the high rates of transfers and the higher productivity of type-B players, it is not surprising that the size of the realized public
good is significantly larger in Transfer relative to that in Baseline. Here, the comparison with the Exogenous condition is more infor­
mative. Theoretically, groups could reach in Exogenous the same maximum public good size as in Transfer. However, as previously
seen, the realized size is not identical. To understand why this is the case, we examine the individual contribution decisions by player
type and realized transfer option.

5.3. Relative contributions to the public good

Fig. 3 presents, according to player type, the average relative contribution rates compared to the available endowment (after
transfer). The average relative contributions of type-A players are significantly smaller in the Transfer condition (23%) than the average
relative contributions in the Baseline condition (54%) (MWW: z = 3.678, p = 0.000). Pairing this with the high rates of realized
transfers shows that type-A players who want to contribute in Transfer mostly did so via the transfer option. In Exogenous, type-A
players contribute on average 61%, which is significantly more than in Transfer (MWW: z = − 3.635, p = 0.000) but not signifi­
cantly different to type-A players’ contributions in Baseline (z = − 1.085, p = 0.282). To assess differences in relative contributions
depending on whether and which type of transfer has been implemented, we use the Wilcoxon signed-rank test (WSRT).17 When type-A
players chose not to make a transfer in the Transfer condition, contributions remain positive but at a much lower level. The average
contribution rates of type-A players who decide against the transfer are at 13% and are at 32% when they decide to transfer only half of
their endowment (exact WSRT, S = 6.000, p = 0.622).18 In Exogenous, where the same transfers were exogenously implemented,
contribution levels are comparable whether or not an exogenous transfer occurred (62% vs. 61%, exact WSRT: S = − 2.000, p = 0.820).
As a result, we see for type-A players in rounds with and without transfers higher contributions in Exogenous than in Transfer (MWW:
transfer rounds: z = − 2.547, p = 0.01, no transfer rounds: z = − 3.990, p = 0.000).
For the “poorer” type-B players, we find, in line with Hauser et al. (2016), that they already contribute in Baseline with 62%—a
higher proportion of their endowment than that seen for the type-A players. In the Transfer condition, this proportion increases to 75%,
which constitutes a significant difference to the contribution levels observed in Baseline (MWW, z = − 2.072, p = 0.038) and in
Exogenous (z = 2.167, p = 0.030) where type-B players contribute, on average, 60%.19
Distinguishing between rounds with and without transfers provides more information regarding type-B players’ motivations to
contribute at high levels. In Transfer, contribution levels are at 37% in rounds when no transfer from the type-B players is received and
at 81% when a transfer is received (WSRT, z = 3.771, p = 0.000). Compared to Baseline, these contribution levels of type-B players are
significantly lower in the case of no transfer (MWW, z = 3.105, p = 0.002) and significantly higher in the case of transfers (z = − 2.863,
p = 0.004).20 This finding provides support for Hypotheses 2a and 2b.
Result 4a: Receiving a voluntary transfer, type-B players reciprocate this action by contributing higher relative levels to the public good as
compared to when no transfer option is given.
Result 4b: When type-A players choose not to make a transfer, type-B players, in contrast, contribute lower relative levels than when no
transfer option is given.
Focusing on the changes in endowments, we find that in Exogenous, type-B players do not change their contributions with higher
endowments resulting from exogenously implemented transfers (no transfer vs. transfer: 62% vs. 63%, WSRT, z = − 0.726, p =
0.489).21 Thus, for the reaction of type-B players, it is important that their increased endowment occurs as a result of a transfer chosen
by their interaction partner. When making a transfer, type-A players in Transfer show trust, and type-B players reciprocate this trusting
action.

