Observation and contagion effects in cooperation

What is it about?

In this paper we experimentally compare two ways in which cooperation (or defection) may spread across a network. Subjects are put in groups of four and then interact pairwise in a social dilemma where they independently decide to exert high or low effort. In interpretation you can think of work colleagues who interact in pairs on projects. Label the four workers A, B, C and D. In our baseline treatment A is always paired with B and C is always paired with D. Moreover, A and B never get to observe what C and D are doing and vice versa. The A-B and C-D pair are completely independent. As such, any correlation in behavior between the A-B pair and C-D pair cannot be due to a spillover from one pair to another. Contrast that with our history treatment where the A-B pair and C-D pair get to see what is happening in the other pair. Here we may see an observations effect in which the behavior of, say, the A-B pair is influenced by what they observe happening in the C-D pair. Next consider a connected treatment in which A is sometimes paired with B and sometimes with C. Similarly, C is sometimes paired with D and sometimes with B. Here we may get a contagion effect in which, say, behavior in the A-B pair is influenced by A's interaction with C. To explore the possibility of an observations and contagion effect we use a 2x2 between subject design that combines the three treatments mentioned above with a connected and history treatment. Subjects played both a minimum effort game and public good game. In the minimum effort game we find evidence of an observations effect and no evidence of a contagion effect. This has the effect of significantly increasing cooperation. In the public good game we find no evidence of an observations or contagion effect. Subjects in our experiment appeared to be relatively strategic and so we suggest that the observations effect in the minimum effort game is due to the success of one pair providing a focal point around which the other pair could aspire.

Featured Image

Why is it important?

There is understandably a lot of interest in endogenous social interaction effects. In empirical data, however, it is almost impossible to distinguish the channels through which one person may be influenced by another because of unobserved exogenous factors. In the lab we have more control to pull apart different mechanisms that might cause endogenous effects. Our experiment was carefully designed to compare an observations effect and contagion effect, both of which could cause an endogenous social interactions effect. An observations effect is arguably more in keeping with the standard notion of endogenous social interactions effect in that a person is influenced by observing another. Our experiment did not, however, point towards 'blind conformity'. Instead the observations effect appeared to be driven more by aspiration - if that pair can succeed then so can we. Interestingly, we saw that subjects had no problem adapting their behavior to others - cooperating with one and defecting with another - based on past behavior. This suggests that the contagion effect is limited. One take away result from our study is that observing others can increase cooperation. Indeed, the increase in cooperation we saw was substantial. Observing the success of one pair allowed the other pair to also coordinate on the cooperative outcome. But we reiterate that this only happened in the minimum effort game. It did nothing to stop the familiar decline in cooperation we see in the public good game.

Perspectives