Normative conflict and the gender gap in cooperation

Full text

Normative conict and
the gender gap in cooperation
Nikos Nikiforakis
Center for Behavioral Institutional Design &
Division of Social Science, New York University Abu Dhabi
e-mail: nikos.nikiforakis@nyu.edu
Ernesto Reuben
Center for Behavioral Institutional Design &
Division of Social Science, New York University Abu Dhabi
e-mail: ereuben@nyu.edu
Robert Stüber
Center for Behavioral Institutional Design,
New York University Abu Dhabi
e-mail: robert.stueber@nyu.edu
Abstract
A normative conict arises when individuals derive dierent benets from cooperation. We ana-
lyze experimental data from three published studies to investigate the impact of normative con-
ict on the cooperative behavior of men and women. We nd that women exhibit signicantly
lower levels of cooperation in the presence of normative conict. We observe no signicant
gender dierences in cooperation in the absence of normative conict.
This version: March 2024
JEL Classication: C92, D70, H41, J16
Keywords: Cooperation, gender, inequality, normative conict, heterogeneous groups
Acknowledgements: The authors gratefully acknowledge nancial support from Tamkeen under the NYU Abu Dhabi Re-
search Institute Award CG005. This is the authors’ version of work that was accepted for publication in Economics Letters.
Changes resulting from the publishing process may not be reected in this document. A nal version is published in
https://doi.org/10.1016/j.econlet.2024.111680.

1. Introduction
Normative conict arises in situations in which multiple plausible rules coexist about how one ought
to behave. A normative conict between equality and equity emerges, for example, when past
choices or investments aect the returns that individuals derive from cooperation. Equality pre-
scribes that those receiving higher returns from cooperation contribute suciently more than oth-
ers so that nal payos are equalized. Equity, on the other hand, stipulates that higher returns
should be associated with higher (and hence unequal) payos. Despite the frequency with which
normative conict is encountered in daily life, little is known about how it aects behavior.
Normative conict between equality and equity has been shown to reduce cooperation and also
to lead to cycles of retaliation (Nikiforakis et al., 2012; Gangadharan et al., 2017; Koch et al., 2021).
In this paper, we explore how normative conict aects the relative willingness of men and women
to cooperate. Evidence suggests that women tend to exhibit a stronger preference for equality,
whereas men can be more tolerant of inequality when it results from past choices (Andreoni and
Vesterlund, 2001; Almås et al., 2020). If women care more about equality than men, we would
expect that women receiving low (high) returns from cooperation will contribute less (more) than
men when equality and equity concerns are at odds.1
To study the impact of normative conict on the cooperative behavior of men and women, we
analyze data from three published papers that employ similar experimental protocols and manip-
ulate the presence of conict between equality and equity (Nikiforakis et al., 2012; Gangadharan
et al., 2017; Koch et al., 2021). The experiments were conducted in experimental laboratories across
three continents.2
2. The experiments
In the three experimental studies we consider, a total of 412 individuals rst participate in a real-
eort task taken from Erkal et al. (2011). Subsequently, they play a nitely-repeated linear public
goods game in xed groups of four or six individuals. In each period, all participants receive the
same endowment and must decide how much of it to contribute to a public account. In all treat-
1When individuals receive the same returns from cooperation, the evidence suggests that there are no gender dierences
in cooperation (see, e.g., Thöni et al., 2021; Exley et al., 2022).
2To the best of our knowledge, these are the only studies on cooperation that specically investigate the interplay be-
tween equality and equity. Other studies have explored the impact of heterogeneity in cooperation games (e.g., Nous-
sair and Tan, 2011; Reuben and Riedl, 2013; Fischbacher et al., 2014). However, there exists no tension between equity
and equality in these studies, as individual choices do not aect the allocation of returns.
1