17
The Wilcoxon signed-rank test takes the dependency in the data (i.e. multiple observations from the same person) into account.
18
Only type-A players who did not transfer their full endowment can still make contributions. As outlined in the previous section, type-A players
predominantly chose to transfer their total endowment, meaning that observations here are relatively small. To account for these small sample sizes,
we perform exact Wilcoxon signed-rank rests at this point.
19
In absolute terms, average contributions of type-B players are 9.35 (sd: 4.39) in Baseline, 20.44 (9.08) in Transfer, and 15.47 (sd: 8.19) in
Exogenous. These contribution levels differ among all three treatment conditions at a 1% significance level.
20
Examining the different constellations of realized transfers gives a more detailed insight into the reactions of the type-B players. As shown in
Section 5.2., type-A players send a transfer equal to half of their endowment in 15% of the cases, which is received by only one of the type-B players.
Those receiving type-B players contribute 76% of their (new) endowment, and the non-receivers contribute 59% (WSRT: z = 2.197, p = 0.030). In
63% of the cases, type-A players transfer their full endowment. They do so primarily by splitting the transfer between the two type-B players (68% of
the time). Type-B players respond in this case by contributing on average 83% of their endowment. When the full transfer is instead directed to only
one type-B player, the receiving type-B player contributes 80% and the non-receiver 67% (WSRT: z = 1.691, p = 0.093).
21
As a consequence, contributions by type-B players in Transfer were also higher than the corresponding type-B contributions in Exogenous when a
transfer was made (MWW, z = 2.740, p = 0.006) and lower in the case of no transfer (MWW, z = - 2.593, p = 0.009).

9
A.-K. Koessler et al. European Economic Review 159 (2023) 104561

Comparing in Transfer the sizes of the transfer, we find that higher transfers (i.e., of the full endowment) also lead to higher
contribution rates among type B players in the Transfer condition. This pattern is absent in Exogenous, reiterating the critical role of the
deliberate decision to transfer to motivate the reciprocity of type-B players in the form of higher contributions to the public good. The
regression analysis shown in Table A3 in the appendix provides support for these results. In summary, we find that the positive
(negative) changes in contribution levels of type-B players following type-A players’ decisions (not) to transfer are a response to the
intentional act of the latter and cannot be fully explained by the resulting changes in endowment as implemented in the Exogenous
treatment condition. With this we find support for Hypothesis 3.
Result 5: Changes in the relative contribution levels of Type-B players can be attributed to the intentional transfer of type-A players and
cannot be fully explained by higher endowments resulting from the transfer.

5.4. Profits

Efficiency gains established by transfers are reflected in the resulting profits. Mirroring the earlier result on the size of the public
good, average profits per round in Transfer are significantly higher than in Baseline and Exogenous (39 tokens vs. 30 and 35 tokens,
MWW, z = − 4.586, p = 0.000 and z = 2.433, p = 0.015). Considering the heterogeneity among players, we find that type-A players
benefit from the option of voluntary transfers (35 tokens in Baseline vs. 41 tokens in Transfer, MWW: z = − 3.965, p = 0.000), but lose
out when exogenous transfers are in place (30 tokens in Exogenous, for Baseline vs. Exogenous: z = 1.912, p = 0.056; and for Transfer vs.
Exogenous: z = 4.292, p-values = 0.000). Type-B players, on the other side, benefit from both transfer schemes compared to their
counterparts in Baseline (27 tokens vs. 38 tokens in Transfer (MWW, z = − 5.472, p = 0.000) and vs. 37 tokens in Exogenous (z = − 4.793,
p = 0.000), with no significant difference between the two transfer schemes (MWW, z = 1.221, p = 0.224).
For type-A players in Transfer, earnings in rounds with a transfer are non-significantly higher than in rounds without a transfer (40
vs. 37 tokens, WSRT: z = − 0.784, p = 0.452).22 By contrast, for type-B players, transfers lead to a remarkable difference in profits (in
Transfer: 43 in rounds with a transfer vs. 20 in rounds without, WSRT: z = 3.920, p = 0.000).23 Lastly, type-B players who in Transfer
always contribute at least the transferred amount fare significantly better than type-B players who do not (MWW, z = 3.311, p =
0.001), with 67% belonging to the group of type-B players that contributed at least the transferred amount in all rounds.
Result 6: Average group profits are significantly higher when a transfer option is provided, with voluntary transfers leading to the best result.
Type-A players earn significantly more under the voluntary transfer scheme. Type-B players earn significantly more when a transfer scheme is in
place, with no difference in whether the transfer is voluntary or exogenously imposed.
The results presented here are derived based on a partner-matching protocol in which players interact in the same group over the
course of the interaction rounds and are able to track the actions of their interaction partners and thus build up a reputation over time
(e.g., Bracht and Feltovich, 2009). As previous research has shown, this setting strengthens the role of trust and reciprocity and
consequently boosts cooperation (see Croson (1996) for public good games and Bohnet and Huck (2004) for trust games), which is
partially because violations against the reciprocity norms can be punished in later rounds. As a consequence, we assumed for our
research that endogenous transfers are realized and reciprocated more often than in a setting where individuals interact anonymously.
To test this theory, we conducted the experiment using a smaller sample under a stranger-matching protocol. The detailed analysis can
be found in the online appendix.24 As expected, average group contributions, public good sizes, and average profits are smaller in both
transfer conditions when the stranger-matching protocol was implemented than when the partner-matching protocol was imple­
mented. Moreover, defection is more common when the stranger-matching protocol is used.

