DP11157 Endogenous Repeated Cooperation and Surplus Distribution - An Experimental Analysis

Author(s): Sibilla di Guida, The Anh Han, Georg Kirchsteiger, Tom Lenaerts, Ioannis Zisis
Publication Date: March 2016
Keyword(s): group formation, repeated cooperation, surplus distribution
JEL(s): C92, D03
Programme Areas: Industrial Organization
Link to this Page: cepr.org/active/publications/discussion_papers/dp.php?dpno=11157

This paper investigates experimentally how the endogenous group formation combined with the possibility of repeated interaction impacts cooperation levels and surplus distribution. We developed a Surplus Production Distribution Game where the cooperation of four agents is needed to produce a surplus. In case of cooperation, two of the four subjects, the distributors, decided how much of surplus each of them wanted to give to the two other agents, the receivers. This game was played repeatedly with different matching procedures. In the Re-match Treatment (RT) the subjects got randomly re-matched every round, while in the Endogenous-match Treatment (ET) a group was maintained as long as its members cooperated. There was also a Base treatment (BT) where cooperation was exogenously enforced. We found that the distributor's contributions were higher in the ET and the RT than in the BT - unsurprisingly, receivers' possibility to refuse cooperation led to more equal surplus distributions. But contrary to commonly hold beliefs, the possibility of repeated interaction did not lead to higher cooperation levels and more equal allocations of the surplus. Instead, endogenous group formation combined with the possibility of repeated interaction led to self-selection of the subjects in the ET. The endogenous group duration varied drastically between different groups in the ET, with long-lived groups exhibiting contributions and cooperation levels higher than in the RT, while short-lived groups showed contributions and cooperation levels lower than in the RT. Furthermore, for given contribution levels, receivers were more likely to refuse cooperation when their average relationship length was short. This shows that long-lived groups consisted of generous distributors and not so demanding receivers, while ungenerous distributors and demanding receivers formed short-lived groups. Hence, the possibility of repeated interaction does not necessarily increase cooperation and efficiency levels when combined with endogenous group formation. Rather, such a situation might lead to self-selection of agents.