Generalized trust and prosocial behavior
Regina Kühne∗
June 11, 2012
This is a DRAFT. Please do not quote or cite.
Comments are very welcome (Email to [email protected])
Abstract
There is enormous empirical evidence from lab and field experiments that people frequently
engage in prosocial behavior. In a simple model, I analyze the evolution of prosocial behavior
where heterogeneity in beliefs about the fellow men’s willingness to help is one of the driving
forces of the evolution. As I abstract from the possibility of reputation building and punishment
between anonymous partners, I remove the main motives for prosocial behavior and reduce it to
a simple non-strategic decision. The motivation to behave prosocially is then mainly intrinsic
and based on preferences for behaving cooperatively and positive beliefs about the behavior
of others. The model suggests that it exists a stable equilibrium with a positive number of
cooperative people in a society.
JEL: D03 D64 D83
Keywords: trust, beliefs, prosocial behavior, conditional cooperation, indirect evolutionary
approach
∗ PhD
Candidate, Institute of Public Economics at Humboldt University Berlin, [email protected]
1
1
Introduction
There is overwhelming evidence from experiments and everyday experience that motivation for
human behavior is more complex than maximizing material payoffs. The classical models which
are based on this assumption often fail to predict behavior in social interactions. One important
conduct that classical models often cannot explain is prosocial behavior: an action that mainly
benefits others at a cost to oneself. This concerns such diverse areas like helping strangers, tax
compliance, voting, donation of blood, giving to charities and the like.
This paper relates prosocial behavior to generalized trust i.e. to positive expectations about
the behavior of others. The concept of generalized trust is to be distinguished from particularized
trust. The latter refers to trust between people who know each other and are hence able to base
their expectation about the other’s behavior on experience from face-to-face interactions or via
third parties. Generalized trust, however, is the extension of interpersonal trust to people who
the trusting individual does not know and hence has no direct information on. The importance
of generalized trust for various economic phenomenons including growth (e.g. Knack and Keefer,
1997) and civic participation (e.g. La Porta, Silanes, Shleifer, and Vishny, 1997) has increasingly
been acknowledged.
Motivated by this evidence, economists have analyzed in lab experiments how prosocial behavior
in encounters with strangers is related to generalized trust (Glaeser, Laibson, Scheinkman, and
Soutter, 2000; Gächter, Herrmann, and Thöni, 2004; Anderson, Mellor, and Milyo, 2004; Thöni,
Tyran, and Wengström, 2012). The authors mainly conclude that generalized trust is positively
related to voluntary contributions. This is in line with the general finding of dictator and public
good game experiments that voluntary contributions are positively correlated with beliefs about
others’ contributions even in non-strategic situation where the dominant strategy is to contribute
nothing. This fascinating empirical result is difficult to explain by conventional models. However,
beliefs and trust is rarely accounted for in economic models.
The novelty of this paper is to incorporate beliefs about the others’ behavior into a simple
theoretical model of prosocial behavior in encounters between strangers. This kind of setting is
especially interesting as interactions between anonymous partners are increasingly frequent and
tend to replace face-to-face interactions in many areas of life. In this setting, I abstract completely
from possibilities of punishment and reputation building which are seen as some of the main driving
forces of cooperation. In this way, there is no incentive to behave prosocially due to strategic reasons.
Instead motives are mainly intrinsic i.e. behavior is driven by preferences for cooperation and beliefs
about the others’ behavior only. As the behavior of one agent is affecting the behavior of others,
it is of special interst to look at the dynamics of preferences and beliefs. I therefore analyze the
2
coevolution of prosocial behavior and generalized trust using an indirect evolutionary approach.
Depending on a few parameters, a stable equilibrium can emerge with a positive probability of
cooperative behavior.