5.5. Summary of the results

To summarize, we found that allowing for a resource transfer from the richer to the poorer but more productive players has an
overall positive effect on the provision of public goods. A significant majority of type-A players transfer half or all of their endowment,
although these transfers were unconditional in our experiment (i.e., players risked low or zero earnings). Type-B players, in turn,
commonly reciprocate the trusting action with high contributions and low defection rates. However, in the case of no transfer,
cooperation rates among type-B players are significantly lower than when no transfer option is given. This suggests that the decision
not to transfer is interpreted as distrust and accordingly reciprocated with low contributions. The additional exogenous condition
shows that the pure reallocation of resources can only explain these behavioral patterns to some degree. When transfers are imple­
mented exogenously, type-B players do not reciprocate the transfer with higher contribution levels. Therefore, the intentional decision
to transfer is vital. By performing a voluntary transfer, type-A players signal their trust and willingness to cooperate, which type-B
players in turn reciprocate.

6. Discussion and conclusion

Previous research has shown that, in the presence of free-riding incentives, cooperation can be difficult to achieve, particularly in

22
In Exogenous, round earnings of type-A players in rounds with a transfer are non-significantly lower than in rounds without (WRST, 30 vs. 34, z
= − 1.167, p = 0.254).
23
For Exogenous: 40 vs. 28 WSRT, z = 3.823, p = 0.000.
24
https://osf.io/9n7d3/?view_only=208c2ea1e96c42b795ea0df507e294d9

