Motivation

Does Group Identification Matter?
Experimental Evidence
Shamlesh Annah
Sebastian Edwards
Christian Freund
Julian Schumacher
Marcin Strzelczyk
Course: Experimental Methods in Economics
Supervisors: Elke Renner, Chris Starmer
7 May 2010
Outline
1.
2.
3.
4.
5.
6.
Motivation
Literature Review
Setup
Experimental Design
Results
Conclusion
Motivation
Most group-based experiments reported in the economic literature so far consider
groups as being a homogeneous entity. Real-world social structures, however,
often exhibit different layers of organization, and different scales on which groups
exist.
Examples:
International Economics: Trade agreements, environmental agreements (Kyoto
Protocol etc.): At least three different scales: The individual, the country, and the
group of countries coming together for this agreement.
Political Economics: Social identities and voter behaviour. At least three different
scales: The individual, the social subgroup it counts itself towards, and the whole
population.
We devised an Public Good experiment designed to illuminate whether the
introduction of an “intermediate social layer”, that is an additional – artificial social
structure between “individual” and “whole group” – affects individual contribution
behaviour in a significant way. The added structure is not reflected in the payoff
structure of the experiment, i.e. we have a psychological game.
Literature Review 1/2
Individual behaviour in PG Games. There seem to be general pattern in public
contribution games, depending on repetition (SOURCE), punishment (SOURCE),
and other factors.
(Isaac, Walker and Williams 1994,Gächter and Renner 2003 and Gächter, Thöni
and Herrmann 2008)
George A. Akerlof and Rachel E. Kranton (2000), for instance, suggest that simply
being able to identify with a group is itself an important source of individual wellbeing.
Groups in Public Good games. (REFERENCE ON GROUP COMPETITION)
PG field experiments. Yoeli (2008) examines the behaviour of customers of an
electricity utility. Work in progress, thus no conclusive results so far.
Literature Review 2/2
The individual and the group. Social comparison theory emphasizes that people in
group settings behave differently than in isolation. In particular, it assumes that
people are motivated both to perceive and present themselves in a socially
desirable way. To accomplish this, a person might react in a way that is closer to
what he regards as the social norm than how he would act in isolation. (Levinger
and Schneider, 1969)
Psychological group effects. Group identification, Insider/Outsider-effects – while
the former would presumably lead to higher contribution (achieve a socially optimal
outcome), the latter might rather decrease average contributions. Hargreaves
Heap and Daniel John Zizzo (2006) find decreasing trust in a group setting, due to
negative discrimination against outsiders.
Groups and Social Capital. Putnam (Julian). Social capital in the form of trust is
most effectively build by experiencing interaction in social networks. Since the
public good game requires trust in the action of others to achieve a social optimal
outcome, this should increase average contributions.
Research Question & Hypothesis
Research Question. Does the existence of an “intermediate social layer”, i.e. of
additional group structure within the population, affect individual contribution
behaviour in a Public Good game in a significant way?
Hypothesis.
H0: Additional structure does not change individual contribution behaviour.
H1: Additional Structure does change individual behaviour (direction not determined)
Experimental Design
Experimental Design
IG treatment
Individual
decision
Preference control
/ repetition bias
PSYCHOLOGICAL
GROUP EFFECT
Repetition control
/ preference bias
Revelation
Group
decision
GI treatment
Group decision
Revelation
iγ,t knows
only φ(πγ,t)!