2
Literature on prosocial behavior in non-strategic situations
Prosocial behavior of individuals has been studied extensively in lab experiments. Public good
games, dictator games and helping games are good instruments to analyze prosocial behavior as
own contributions mainly increase the payoff of others. In all games, the dominant strategy is to
give nothing in order to maximize economic payoffs. However, it has been shown that there is much
more cooperation than predicted by standard theory even in anonymous situations1 . In dictator
games, individuals tend to give on average between 20 % and 30 % of their endowment to the other
player. However, contributions tend to decay over time with repeated games. In addition, there is a
solid body of empirical evidence that individuals do care about how their behavior compares to the
actions of others. Informing individuals of others’ decision tends to change the individual’s behavior
as they adjust to the amount given by others. This has been shown in lab experiments with dictator
games (e.g. Krupka and Weber, 2009; Duffy and Kornienko, 2010) and public good games (e.g.
Kurzban, McCabe, Smith, and Wilson, 2001; Fischbacher, Gachter, and Fehr, 2001; Zafar, 2011)
as well as in field experiments on charitable giving (e.g. Chen, Harper, Konstan, and Li, 2010;
Shang and Croson, 2009; Frey and Meier, 2004; Andreoni and Scholz, 1998) and contributions to an
online community (Chen, Harper, Konstan, and Li, 2010) where information on the amount given
by others or the share of people giving changed contribution behavior in the way that individuals
tend to adjust to the mean contribution.
If there is no information available on the others’ behavior, individuals also tend to give more
when they belief that others are giving more (e.g. Iriberri and Rey-Biel, 2008; Zafar, 2011; Thöni,
Tyran, and Wengström, 2012). However, this kind of correlation of believes and contributions
does not establish causation. The correlation might as well be due to a false consensus effect
which causes individuals to believe that others tend to contribute the same as them (Ross, Greene,
and House, 1977). To overcome this problem, Fischbacher, Gachter, and Fehr, 2001 developed
an experiment where people indicate how much they would contribute to a public good given all
possible average contributions of the other group members. They find that approximately 50 % of
the people were conditional cooperators who increase their contributions when others contribute
more, whereas about 25 % where always free-riding and the rest showed more complicated patterns.
1 See
Camerer, 2003 and Engel, 2011 for an overview of evidence on dictator games
3
Similar observations have been made when people play public good games repeatedly and update
their beliefs. Fischbacher and Gachter, 2010 find that when people learn about the behavior of
others they update their beliefs about how much others are contributing in the next round and will
adjust their behavior.
A similar tendency has also been observed in laboratory experiments when beliefs about what
others are doing are not explicitly elicited but rather a wider measure of beliefs is applied, namely
a measure of generalized trust. Gächter, Herrmann, and Thöni, 2004; Anderson, Mellor, and Milyo,
2004 and Thöni, Tyran, and Wengström, 2012 use survey questions2 to elicit individual level of trust
in order to investigate the link between trust and voluntary contributions in public good games.
They mainly find a positive correlation between generalized trust and voluntary contributions.
However, the causal relationship is ambiguous. There is no study which investigates how a change
in individual trust level changes prosocial behavior.
Though experiments vary in their settings, there are two stylized facts that can be drawn from
these studies: (i) individuals differ with respect to their preferences for cooperation, and (ii) beliefs
about other people’s willingness to contribute matter. There are two major competing explanations
for prosocial behavior which is conditional of the behavior of others. It may be triggered by notions
of fairness such as reciprocity. Or it might be a result of the wish to behave appropriately by
conforming to a norm3 . In both cases, beliefs about the behavior of others are taken as a signal
either for an appropriate or a fair behavior. A few studies try to discriminate whether individuals
engage in social comparison due to reciprocity or conformity (Bardsley and Sausgruber, 2005;
Falk, 2004; Bohnet and Zeckhauser, 2004) but results are unclear and both seems to be a driving
motivation.