10
A.-K. Koessler et al. European Economic Review 159 (2023) 104561

situations involving heterogeneous actors, which is often the case outside the laboratory. In the case of an asymmetric distribution of
endowment and productivity heterogeneity, meaning that the “poor” are more productive than the “rich”, our results show that the
simple mechanism of allowing a transfer of resources can significantly increase public good provision. These transfers in our exper­
iment are purely voluntary and do not require any exogenous intervention. Successful groups use the transfers to exploit efficiency
gains and to overcome conflicting contribution norms by establishing a positive reciprocal relationship. However, not all groups
succeed in cooperation.
To be more specific, we find that a surprisingly large share of players transfer their entire endowment, even in the first round of
play. Numerous experiments have shown that humans exhibit a “taste” for efficiency. For instance, in a study by Kritikos and Bolle
(2001), individuals accept a more unequal distribution of payoffs to their disadvantage in order to achieve a more efficient outcome
overall. Galeotti et al. (2019) show that players are more concerned with efficiency than equity in bargaining games. Barrett and
Dannenberg (2017) demonstrate that subjects prefer the more efficient game (in terms of higher maximum payoff possible) over the
game with lower payoffs but higher chances of realizing these lower payoffs. This concern for efficiency can explain why type-A players
in our experiment are willing to risk their endowment for the opportunity to achieve the most efficient solution.
Second, type-B players with low endowment but high productivity usually reciprocate transfers with very high contribution rates,
even when the other B-player was the recipient of the transfer. The comparison with the exogenous treatments shows that these high
cooperation rates are only realized when an endogenous transfer occurs. When transfers are exogenously imposed, we do not find that
type-B players respond to the transfer in the same way. This difference cannot be explained by standard economic preferences or
inequality aversion, but the behavior fits relatively well the predictions based on intention-based reciprocity. The positive correlation
between higher transfers and higher subsequent contributions in the transfer treatments also corresponds to the predictions based on
reciprocity concerns.
Third, if the option of a transfer is given but not used, this seemed to be perceived by the players as a signal of non-cooperation or
unwillingness to trust and was reciprocated accordingly. It is therefore important to investigate the conditions under which a transfer is
more likely to occur.
Fourth, compared to the asymmetric setup adopted by Hauser et al. (2019), the contribution level in our baseline treatment is
considerably higher: 63% compared to 46% in Hauser et al. (2019). There were several differences between the two experiments:
Subjects in Hauser et al. (2019) were recruited via Mturk and play online, they play in groups of two players with a productivity
parameter of 1.3 and 1.9, and the game lasts at least 20 rounds with an indefinite ending thereafter. While the first difference most
likely has a negative effect on contributions, smaller group sizes and indefinite endings usually have a positive impact. Meanwhile, the
effect of the difference in the productivity parameter is less clear as it differs from our setup in terms of its level as well as in its
skewness.
The results of this paper are good news for heterogeneous groups in situations with repeated interactions and observability of
behavior. By simply allowing group members to enter into a trusting and reciprocal relationship and to exploit efficiency gains,
cooperation rates and thereby public good provision can be increased significantly.
While the use of a laboratory experiment enabled us to identify causal effects, the simplification and level of abstraction that came
with it impose a number of limitations. Essential for us are the following: Transfers outside the laboratory are typically not given
unconditionally. In our experiment the players with high endowment but low productivity (type-A) are able to observe other players’
behavior but are unable to tie their transfers to conditions or to punish in case recipient’s behavior is not as intended. While such an
enforcement mechanism may be important, in reality it is often difficult or even impossible for the donor to determine whether funds
are actually used as intended (see Devarajan and Swaroop (2000) for a discussion on the fungibility of development aid). Conditions or
sanctions, though declared ex-ante, may thus in reality not be enforceable.
Furthermore, our experiment allows for substantial efficiency gains through transfers. With the high productivity parameter, type-
B players have to contribute only 50% or more of the received transfer amount so that the transfer pays off. Additionally, the high
productivity parameter reduces the cost of contributing to the public good and thus facilitates reciprocation. A valuable investigation
for future research may be to examine how sensitive the observed reciprocity is to variations in the productivity heterogeneity.
Lastly, a feature that is usually present in real-world situations, such as teamwork settings or in the international community, but is
absent in our experiment is communication. Communication has been shown to increase cooperation in homogenous groups (Ostrom
et al., 1992; Balliet, 2010), while its impact has been found to be more ambiguous in heterogeneous groups (Hackett et al., 1994; Chan
et al., 1999). Thus, it is a priori not clear how communication would interact with the transfer possibility. On the one hand, the ability
to communicate could strengthen the group feeling and increase the willingness to trust, thereby increasing the number of positive
transfers and subsequent contributions. On the other hand, with prior bargaining the transfers could be interpreted more as a strategic
action than a trusting one, which could increase the likelihood of non-reciprocal behavior accordingly. Hence, incorporating
communication could provide an interesting starting point for an extension of this experiment.

