Social Preferences, Beliefs, and the Dynamics of Free Riding in Public Goods Experiments

Abstract

One lingering puzzle is why voluntary contributions to public goods decline over time in experimental and real-world settings. We show that the decline of cooperation is driven by individual preferences for imperfect conditional cooperation. Many people's desire to contribute less than others, rather than changing beliefs of what others will contribute over time or people's heterogeneity in preferences makes voluntary cooperation fragile. Universal free riding thus eventually emerges, despite the fact that most people are not selfish. (D12, D 83, H41, Z13)

Full text

Page 1

Social Preferences, Beliefs, and the Dynamics of
Free Riding in Public Good Experiments



URS FISCHBACHER
SIMON GÄCHTER


CESIFO WORKING PAPER NO. 2491
CATEGORY 2: PUBLIC CHOICE
DECEMBER 2008






An electronic version of the paper may be downloaded
• from the SSRN website: www.SSRN.com
• from the RePEc website: www.RePEc.org
• from the CESifo website: Twww.CESifo-group.org/wpT

Page 2

CESifo Working Paper No. 2491



Social Preferences, Beliefs, and the Dynamics of
Free Riding in Public Good Experiments


Abstract

We provide a test of the role of social preferences and beliefs in voluntary cooperation and its
decline. We elicit individuals’ cooperation preferences in one experiment and use them – as
well as subjects’ elicited beliefs – to explain contributions to a public good played repeatedly.
We find substantial heterogeneity in people’s preferences. With simulation methods based on
this data, we show that the decline of cooperation can be driven by the fact that most people
have a preference to contribute less than others, rather than by their changing beliefs of
others’ contribution over time. Universal free riding is very likely despite the fact that most
people are not selfish.
JEL Code: C91, C72, H41, D64.
Keywords: public goods experiments, social preferences, conditional cooperation, free riding.



Urs Fischbacher
University of Konstanz
P.O. Box D 131
78457 Konstanz
Germany
urs.fischbacher@uni-konstanz.de
Simon Gächter
University of Nottingham
School of Economics
Sir Clive Granger Building
University Park
Nottingham NG7 2RD
United Kingdom
simon.gaechter@nottingham.ac.uk


November 25, 2008
This paper is part of the MacArthur Foundation Network on Economic Environments and the
Evolution of Individual Preferences and Social Norms and the Research Priority Program of
the University of Zurich on “The Foundations of Human Social Behavior: Altruism versus
egoism”. Support from the EU-TMR Research Network ENDEAR (FMRX-CT98-0238) is
gratefully acknowledged. Sybille Gübeli, Eva Poen, and Beatrice Zanella provided very able
research assistance. We also thank very constructive referees and Nick Bardsley, Rachel
Croson, Armin Falk, Ernst Fehr, Justina Fischer, Martin Kocher, Michael Kosfeld, Andreas
Kuhn, Charles Noussair, Eva Poen, Daniel Schunk, Martin Sefton, and various conference,
workshop and seminar participants for their helpful comments. Simon Gächter thanks the
University of St. Gallen, CES Munich, CMPO Bristol and the University of Copenhagen for
the hospitality he enjoyed while working on this paper.

Page 3

1
In this paper we investigate the role of social preferences and beliefs for voluntary
cooperation. This question is interesting because numerous public goods experiments have
shown that many people contribute more to the public good than pure self-interest can easily
explain. However, an equally important observation is that free riding increases in repeatedly
played public good experiments across various parameters and participant pools.1 The facts
are clear, but their explanation is not.
An obvious candidate for explaining the decline of cooperation was learning the free rider
strategy. However, Andreoni (1988) showed that cooperation resumed after a restart, which is
inconsistent with a pure learning argument. In a subsequent paper he interpreted the decline in
cooperation “to be due to frustrated attempts at kindness, rather than learning the free-riding
incentives” (Andreoni (1995), p. 900). Several papers since argue that contributions that are
not due to confusion might possibly be explained by other-regarding preferences (e.g., Palfrey
and Prisbrey (1997); Brandts and Schram (2001); Goeree, Holt and Laury (2002); Ferraro and
Vossler (2005); Kurzban and Houser (2005)). One type of social preference – long argued by
social psychologists (e.g., Kelley and Stahelski (1970)) – is many people’s propensity to
cooperate provided others cooperate as well (e.g., Keser and van Winden (2000); Fischbacher,
Gächter and Fehr (2001); Frey and Meier (2004); Croson (2007); Ashley, Ball and Eckel
(2008); see Gächter (2007) for an overview). Such “conditional cooperation” is an interesting
candidate for explaining the fragility of cooperation because this motivation depends directly
on how others behave or are believed to behave. Conditional cooperators who observe (or