There are a few theoretical models which explain conditional cooperation in non-strategic settings including model of altruism, self-signaling, reciprocity, conformity, inequality aversion and
norm compliance. The most closely related to the model developed in this paper is Traxler and
Spichtig, 2011. The authors assume that there is a norm to contribute to a public good whose
strength is perceived differently by individuals. In their model, an individual voluntarily contributes to a public good if internal or external sanction from norm violation outweighs the cost
of norm compliance. As the strength of the norm is determined by the share of people complying
with the norm, they are able to show the evolution of the norm.
In the model developed here, individuals have a preference for cooperation and these preferences
2 There
are several survey questions that try to capture different aspects of trust. However, the standard trust
question of the General Social Survey (“Generally speaking, would you say that most people can be trusted or that
you can’t be too careful in dealing with people”) is primarily taken to investigate the link between trust and different
economic phenomena.
3 Other explanations include competitive motives and inequality aversion. See Fehr and Schmidt, 2006 and Meier,
2007 for a survey of theories.
4
are endogenous which means that they may change through experience in encounters. Instead of
heterogeneous norm sensitivities, subjects have subjective beliefs about how others behave incorporating the notion of generalized trust. Besides their preferences for cooperation, they condition
their behavior on what they belief others are doing.
3
A simple evolutionary model of prosocial behavior
Consider the following situation. There are randomly chosen pairwise encounters between members
of a large population. When they meet, one subject is in the need of help and can only be helped
by the other subject. The latter can either cooperate and thus help the one in need or shirk. His
move is denoted by
xi ∈ {0, 1}
where xi = 1 when he cooperates and xi = 0 when he shirks. Shirking is costless but adds nothing
to the payoff of the other agent. Cooperation costs c but results in a payoff b for the other individual
where b > c > 0. The payoff Πi (xi ) for move xi is Πi (xi ) = −xi c for agent i and Πj (xi ) = xi b for
agent j. Cooperating in this manner is socially efficient.
Preferences
The two subjects are strangers who cannot rely on past of future behavior to establish mutual benefit from cooperation. In addition, those encounters are not directly observable by other
members of the population. In this way punishment or reputation building are not possible and the
decision is therefore a non-strategic one with a strong incentive to free-ride. One of the motives that
drives prosocial behavior in this kind of situations is altruism: the tendency to behave prosocially
towards unrelated others. If a population member derives utility from the benefit of others, the
individual would decide to cooperate. On the contrary, a selfish individual would decide to shirk.
However, as seen above, individuals do not decide independently of what they expect others would
do in their situation. If a altruistic individual assumes that a considerable part of her fellow men
would not cooperate, he would suffer from not behaving appropriately. The same holds true for
selfish individuals. If they assume that a considerable share people would behave cooperatively,
they would gain utility from doing the same. They would comply to what they perceive to be the
norm.
The Utility that person i derives from Πi (xi ) and Πj (xi ) draws heavily on a preference-model
developed by Levine, 1998:
Ui (xi ) = Πi (xi ) + βij Πj (xi ) for all i 6= j
5
βij =
ai + λasi
with λ > 0
1+λ
The coefficient of the own monetary payoff is normalized to one for all individuals. However,
the weight on person j’s payoff varies over individuals and depends on the parameters ai and asi .
ai ∈ {0, 1} is person i’s unconditional goodwill towards j. If ai = 1, j’s payoff enters i’s utility
function directly; i is altruistic. If ai = 0, he is unaffected by j’s payoff. The parameter asi
indicates i’s subjective belief of how many of his fellow men would help in this situation, reflecting
the individual level of generalized trust. The beliefs are often flawed since the encounters are
not directly observable by others. However, the expected value corresponds to the true share of
altruistic people and people helping in the population i.e. asi ˜N (α, σ). As subjects talk about their
experience in the encounters, they distribute limited information. This process is not explicitly
modeled here. λ incorporates an element of fairness or norm compliance. If is λ 6= 0, this causes
subject i to consider what others are doing in his decision. Here, I assume that λ > 0 which is in
line with finding in experiments (see section 2).