Declaration of Competing Interest

None.

Data availability

The data, analysis files and experimental instructions are publicly available via the Center for Open Science: https://osf.io/9n7d3/?
view_only=208c2ea1e96c42b795ea0df507e294d9

11
A.-K. Koessler et al. European Economic Review 159 (2023) 104561

Acknowledgements

This work was financially supported by the European Union (EU) Horizon 2020 program, action ERC-2014-STG, Project HUCO,
Grant Number 636746. The experiment was approved by the Institutional Review Board of the German Association for Experimental
Economic Research (https://gfew.de/ethik/qpAfJtwu). We thank Astrid Dannenberg, the editor David K. Levine and three anonymous
reviewers for their valuable comments. We thank G. Simmonett for the proofreading of the manuscript.

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.euroecorev.2023.104561.

References

Andreoni, J., Varian, H., 1999. Preplay contracting in the prisoners’ dilemma. In: Proceedings of the National Academy of Sciences of the United States of America, 96,
pp. 10933–10938. https://doi.org/10.1073/pnas.96.19.10933.
Balliet, D., 2010. Communication and cooperation in social dilemmas: a meta-analytic review. J. Confl. Resolut. 54 (1), 39–57. https://doi.org/10.1177/
0022002709352443.
Barrett, S., Dannenberg, A., 2017. Tipping versus cooperating to supply a public good. J. Eur. Econ. Assoc. 15 (4), 910–941. https://doi.org/10.1093/jeea/jvw022.
Berg, J., Dickhaut, J., McCabe, K., 1995. Trust, reciprocity, and social history. Games Econ. Behav. 10 (1), 122–142. https://doi.org/10.1006/game.1995.1027.
Blanco, E., Haller, T., Walker, J.M., 2018. Provision of environmental public goods: unconditional and conditional donations from outsiders: unconditional and
conditional donations from outsiders. J. Environ. Econ. Manage. 92, 815–831. https://doi.org/10.1016/j.jeem.2017.10.002.
Blount, S., 1995. When social outcomes aren′ t fair: the effect of causal attributions on preferences. Organ. Behav. Hum. Decis. Process. 63 (2), 131–144.
Bock, O., Baetge, I., Nicklisch, A., 2014. hroot: Hamburg registration and organization online tool. Eur. Econ. Rev. 71, 117–120. https://doi.org/10.1016/j.
euroecorev.2014.07.003.
Bohnet, I., Huck, S., 2004. Repetition and reputation: implications for trust and trustworthiness when institutions change. Am. Econ. Rev. 94 (2), 362–366. https://
doi.org/10.1257/0002828041301506.
Bolton, G., Ockenfels, A., 2000. ERC: A Theory of Equity, Reciprocity, and Competition. Am. Econ. Rev. 90 (1), 166–193. https://doi.org/10.1257/aer.90.1.166.
Boucher, D., Roquemore, S., Fitzhugh, E., 2013. Brazil’s success in reducing deforestation. Trop. Conserv. Sci. 6 (3), 426–445. https://doi.org/10.1177/
194008291300600308.
Bracht, J., Feltovich, N., 2009. Whatever you say, your reputation precedes you: observation and cheap talk in the trust game. J. Public Econ. 93 (9–10), 1036–1044.
https://doi.org/10.1016/j.jpubeco.2009.06.004.
Bruttel, L.V., Güth, W., 2018. Asymmetric voluntary cooperation: a repeated sequential best-shot experiment. Int. J. Game Theory 47 (3), 873–891. https://doi.org/
10.1007/s00182-018-0633-y.
Bruttel, L., Güth, W., Nithammer, J., Orland, A., 2022. Inefficient cooperation under stochastic and strategic uncertainty. J. Confl. Resolut. 66 (4–5), 755–782. https://
doi.org/10.1177/00220027211066614.
Buckley, E., Croson, R.T.A., 2006. Income and wealth heterogeneity in the voluntary provision of linear public goods. J. Public Econ. 90 (4–5), 935–955. https://doi.
org/10.1016/j.jpubeco.2005.06.002.
Chan, K.S., Mestelman, S., Moir, R., Muller, R.A., 1999. Heterogeneity and the voluntary provision of public goods. Exp. Econ. 2 (1), 5–30. https://doi.org/10.1023/A:
1009984414401.
Charness, G., Fréchette, G.R., Qin, C.Z., 2007. Endogenous transfers in the prisoner’s dilemma game: an experimental test of cooperation and coordination. Games