ments, individuals have a pecuniary incentive not to contribute to the public account, but contri-
butions increase the earnings of the other group members. The presence of normative conict
depends on the returns that each group member obtains from the public account. Specically, in
treatments with normative conict, participants are told that the two best-performing group mem-
bers in the real-eort task will receive a higher rate of return from the public account. To prevent
other-regarding preferences from driving selection into high returns, participants are not provided
details about the public goods game when performing the real-eort task (see Erkal et al., 2011).3
In treatments without normative conict, all participants receive the same returns from the public
account but are still required to complete a xed number of exercises in the real-eort task. The
average rate of return across group members is held constant in treatments with and without nor-
mative conict.4
To evaluate the impact of normative conict on the willingness of men and women to cooperate,
we compare contributions in treatments with and without normative conict. In all treatments, joint
maximal contributions to the public account maximize group earnings. However, in treatments with
normative conict, joint maximal contributions also maximize pay inequality between group mem-
bers.5These treatments include the Asymmetric and the No Feud treatments in Nikiforakis et al.
(2012), the Het-NC treatment in Gangadharan et al. (2017), and treatment ComLate in Koch et al.
(2021). If women care more about equality than men, we would expect that those who receive a
low return from the public account will contribute less than their male counterparts in these treat-
ments, whereas those who receive a high return will contribute more than their male counterparts.6
By contrast, we predict no gender dierences in contributions in treatments where participants ob-
tain the same returns from the public account. These treatments include treatment Symmetric in
Nikiforakis et al. (2012) and treatments Hom-NC and Hom-C in Gangadharan et al. (2017).
3The fraction of men assigned the role of high return is similar to that of women in all three studies: (i) Nikiforakis et al.
(2012): 51% men, 𝑝 = 1.000; (ii) Gangadharan et al. (2017): 51% men, 𝑝 = 1.000; (iii) Koch et al. (2021): 55% men,
𝑝 = 0.646 (Fisher’s exact tests).
4As an example, in Nikiforakis et al. (2012), earnings at the end of the contribution stage were given by: 𝜋𝑖= 20−𝑐𝑖+ 𝑚𝑖×
Σ4
𝑗=1𝑐𝑗, with 𝑚𝑖= 0.4 under homogeneous returns, and 𝑚𝑖= {0.3, 0.5} under normative conict, respectively. Similar
incentives were provided in Gangadharan et al. (2017) and Koch et al. (2021).
5For instance, in Nikiforakis et al. (2012), if all group members contribute maximally, everyone earns 𝜋𝑖= 32 in the
absence of normative conict. By contrast, with normative conict, high-return individuals earn 𝜋𝑖= 40, and low-return
individuals earn 𝜋𝑖= 24. In all treatments, if nobody contributes, everyone earns 𝜋𝑖= 20.
6More specically, let 𝑐𝑖= 𝔼[ 1
3Σ4
𝑗=1,𝑗≠𝑖𝑐𝑗]denote the average contribution that 𝑖expects of her group members. Then, if i
receives a low return from the public account and cares suciently strongly about inequality, she will contribute 𝑐𝑖< 𝑐,
if 0 < 𝑐 ≤ 20. If, on the other hand, she receives a high return, she will contribute 𝑐𝑖> 𝑐 as doing so reduces pay
inequality, at least if the other high-return player does the same.
2

a. Normative conict b. No normative conict
c. Normative conict and low return d. Normative conict and high return
Figure 1. Contributions to the public account by gender
Note: Error bars depict 95% condence intervals. Participants in Gangadharan et al. (2017) interact for 20 rounds. For
illustration purposes, we drop every second period. All data is used in the statistical analysis.
3. Results
Figure 1 presents the evolution of contributions for men and women in the presence and absence
of normative conict. As seen in Panel A, on average, men contribute 2.21 ECU (33%) more than
women in the presence of normative conict. The dierence is noticeable in all periods but the
last. In contrast, as seen in Panel B, in the absence of normative conict, there are no noticeable
gender dierences in contributions. The latter is in line with previous ndings in cooperation from
homogeneous groups (e.g., Thöni et al., 2021; Exley et al., 2022). In Figure A2 in the Appendix, we
show that the same pattern is observed in each of the three studies.
To obtain statistical support for these ndings, Table 1 presents the results from a regression
analysis. The dependent variable is an individual’s contribution in a given round. The independent
variables include dummy variables for being female, being exposed to normative conict, and the
interaction of the two variables. The regressions also include a variable to control for time eects
3