1 See, for instance, Isaac, Walker and Thomas (1984); Andreoni (1988); Andreoni (1995); Weimann (1994);
Laury, Walker and Williams (1995); Croson (1996); Burlando and Hey (1997); Gächter and Fehr (1999);
Sonnemans, Schram and Offerman (1999); Keser and van Winden (2000); Fehr and Gächter (2000); Park
(2000); Masclet, Noussair, Tucker and Villeval (2003); Carpenter (2007); Sefton, Shupp and Walker (2007);
Egas and Riedl (2008); Herrmann, Thöni and Gächter (2008) and Neugebauer, Perote, Schmidt and Loos
(forthcoming). The decline of cooperation has also been observed in prisoner’s dilemma experiments. See, e.g.,
Selten and Stoecker (1986); Andreoni and Miller (1993) and Cooper, DeJong, Forsythe and Ross (1996).

Page 4

2
believe) that others take a free ride, will reduce their contributions and thus contribute to the
decline of cooperation.2 It is unknown, however, to what extent conditional cooperation, and
inter-individual differences in this regard, can explain (the decline of) cooperation. Our paper
aims to shed empirical light on this question. For a related theoretical approach see Ambrus
and Pathak (2007).
The novelty of our paper is to combine two observations from previous research: beliefs
about other people’s contributions matter and people differ in their cooperative preferences –
some are free rider types, whereas others are conditional cooperators.3 In our approach we
measure people’s cooperation preferences in a specially-designed public good game played in
the strategy method (called the “P-experiment”) and then observe the same people in a
sequence of ten one-shot games (labeled the “C-experiment”), in which we also elicit their
beliefs about others’ contributions. This allows us to quantify how preference heterogeneity
and beliefs interact in voluntary cooperation. Specifically, we can disentangle whether
contributions decline because of cooperation preferences and/or because of the way people
form (and change) their beliefs about how others will behave.
Our data from the P-experiment show that people differ strongly in their contribution
preferences. This is consistent with previous evidence. The biggest groups of people are (i)
conditional cooperators who cooperate if others cooperate, and (ii) free riders who never
contribute anything, irrespective of how much others contribute. We push beyond this
observation of preference heterogeneity by investigating how measured preferences and

2 Experiments on “assortative regrouping” of participants provide suggestive evidence for this explanation: in
these experiments (e.g., Burlando and Guala (2005); Gächter and Thöni (2005); Page, Putterman and Unel
(2005); Gunnthorsdottir, Houser and McCabe (2007)) contributors are matched up with contributors, and free
riders with free riders. The results show that contributions are largely stabilized if free riders are moved out of
the group.
3 See, e.g., Fischbacher, et al. (2001); Burlando and Guala (2005); Kurzban and Houser (2005); Bardsley and
Moffatt (2007); Kocher, Cherry, Kroll, Netzer and Sutter (2008); Herrmann and Thöni (forthcoming)).

Page 5

3
beliefs are related to observed contribution behavior. We have therefore designed our
experiments such that we can use the P-experiment to make a point prediction for each
participant about his or her contribution in the C-experiment, given his or her beliefs.
Our approach allows us to answer the following three questions: how do people form their
beliefs about others’ contributions in a given period? What is the respective role of elicited
contribution preferences and beliefs for determining contributions? How do beliefs and
preference heterogeneity impact on the decline of cooperation?
Our results, which we detail in Section II, answer these questions as follows: first, belief
formation can be described as a partial adjustment of one’s belief into the direction of the
observed contribution of others in the previous period. More specifically, beliefs in a given
period are a weighted average of what others contributed in the previous period and one’s
own belief in the previous period. As we will show with the help of simulation methods, our
estimated belief formation process implies that beliefs decline only if contributions decline,
but not vice versa. Second, contributions are significantly positively influenced by predicted
contributions, that is, the elicited preferences. In addition to their preferences, people’s
contributions also depend directly on their beliefs about others’ contributions. This implies
that the P-experiment underestimates the extent of conditional cooperation that occurs in the
C-experiment. Third, contributions decline because on average people are “imperfect
conditional cooperators” who only partly match others’ contributions. The presence of free
rider types is not necessary for this result; contributions also decline if everyone is an
imperfect conditional cooperator.