Econ. Behav. 60 (2), 287–306. https://doi.org/10.1016/j.geb.2006.10.010.
Charness, G., 2004. Attribution and reciprocity in an experimental labor market. J. Labor Econ. 22 (3), 665–688. https://doi.org/10.1086/383111.
Chatziathanasiou, K., Hippel, S., Kurschilgen, M., 2021. Property, redistribution, and the status quo: a laboratory study. Exp. Econ. 24, 919–951. https://doi.org/
10.1007/s10683-020-09685-5.
Chatziathanasiou, K., Hippel, S., Kurschilgen, M., 2022. Does the threat of overthrow discipline the elites? Evidence from a laboratory experiment. J. Legal. Stud. 51
(2), 289–320.
Cherry, T.L., Kroll, S., Shogren, J.F., 2005. The impact of endowment heterogeneity and origin on public good contributions: evidence from the lab. J. Econ. Behav.
Organ. 57 (3), 357–365. https://doi.org/10.1016/j.jebo.2003.11.010.
Chaudhuri, A., 2011. Sustaining cooperation in laboratory public goods experiments: a selective survey of the literature. Exp. Econ. 14 (1), 47–83. https://doi.org/
10.1007/s10683-010-9257-1.
Cox, J.C., 2004. How to identify trust and reciprocity. Games Econ. Behav. 46 (2), 260–281. https://doi.org/10.1016/S0899-8256(03)00119-2.
Croson, R.T., 1996. Partners and strangers revisited. Economics Letters 53 (1), 25–32. https://doi.org/10.1016/S0165-1765(97)82136-2.
Croson, R.T.A., Marks, M., 2001. The effect of recommended contributions in the voluntary provision of public goods. Econ. Inq. 39 (2), 238–249. 10.1111/j.1465-
7295.2001.tb00063.x.
de Oliveira, A.C.M., Croson, R.T.A., Eckel, C., 2015. One bad apple?: heterogeneity and information in public good provision. Exp. Econ. 18 (1), 116–135. https://doi.
org/10.1007/s10683-014-9412-1.
Devarajan, S., Swaroop, V., 2000. The implications of foreign aid fungibility for development assistance. In: Gilbert, C.L., Vines, D. (Eds.), The World Bank. Structure
and Policies. Cambridge Univ. Press, Cambridge, pp. 196–209.
Dufwenberg, M., Kirchsteiger, G., 2004. A theory of sequential reciprocity. Games Econ. Behav. 47 (2), 268–298. https://doi.org/10.1016/j.geb.2003.06.003.
Falk, A., Fischbacher, U., 2000. A Theory of Reciprocity. Institute for Empirical Research in Economics Working Paper Series 6. University of Zurich.
Fehr, E., Gächter, S., Kirchsteiger, G., 1997. Reciprocity as a contract enforcement device. Econometrica 65, 833–860. https://doi.org/10.2307/2171941, 1997.
Fehr, E., Fischbacher, U., 2004. Social norms and human cooperation. Trends Cogn. Sci. (Regul. Ed.) 8 (4), 185–190. https://doi.org/10.1016/j.tics.2004.02.007.
Fehr, E., Gächter, S., 2000. Fairness and retaliation: the economics of reciprocity. J. Econ. Perspect. 14 (3), 159–181. https://doi.org/10.1257/jep.14.3.159.
Fellner, G., Iida, Y., Kröger, S., Seki, E., 2011. Heterogeneous productivity in voluntary public good provision: an experimental analysis. IZA Discussion Paper.
Fischbacher, U., 2007. z-Tree: Zurich toolbox for ready-made economic experiments. Exp. Econ. 10 (2), 171–178. https://doi.org/10.1007/s10683-006-9159-4.
Fischbacher, U., Gächter, S., Fehr, E., 2001. Are people conditionally cooperative? Evidence from a public goods experiment. Econ. Lett. 71 (3), 397–404. https://doi.
org/10.1016/S0165-1765(01)00394-9.
Fisher, J., Isaac, R.M., Schatzberg, J.W., Walker, J.M., 1995. Heterogenous demand for public goods: behavior in the voluntary contributions mechanism. Public
Choice 85 (3–4), 249–266. https://doi.org/10.1007/BF01048198.
Fung, J.M., Au, W., 2014. Effect of inequality on cooperation: heterogeneity and hegemony in public goods dilemma. Organ. Behav. Hum. Decis. Process. 123 (1),
9–22. https://doi.org/10.1016/j.obhdp.2013.10.010.
Galeotti, F., Montero, M., Poulsen, A., 2019. Efficiency versus equality in bargaining. J. Eur. Econ. Assoc. 17 (6), 1941–1970. https://doi.org/10.1093/jeea/jvy030.