Table 1. Gender dierences in contributions with and without normative conict
I II III IV V VI
Female 1.501 1.501 0.046 −0.051 −0.051 −0.052
(0.976) (1.113) (0.635) (0.254) (0.244) (0.242)
Normative conict −3.217∗∗ −3.217∗−4.479∗∗∗ −4.809∗∗∗ −5.017∗∗∗ −5.310∗∗
(1.486) (0.986) (1.341) (1.326) (1.942) (2.306)
Female×Normative conict −3.876∗∗∗ −3.876∗∗ −1.428∗−1.115∗∗∗ −0.639 −1.598∗∗∗
(1.234) (0.893) (0.816) (0.349) (0.434) (0.454)
Observations 5,560 5,560 5,560 5,560 2,780 2,780
Controls for return rate ✓✓✓✓✓✓
Clustered standard errors: subjects ✓
Clustered standard errors: groups ✓ ✓
Clustered standard errors: studies ✓
Subject random eects ✓
Group random eects ✓✓✓✓
Study random eects ✓✓✓
Note: Regressions of individual contributions on indicator variables for gender, being in a treatment with normative
conict, and their interaction. The regressions also include controls for the rate of return and the period of the game
(not reported). Columns I to IV present full sample estimates. Columns V and VI estimate the model separately for
low-return and high-return players, respectively ∗𝑝 < 0.10,∗∗ 𝑝 < 0.05,∗∗∗ 𝑝 < 0.01.
and a dummy indicating whether the individual received high returns from contributing to the pub-
lic account. To assess the robustness of our claims, in columns I to IV, we present estimates of our
model under dierent assumptions. Columns V and VI estimate the model separately for low-return
players, and high-return players, respectively.
The coecient for Female is statistically insignicant in columns I to VI in Table 1. This signies
that, in the absence of normative conict, there is no statistically signicant gender dierence in
contributions. The coecient for Normative conict is always negative and statistically signicant,
indicating that men and women contribute less in the presence of normative conict. The interac-
tion term Female×Normative conict in columns I to IV indicates that women decrease their contri-
butions signicantly more in the presence of normative conict than men, resulting in a statistically
and economically signicant gender dierence.
Interestingly, the coecients in columns V and VI in Table 1 reveal that the gender dierence
in contributions is driven by high- and not low-return women. This pattern, which can be seen
clearly in Panels C and D in Figure 1 and is present in each of the three studies (see Figure A1 in
the Appendix), is not in line with the hypothesis that women care more about equality than men.
If this were the case, women receiving low (high) returns would contribute less (more) than their
male counterparts. This is clearly not the case.
Evidence that women do not care more about equality than men can also be found in the post-
experiment questionnaire in Gangadharan et al. (2017). The authors had participants express their
4

agreement to the following two statements, using a seven-point scale: (i) “Ideally, all group mem-
bers should earn more or less the same from the experiment irrespective of whether their returns
are low or high”; (ii) “Ideally, all group members should allocate the same amount to the public
account irrespective of whether they have low or high returns and thus earn dierent amounts.”
Agreement with the rst statement indicates that an individual favors pay equality over equity,
whereas agreement with the second sentence indicates one’s support for equity. We do not nd
evidence of a gender dierence in responses for either statement (𝑝 = 0.221 and 𝑝 = 0.969, re-
spectively, Mann-Whitney U tests).
4. Discussion
We have presented robust evidence from three experimental studies conducted in three dierent
continents over a decade, showing that normative conict can cause gender dierences in cooper-
ation to emerge. This nding shows that, when it comes to gender dierences in cooperation, one
must take into account the specic context: gender dierences may not emerge in homogeneous
groups but are more likely to emerge in heterogeneous groups where there is normative conict.
The underlying mechanism behind the gender gap in contributions remains unclear. The ob-
served patterns do not align with the assumption that women care more about equality than men
or that men care more about equity than women. If this were the case, we would anticipate seeing
lower contributions by women among low-return players and lower contributions by men among
high-return players, which is not what we observe.
One explanation, supported by the data, is that strategic motivations prompt men to contribute
more. Men were found to make signicantly higher contributions at the outset of the experiments,
yet a marked decline in their contributions was observed in the nal round of all three studies.
Consequently, there is no discernible gender disparity in contributions during the last round of
the experiments (see Figure 1). Dierences in strategic motives between men and women could
also account for the fact that the gender gap is observed only among high-return players; it is they
who stand to benet most from higher levels of cooperation.7Identifying the precise mechanism
behind the gender gap in cooperation in the presence of normative conict is an interesting topic
for future research.
7The evidence on the relative strategic sophistication of men and women is mixed, with some studies nding men exhibit
greater strategic sophistication (Cubel and Sanchez-Pages, 2022; Gauriot et al., 2023) and others nding no dierence
(Brañas-Garza et al., 2012; Burnham et al., 2009).
5