Person i will decide to cooperate if Ui (1, α) > Ui (0, α). This is true iff
−c +
ai + λasi
b>0
1+λ
asi >
c(1 + λ) ai
−
bλ
λ
asi > Z(ai )
Z(1) < Z(0)
In order to allow individuals to be conditional cooperators, an assumption need to be made
regarding the values of b, c, and λ:
Assumption A1: The values of b, c, and λ need to allow that 1 < Z(0) < Z(1) < 0 i.e.
λ
1+λ
>
c
b
>
1
1+λ .
The probability that person i helps person j is then:
pt = αt f + (1 − αt )g
where f = F (Z(1)) = P (Z(1) < asi ), asi ˜N (αt , σ)
and g = F (Z(0)) = P (Z(0) < asi ), asi ˜N (αt , σ)
6
pt
1.0
0.8
0.6
pt HΑt L
0.4
0.2
0.2
0.4
0.6
0.8
1.0
Αt
Figure 1: Probability that individual j receives help with varying α and 45◦ -line (dashed).
This means that f is the probability that person i behaves cooperatively given that he is altruistic; g is the probability that i is cooperative, given he is not altruistic. Since Z(1) < Z(0),
1 > f > g > 0 needs to be true too. Both g and f increase with α as individuals condition
their behavior to some extend on what they believe others would do in this situation. As a result
of conditional cooperation, the probability that someone helps is higher (lower) than the share of
altruistic individuals in the population for high (low) levels of α (see figure 1 for an illustration).
Equilibrium
I do not think altruistic preferences are genetically given, but rather think it is a trait acquired
through socialization or experience. In that sense, subject j - the subject in the need of help - is
able to develop or lose altruistic preferences through experience made in the encounter. I assume a
subject to change his preferences if the behavior of the other deviates strongly from his expectation
i.e. if he did not anticipate to be helped yet received help or, vice versa, he was expecting to be
helped but did not receive any help. When they adopt a trait, they change their preferences in
the direction of the experience they had. Who received help is then more likely to return help
given his beliefs. Assume that subject j has altruistic preferences (i.e. ai = 1) and he expected a
relatively high share of his fellow men to behave cooperatively (i.e. asi > y, with y denoting the
threshold for learning; 0.5 < y < 1) then he would be disappointed and change his preferences.
Vice versa for a subject j with non-altruistic preferences. He would not change his preferences
through his experience if he expected to be helped. Only if his expectations where sufficiently low
(i.e. asi < (1 − y), with (1 − y) denoting the threshold for learning) he would change his preferences.
Note that the preference of his opponent does not have any influence but only his behavior.
7
Η,Ω
0.8
0.6
Ω(Α)
Η(Α)
0.4
0.2
0.2
0.4
0.6
0.8
1.0
Αt
Figure 2: Learning probabilities η and ω with varying α.
Let η denote the probability that a subject with altruistic preferences changes his preferences
and ω denote the probability that a selfish subject would change his preferences, then we can write
them as:
η
= P (asi > y) mit asi ˜N (αt , σ)
ω
= P (asi < 1 − y) mit asi ˜N (αt , σ)
While η increases with increasing α, ω decreases with growing α. Figure 2 depicts both learning
probabilities with varying α.
As the behavior of one subject influences the behavior and preferences of others, it is interesting
to look at the dynamic development of altruistic preferences. The share of the population with
altruistic preferences is determined by the share in the previous period plus the share of people that
changed their behavior.
αt
∆α
= αt−1 + ω(1 − αt−1 )pt−1 − ηαt−1 (1 − pt−1 )
(1)
= ω(1 − αt−1 )pt−1 − ηαt−1 (1 − pt−1 )
(2)
8
DΑ
0.02
0.01
0.2
0.4
0.6
0.8
1.0
Αt
-0.01
-0.02
-0.03
-0.04
-0.05
Figure 3: Evolution of α
The share of altruistic people in the population will then increase if:
ω(1 − αt−1 )pt−1
ω
η
>
>
ηαt−1 (1 − pt−1 )
αt−1 (1 − pt−1 )
(1 − αt−1 )pt−1
This holds true for low level of α. However, for high level of α, we will find that
αt−1 (1 − pt−1 )
ω
<
η
(1 − αt−1 )pt−1
Figure 3 shows the evolution of people with altruistic preferences. It emerges an inner stable
equilibrium given parameters do fulfill assumption A1.