12
A.-K. Koessler et al. European Economic Review 159 (2023) 104561

Güth, W., Levati, M.V., Sutter, M., van der Heijden, E., 2007. Leading by example with and without exclusion power in voluntary contribution experiments. J. Public
Econ. 91 (5–6), 1023–1042. 10.1016/j.jpubeco.2006.10.007.
Hackett, S., Schlager, E., Walker, J.M., 1994. The role of communication in resolving commons dilemmas: experimental evidence with heterogeneous appropriators.
J. Environ. Econ. Manage. 27 (2), 99–126. 10.1006/jeem.1994.1029.
Hargreaves Heap, S.P., Ramalingam, A., Stoddard, B.v., 2016. Endowment inequality in public goods games: a re-examination. Econ. Lett. 146, 4–7. https://doi.org/
10.1016/j.econlet.2016.07.015.
Hauser, O.P., Hilbe, C., Chatterjee, K., Nowak, M.A., 2019. Social dilemmas among unequals. Nature 572 (7770), 524–527. https://doi.org/10.1038/s41586-019-
1488-5.
Hofmeyr, A., Burns, J., Visser, M., 2007. Income inequality, reciprocity and public good provision: an experimental analysis. South Afr. J. Econ. 75 (3), 508–520.
https://doi.org/10.1111/j.1813-6982.2007.00127.x.
Kesternich, M., Lange, A., Sturm, B., 2014. The impact of burden sharing rules on the voluntary provision of public goods. J. Econ. Behav. Organ. 105, 107–123.
https://doi.org/10.1016/j.jebo.2014.04.024.
Kimbrough, E.O., Sheremeta, R.M., 2013. Side-payments and the costs of conflict. Int. J. Ind. Organ. 31 (3), 278–286. https://doi.org/10.1016/j.
ijindorg.2012.01.005.
Kingsley, D.C., 2016. Endowment heterogeneity and peer punishment in a public good experiment: cooperation and normative conflict. J. Behav. Exp. Econ. 60,
49–61. https://doi.org/10.1016/j.socec.2015.12.002.
Koessler, A.K., Ortiz-Riomalo, J.F., Janke, M., et al., 2021. Structuring Communication Effectively—The Causal Effects of Communication Elements on Cooperation in
Social Dilemmas. Environ. Resource Econ. 79, 683–712. https://doi.org/10.1007/s10640-021-00552-2.
Kölle, F., 2015. Heterogeneity and cooperation: the role of capability and valuation on public goods provision. J. Econ. Behav. Organ. 109, 120–134. https://doi.org/
10.1016/j.jebo.2014.11.009.
Kölle, F., Sliwka, D., Zhou, N., 2016. Heterogeneity, inequity aversion, and group performance. Soc. Choice Welfare 46 (2), 263–286. https://doi.org/10.1007/
s00355-015-0912-5.
Kritikos, A., Bolle, F., 2001. Distributional concerns: equity- or efficiency-oriented? Econ. Lett. 73 (3), 333–338. https://doi.org/10.1016/S0165-1765(01)00503-1.
