Repeated Interaction and Its Impact on Cooperation and Surplus Allocation - An Experimental Analysis

Sibilla Di Guida, The Anh Han, Georg Kirchsteiger, Tom Lenaerts, Ioannis ZisisGames 2021 · DOI

Sure, here's a blog-style summary of the paper "Repeated Interaction and Its Impact on Cooperation and Surplus Allocation - An Experimental Analysis":

TL;DR

This paper uses a controlled experiment to investigate how repeated interactions and the ability to choose group members affect cooperation and how surplus is shared. They found that while repeated interaction doesn't improve overall cooperation rates, it leads to self-selection into groups with varying lifespans, suggesting that "greedy" distributors acting out of fear of rejection, rather than pure generosity, lead to the most efficient and long-lasting groups.

What Problem Does This Paper Address?

Cooperation is essential for producing surplus, but it can also lead to conflicts over how that surplus is distributed. In many real-world scenarios, like workplaces, individuals have unequal bargaining power, and it's not always possible to pre-negotiate contracts. This can lead to situations where those with less power might refuse to cooperate, or those with more power might exploit them. The paper explores how the ability to interact repeatedly with the same partners and to choose who you work with can influence cooperation and fairness in these unequal situations. Specifically, it asks: does knowing you'll work with someone again, or being able to choose your partners, lead to more cooperation and fairer outcomes when power is uneven?

Methods

The researchers designed a "surplus allocation game" with groups of four: two distributors and two receivers. Distributors had the power to decide how to split the surplus, while receivers could only choose to cooperate or not. They conducted three experimental treatments:

  1. Baseline Treatment (BT): Cooperation was enforced, and groups were randomly re-matched each round. This served as a control.
  2. Re-Match Treatment (RT): Cooperation was not enforced, and groups were randomly re-matched each round. Participants were informed about the previous three rounds' distributor contributions.
  3. Endogenous-Match Treatment (ET): Similar to RT, but groups were maintained if all members agreed to cooperate. If anyone refused, the group dissolved, and members were re-matched. This allowed for endogenous group formation.

They collected data from 400 participants over 30 rounds and analyzed contributions, earnings, and cooperation rates. They also developed a behavioral model to explain the observed dynamics.

Datasets

Not discussed in the paper. The study relies on data generated from their own experimental design.

Key Insights and Ideas

  • Fear of Rejection, Not Pure Generosity: The most surprising finding is that "generous" distributors (those who contribute more) are likely motivated by a fear of rejection by receivers, rather than intrinsic prosociality. This suggests that the threat of losing future cooperation is a powerful driver of fairer outcomes.
  • Self-Selection is Key: While repeated interaction alone didn't boost overall cooperation rates, it enabled a crucial self-selection process. Individuals sorted themselves into groups with different lifespans, leading to a heterogeneity of outcomes.
  • Long-Lived Groups are Efficient: The most efficient and long-lasting groups were composed of "modest receivers" (those with lower acceptance thresholds) and "greedy distributors" (those acting out of fear of rejection). This highlights that optimal group composition isn't necessarily about having all "prosocial" individuals.
  • Unequal Power Amplifies Differences: Instead of leveling the playing field, repeated interaction in unequal power structures seems to amplify existing behavioral differences across groups, leading to both highly cooperative and highly uncooperative teams.

Key Results

  • Contributions: Average contributions were lower in the Baseline Treatment (BT) compared to the Re-Match (RT) and Endogenous-Match (ET) treatments.
  • Earnings: Distributors earned more in BT, while receivers earned more in RT and ET.
  • Cooperation Rates: Average cooperation rates did not significantly differ between RT and ET.
  • Group Duration: In ET, there was significant heterogeneity in group durations, with many short-lived groups and some long-lasting ones.
  • Receiver Behavior: Receivers' likelihood to cooperate increased with distributors' past contributions. Receivers who belonged to many short-lived groups ("multi-group receivers") were more demanding and less likely to cooperate than those in fewer, longer-lived groups.
  • Distributor Behavior: Distributors who belonged to fewer groups tended to contribute more, suggesting that more stable relationships encourage greater generosity (or at least, the appearance of it).

Limitations and Future Work

The authors acknowledge that their model makes a strong assumption that distributors believe receivers are like themselves. They suggest that future work could explore scenarios with asymmetric information and heterogeneous beliefs about social preferences, which could be analyzed using Bayesian Psychological Games. Additionally, they note that their results are qualitative and aim to provide guidelines rather than a precise real-life scenario. Future research could also investigate the impact of unequal donations within a group and how other distributors' donations influence individual choices.

Summary updated

BibTeX
@article{guida2021repeated,
  author = {Sibilla Di Guida and The Anh Han and Georg Kirchsteiger and Tom Lenaerts and Ioannis Zisis},
  title = {{Repeated Interaction and Its Impact on Cooperation and Surplus Allocation - An Experimental Analysis}},
  journal = {Games},
  year = {2021},
  doi = {10.3390/g12010025},
}
Figure