Comparative statics
The equilibrium share of altruistic people in a population depends on several factors:
- An decreases of the cost to benefit ratio will increase the equilibrium value of α. The higher
individuals perceive the benefit of helping to be relative to the cost which helping creates to
them, the more they are likely to help. This holds true for both individuals with altruistic
and non-altruistic preferences.
- Biased beliefs about the behavior of others also changes the equilibrium level of α. A high
(low) level of generalized trust in a society results in a relatively high (low) level of cooperative
behavior.
- The effect of a change of the threshold parameter for learning (y) and the parameter incor9
porating fairness or norm compliance (λ) is ambiguous. It can both increase and decreases
the equilibrium share depending on a certain threshold level of α, shifting the equilibrium
towards the extreme points of α.
4
Concluding remarks
Encounters with strangers constitute an increasingly large proportion of our everyday social
interactions. We are interacting with people who we do not know, to whom we are not liked by
any obligation or affection and thus to whom we owe nothing. Still, we are frequently engaging
in behavior that is costly and mainly benefits a stranger. The model shown in this paper
explains variation in behavior in this kind of situations. As shown in many lab experiments,
individuals are more likely to behave cooperatively when they think others do so as well. The
model suggests that a high level of generalized trust in a society promotes a relatively high
level of cooperative behavior even among strangers. This is in line with findings of empirical
studies that relate generalized trust to different economic phenomenons (e.g. Knack and
Keefer, 1997; La Porta, Silanes, Shleifer, and Vishny, 1997). Another factor that facilitates
cooperative behavior is the perception of the benefit of behaving cooperatively. The higher a
person perceives the benefit of his action the more likely he is to behave cooperatively.
This has certain policy implications. Especially in situations where formal contracts are limited or too onerous and personal relations are not possible, information on positive behavior
of others may help to improve cooperation by correcting exceedingly pessimistic beliefs. Additionally, making people more aware of the benefit of their behavior may promote cooperative
behavior.
References
Anderson, Lisa R., Jennifer M. Mellor, and Jeffrey Milyo (2004). “Social Capital and Contributions in a Public-Goods Experiment”. In: The American Economic Review 94.2, pp.
373–376.
Andreoni, James, and John Karl Scholz (1998). “An Econometric Analysis of Charitable Giving with Interdependent Preferences”. In: Economic Inquiry 36.3, pp. 410–28.
Bardsley, Nicholas, and Rupert Sausgruber (2005). “Conformity and reciprocity in public good
provision”. In: Journal of Economic Psychology 26.5, pp. 664 –681.
10
Bohnet, Iris, and Richard Zeckhauser (2004). “Social Comparisons in Ultimatum Bargaining”.
In: Scandinavian Journal of Economics 106.3, pp. 495–510.
Camerer, C (2003). Behavioral Game Theory: Experiments in Strategic Interaction. Princeton
University Press.
Chen, Yan, F. Maxwell Harper, Joseph Konstan, and Sherry Xin Li (2010). “Social Comparisons and Contributions to Online Communities: A Field Experiment on MovieLens”. In:
American Economic Review 100.4, pp. 1358–98.
Duffy, John, and Tatiana Kornienko (2010). “Does competition affect giving?” In: Journal of
Economic Behavior & Organization 74.1-2, pp. 82–103.
Engel, Christoph (2011). “Dictator games: a meta study”. In: Experimental Economics 14 (4),
pp. 583–610.
Falk, Armin (2004). Charitable Giving as a Gift Exchange: Evidence from a Field Experiment.
IZA Discussion Papers 1148. Institute for the Study of Labor (IZA).