Levati, M.V., Sutter, M., van der Heijden, E., 2007. Leading by example in a public goods experiment with heterogeneity and incomplete information. J. Confl.
Resolut. 51 (5), 793–818. https://doi.org/10.1177/0022002707302796.
McCabe, K.A., Rassenti, S.J., Smith, V.L., 1998. Reciprocity, trust, and payoff privacy in extensive form bargaining. Games Econ. Behav. 24 (1–2), 10–24. https://doi.
org/10.1006/game.1998.0638.
McCabe, K.A., Rigdon, M.L., Smith, V.L., 2003. Positive reciprocity and intentions in trust games. J. Econ. Behav. Organ. 52 (2), 267–275. https://doi.org/10.1016/
S0167-2681(03)00003-9.
Ministry of Climate and Environment, 2018. Brazil. Norwegian Government. https://www.regjeringen.no/en/topics/climate-and-environment/climate/climate-and-
forest-initiative/kos-innsikt/brazil-and-the-amazon-fund/id734166/?expand=factbox2621335. Accessed 16 November 2020.
Nikiforakis, N., Noussair, C.N., Wilkening, T., 2012. Normative conflict and feuds: the limits of self-enforcement. J. Public Econ. 96 (9–10), 797–807. https://doi.org/
10.1016/j.jpubeco.2012.05.014.
Norway’s International Climate and Forest Initiative, 2020. Partner Countries: Brazil. https://www.nicfi.no/partner-countries/brazil/. Accessed 16 November 2020.
Noussair, C.N., Tan, F., 2011. Voting on punishment systems within a heterogeneous group. J. Public Econ. Theory 13 (5), 661–693. https://doi.org/10.1111/j.1467-
9779.2011.01516.x.
OECD, 2009. The Economics of Climate Change Mitigation. OECD Publishing, Paris.
Ostrom, E., Walker, J.M., Gardner, R., 1992. Covenants with and without a sword: self-governance is possible. Am. Polit. Sci. Rev. 86 (2), 404. https://doi.org/
10.2307/1964229.
Rabin, M., 1993. Incorporating fairness into game theory and economics. Am. Econ. Rev. 83 (5), 1281–1302.
Reuben, E., Riedl, A., 2009. Public goods provision and sanctioning in privileged groups. J. Confl. Resolut. 53 (1), 72–93. https://doi.org/10.1177/
0022002708322361.
Reuben, E., Riedl, A., 2013. Enforcement of contribution norms in public good games with heterogeneous populations. Games Econ. Behav. 77 (1), 122–137. https://
doi.org/10.1016/j.geb.2012.10.001.
Tan, F., 2008. Punishment in a linear public good game with productivity heterogeneity. Economist (Leiden) 156 (3), 269–293. https://doi.org/10.1007/s10645-008-
9094-1.
van Dijk, E., Wilke, H., 1994. Asymmetry of wealth and public good provision. Soc. Psychol. Q. 57 (4), 352. https://doi.org/10.2307/2787161.
Zelmer, J., 2003. Linear public goods experiments: a meta-analysis. Exp. Econ. 6 (3), 299–310. https://doi.org/10.1023/A:1026277420119.

13