References
Almås, I., Cappelen, A. W., and Tungodden, B. (2020). Cutthroat capitalism versus cuddly socialism: Are
americans more meritocratic and eciency-seeking than scandinavians? Journal of Political Economy,
128(5):1753–1788.
Andreoni, J. and Vesterlund, L. (2001). Which is the fair sex? Gender dierences in altruism. The Quarterly
Journal of Economics, 116(1):293–312.
Brañas-Garza, P., García-Muñoz, T., and González, R. H. (2012). Cognitive eort in the beauty contest game.
Journal of Economic Behavior & Organization, 83(2):254–260.
Burnham, T. C., Cesarini, D., Johannesson, M., Lichtenstein, P., and Wallace, B. (2009). Higher cognitive abil-
ity is associated with lower entries in a p-beauty contest. Journal of Economic Behavior & Organization,
72(1):171–175.
Cubel, M. and Sanchez-Pages, S. (2022). Gender dierences in equilibrium play and strategic sophistication
variability. Journal of Economic Behavior & Organization, 194:287–299.
Erkal, N., Gangadharan, L., and Nikiforakis, N. (2011). Relative earnings and giving in a real-eort experiment.
American Economic Review, 101(7):3330–3348.
Exley, C. L., Hauser, O. P., Moore, M., Pezzuto, J.-H., et al. (2022). Beliefs about gender dierences in social
preferences. University of Exeter Department of Economics Discussion Papers 22/04, University of Exeter.
Fischbacher, U., Schudy, S., and Teyssier, S. (2014). Heterogeneous reactions to heterogeneity in returns from
public goods. Social Choice and Welfare, 43(1):195–217.
Gangadharan, L., Nikiforakis, N., and Villeval, M. C. (2017). Normative conict and the limits of self-
governance in heterogeneous populations. European Economic Review, 100:143–156.
Gauriot, R., Page, L., and Wooders, J. (2023). Expertise, gender, and equilibrium play. Quantitative Economics,
14(3):981–1020.
Koch, C., Nikiforakis, N., and Noussair, C. N. (2021). Covenants before the swords: The limits to ecient
cooperation in heterogeneous groups. Journal of Economic Behavior & Organization, 188:307–321.
Nikiforakis, N., Noussair, C. N., and Wilkening, T. (2012). Normative conict and feuds: The limits of self-
enforcement. Journal of Public Economics, 96(9-10):797–807.
Noussair, C. N. and Tan, F. (2011). Voting on punishment systems within a heterogeneous group. Journal of
Public Economic Theory, 13(5):661–693.
Reuben, E. and Riedl, A. (2013). Enforcement of contribution norms in public good games with heterogeneous
populations. Games and Economic Behavior, 77(1):122–137.
Thöni, C., Volk, S., and Cortina, J. M. (2021). Greater male variability in cooperation: Meta-analytic evidence
for an evolutionary perspective. Psychological Science, 32(1):50–63.
6

A. Online Appendix: Additional gures
0 2 4 6 8 10 12 14 16 18 20
Average contribution
1 2 3 4 5 6 7 8 9 10
Period
Female
Male
a. Low return (NNW2012)
0 2 4 6 8 10 12 14 16 18 20
Average contribution
1 2 3 4 5 6 7 8 9 10
Period
Female
Male
b. High return (NNW2012)
0 2 4 6 8 10 12 14 16 18 20
Average contribution
1 2 3 4 5 6 7 8 9 10
Period
Female
Male
c. Low return (GNV2017)
0 2 4 6 8 10 12 14 16 18 20
Average contribution
1 2 3 4 5 6 7 8 9 10
Period
Female
Male
d. High return (GNV2017)
0 2 4 6 8 10 12 14 16 18 20
Average contribution
1 2 3 4 5 6 7 8 9 10
Period
Female
Male
e. Low return (KNN2021)
0 2 4 6 8 10 12 14 16 18 20
Average contribution
1 2 3 4 5 6 7 8 9 10
Period
Female
Male
f. High return (KNN2021)
Figure A1. Contributions to the public account by gender, return, and study
Note: The gure presents contributions separately by gender, rate of return, and study. The gender dierence tends
to be more pronounced for high returns in all three studies. We nd a statistically signicant dierence in NNW2012
under high returns (dierence 3.20; 𝑝 = 0.048) but not low returns (𝑝 = 0.106) and a statistically signicant dierence
in GNV2017 under high returns (dierence 3.60; 𝑝 = 0.033) but not low returns (𝑝 = 0.760). For KNN2021, the
dierence is much larger for high returns (2.59) than for low returns (0.62) but statistically insignicant for both rates
of returns (𝑝 = 0.162 and 𝑝 = 0.770).
7

a. Asymmetric, NNW2012 b. No Feud, NNW2012 c. Symmetric, NNW2012
d. Het-NC, GNV2017 e. Hom-NC, GNV2017 f. Hom-C, GNV2017
g. ComLate, KNN2021
Figure A2. The evolution of gender dierences in contributions by treatment
Note: The gure displays the evolution of gender dierences in contributions separately for each of the treatments
used for the analysis in our paper. Panels A, B, and C display the treatments from Nikiforakis et al. (2012). Panel
D, E, and F do the same for the treatments from Gangadharan et al. (2017). Finally, Panel G displays the evolution
of contributions, for the rst ten rounds, in the ComLate treatment from Koch et al. (2021).8Panels A, B, D, and G
present treatments with normative conict. Panels C, E, and F are treatments without normative conict. Note that
the dierences in levels and trends across treatments are due to dierent mechanisms used (e.g., communication in
Hom-C or punishment in ComLate.) For details about these mechanisms, please see the original studies. “NNW2012”
refers to Nikiforakis et al. (2012), “GNV2017” to Gangadharan et al. (2017), and “KNN2021” to Koch et al. (2021).
8