Revelation
Preference control
/ repetition bias
Individual
decision
Revelation
Payoffs
Preference bias
Repetition bias
Changing beliefs
about other’s
contributions
Imperfect
conditional
cooperation
Declining contributions
Lack of
enforcement/
punishment
Unequal distribution of free riding / conditional
cooperation / social preferences between groups
Confunding group effect with heterogenous
preferences
Mitigate by N → ∞
Results 1/3
Between Group-/Individual Comparison (stage 1)
0.5
Individual
Group
Relative frequency
0.4
0.3
0.2
0.1
0
0
•
•
•
1
2
3
4
5
6
Contribution to Public Good
7
8
9
Higher mean for individual setting (3.78) than for group setting (2.33)
Strong clustering of group contributions, almost uniform contribution of
individual contributions
No significant difference in distributions (Mann-Whitney)
10
Results 2/3
IG Treatment
0.25
Individual
Relative frequency
0.2
Group
0.15
0.1
0.05
0
0
1
2
3
4
5
6
7
8
9
Contribution to Public Good
•
•
•
Higher mean for individual setting (3.78) than for group setting (3.44)
Strong clustering of group contributions, almost uniform contribution of
individual contributions
No significant difference in distributions (Mann-Whitney)
10
Results 3/3
GI Treatment
0.6
Relative frequency
0.5
Individual
Group
0.4
0.3
0.2
0.1
0
0
1
2
3
4
5
6
7
8
9
Contribution to Public Good
•
•
•
Higher mean for group setting (2.33) than for individual setting (1.33)
Overall lower contributions than in IG treatment
No significant difference in distributions (Mann-Whitney)
10
Results: Additional remarks
• Weakly significant mean-difference across
treatment groups (t = -1.64)
heterogenous preference bias?
• Weakly significant difference of individual
contribution across treatments in the
individual setting (z = -1.81)  preference
and repetition bias?
Conclusions
•
•
No significant differences between group/individual contributions
However, these quantitative results are questionable due to limitations:
– subject knew each other, were familiar with experimental design and game
theoretic prediction
– given the time slot, it was not possible to pose control questions in order to make
sure each participant fully understood the game, especially the lack of any
connection between the group structure and the payoff function.
– very small sample size: strong evidence of preference bias
– Resources not sufficient to investigate further group effects
•
•
Slight hint of repetition bias: across-treatment comparison more meaningful
Possible future directions:
– varying the group structure (larger subgroups / different subgroups)
– varying the extent to which these groups are constituted. Here: lowest possible
level, mere identification with a n anonymous group. May be extended to groups
with known members, communication, …
Old slides – not part of
presentation!
Literature Review
Andreoni 1988 runs a PG game to try to solve the
puzzle. Finds neither the learning hypothesis or a
rational strategy can explain the puzzle.
• Fehr and Schmidt 1999 model Inequality aversion
people resist inequitable outcomes meaning they are
willing to give up some material payoff to move in a
direction of more equitable outcomes.
1.
2.
3.
4.
References:
Fischbacher & Gächter – Social Preferences, Beliefs and
the Dynamics of Free Riding in Public Good Experiments,
forthcoming in AER
Fehr & Gächter – Cooperation and Punishment in Public
Goods Experiments, AER 2000
Fehr & Schmidt – A theory of fairness, competition and
cooperation, QJE 1999
Andreoni – Cooperation in Public-Goods Experiments:
Kindness or Confusion?, AER 1988
• This has lead to beliefs that people get utility from the
fact others are benefitting ‘warm glow’.
Literature Review
The social comparison theory emphasizes that people in group settings
behave differently than in isolation. In particular, it assumes that people
are motivated both to perceive and present themselves in a socially
desirable way. To accomplish this, a person might react in a way that is
closer to what he regards as the social norm than how he would act in
isolation. (Levinger and Schneider, 1969)
The persuasive argument theory, deliberation drives group decisions in
a particular direction because it is more persuasive. A related
explanation of group shifts is that people with certain preferences tend
to be more persuasive than others (for example, more selfish
individuals are also more aggressive in deliberation).
(Burnstein et al., 1973; Brown, 1974)
Results (between treatments)
Results (stage 2)
• Group mean (3.44) higher individual mean
(1.33)
• Strong clustering of individual
contributions around 0, rather uniform
group contributions
• Weakly significant (10 %-level) difference
in distributions (Mann-Whitney)