Relative gain maximization in sequential 3-person characteristic function games

Abstract

The present study tested competitively three descriptive models of coalition formation and payoff disbursement in sequential 3-person games in which each player seeks to maximize the rank of his or her total score in a sequence of interdependent characteristic function games with sidepayments. Two of the models were originally proposed and tested by J. D. Laing and R. J. Morrison. A third mixed-signal model is proposed, postulating that the starting rank position and the values of the characteristic function, which operate as two independent signals, are combined to determine both coalition frequencies and payoff division. To test these models, 25 subjects played 46 different sequences for a total of 236 games in a new experimental paradigm, which generalizes previous research by assigning different values to the three 2-person coalitions, introducing dependency between successive characteristic functions, and climinating face-to-face bargaining. The results support the mixed-signal model over its compentors.