Fehr, Ernst, and Klaus M. Schmidt (2006). “The Economics of Fairness, Reciprocity and
Altruism - Experimental Evidence and New Theories”. In: ed. by S. Kolm and Jean Mercier
Ythier. Vol. 1. Handbook on the Economics of Giving, Reciprocity and Altruism. Elsevier.
Chap. 8, pp. 615–691.
Fischbacher, Urs, and Simon Gachter (2010). “Social Preferences, Beliefs, and the Dynamics of
Free Riding in Public Goods Experiments”. In: American Economic Review 100.1, pp. 541–
56.
Fischbacher, Urs, Simon Gachter, and Ernst Fehr (2001). “Are people conditionally cooperative? Evidence from a public goods experiment”. In: Economics Letters 71.3, pp. 397–
404.
Frey, Bruno S., and Stephan Meier (2004). “Social Comparisons and Pro-Social Behavior:
Testing ”Conditional Cooperation” in a Field Experiment”. In: The American Economic
Review 94.5, pp. 1717–1722.
Gächter, Simon, Benedikt Herrmann, and Christian Thöni (2004). “Trust, voluntary cooperation, and socio-economic background: survey and experimental evidence”. In: Journal of
Economic Behavior & Organization 55.4, pp. 505 –531.
Glaeser, Edward L., David I. Laibson, José A. Scheinkman, and Christine L. Soutter (2000).
“Measuring Trust”. In: The Quarterly Journal of Economics 115.3, pp. 811–846.
11
Iriberri, Nagore, and Pedro Rey-Biel (2008). Elicited beliefs and social information in modified
dictator games: What do dictators believe other dictators do? Economics Working Papers
1137. Department of Economics and Business, Universitat Pompeu Fabra.
Knack, Stephen, and Philip Keefer (1997). “Does Social Capital Have an Economic Payoff? A
Cross-Country Investigation”. In: The Quarterly Journal of Economics 112.4, pp. 1251–
1288.
Krupka, Erin, and Roberto A. Weber (2009). “The focusing and informational effects of norms
on pro-social behavior”. In: Journal of Economic Psychology 30.3, pp. 307 –320.
Kurzban, Robert, Kevin McCabe, Vernon L. Smith, and Bart J. Wilson (2001). “Incremental
Commitment and Reciprocity in a Real-Time Public Goods Game”. In: Personality and
Social Psychology Bulletin 27.12, pp. 1662–1673.
La Porta, Rafael, Florencio Lopez-de Silanes, Andrei Shleifer, and Robert W. Vishny (1997).
“Trust in Large Organizations”. In: The American Economic Review 87.2, pp. 333–338.
Levine, David K. (1998). “Modeling Altruism and Spitefulness in Experiments”. In: Review of
Economic Dynamics 1.3, pp. 593 –622.
Meier, Stephan (2007). “A survey of economic theories and field evidence on pro-social behavior”. In: ed. by Bruno S. Frey and Alois Stutzer. Economics and Psychology. The MIT
Press, pp. 51–87.
Ross, Lee, David Greene, and Pamela House (1977). “The “false consensus effect”: An egocentric bias in social perception and attribution processes”. In: Journal of Experimental
Social Psychology 13.3, pp. 279 –301.
Shang, Jen, and Rachel Croson (2009). “A Field Experiment in Charitable Contribution:
The Impact of Social Information on the Voluntary Provision of Public Goods”. In: The
Economic Journal 119.540, pp. 1422–1439.
Thöni, Christian, Jean-Robert Tyran, and Erik Wengström (2012). “Microfoundations of Social Capital”. In: Journal of Public Economics 0, pp. –.
Traxler, Christian, and Mathias Spichtig (2011). “Social norms and the indirect evolution of
conditional cooperation”. In: Journal of Economics 102 (3), pp. 237–262.
Zafar, Basit (2011). “An experimental investigation of why individuals conform”. In: European
Economic Review 55.6, pp. 774 –798.
12