June 21, 11.50-12.20: Georg Sator, University of Vienna
Peer Effects and Social Closeness
We introduce social closeness to the economic literature on peer effects by studying how their strength and magnitude depends on how closely a decision-maker identifies with its peer.
There is a large psychological literature on what characterizes social closeness. We work with the definition of unilateral subjectively perceived closeness to another person. Most importantly this is sharply distinct from social proximity, which aims at how many, what kind of or how important personal characteristics someone may share with someone else.
We speak of peer effects if the mere presence of other people or knowledge of their behaviour, changes a decision-maker’s preferences and hence her choices. We carefully demarcate peer effects conceptually and empirically from influences stemming from other sources of interaction such as strategic considerations, information spillovers or signalling.
A thorough understanding of peer effects in general, their magnitude and the circumstances under which they can be expected to arise is important because of the economic significance of the settings in which they have been shown to play a role. Peer effects have been shown to be a crucial ingredient in determining investments in education, effort provision at the workplace, charitable giving or the sustainability of cooperation when institutions are weak.
Despite the fact that there is plenty of evidence for the relevance of peers for people’s choices, little is known about the conditions under which we can expect them to occur. Since peer effects rely on reputational concerns and imitation and both of these concepts are likely to be affected by who the other person is and how she relates to oneself, we think that social closeness between peers is a likely candidate for their occurrence. So far peer effects have been studied in (typically mutually anonymous) lab settings, where the decision-maker had no information about her peers and how she compares to herself or in field settings with little control over the social closeness between peers.
We use an experimental design that allows us to systematically vary social closeness between peers and hence study its effect on choices. On top of that we isolate the two channels though which peer effects can operate: first, the effect on an observed agent, who may make a different choice due to the fact that the peer can see what she is doing, or second, the effect on the observing agent, who, prior making her own choice, receives information about what her peer did in an identical situation.
A key feature of our design constitutes the unusual experimental procedure. We recruited subjects pairwise with a friend and moreover, we introduce them to a stranger prior to the start of the actual experiment. In the course of the experiment, interaction between peers then takes place between either a pair of friends, strangers, anonymous subjects, or in the control, not at all.
Since we are interested in cooperation we use a standard linear public good game.
We find that peer effects are indeed affected by social closeness.
Observed subjects choose more prosocially than subjects who choose in private. Interestingly, when differentiating with respect to social closeness, we see that this general result is exclusively driven by subjects who are observed by people they do not or hardly know. If observations are made between agents who know is other very well from outside the lab the effect completely vanishes.
With respect to those subjects who observe a peer’s decision prior to making their own decision we find that their choices are more similar to the choice they have observed that two subjects’ choices are when there was no observation between them. Differentiating with respect to social closeness reveals that this effect is driven by those subjects who observed socially close people.
In the light of these results we argue that the occurrence of peer effects depends on the social closeness between the peers in the following way: The drivers of peer effects seem to be reputational concerns vis-à-vis strangers and imitation of friends. We conduct multiple robustness checks to rule out alternative explanations. Most notably we control for similarity of socially close people, such as friends, in unobserved characteristics, which is a notorious difficulty in research on peer effects.