Evaluating Coalition Theories SOME lNDlCES

Until recently the choice of methods to evaluate coalition theories has not been a serious problem because the vast majority of studies has been restricted to simple. thrcc-person games. When larger groups are used to test various theories. a larger number of coalitions is possible and the coalitions may also vary in size; hence, i t is difficult to compare and evaluate theories when group size is varied. Accordingly. some indices measuring the accuracy of theoretical predictions of coalition likelihood are proposed, compared, and evaluated. Indices to evaluate theoretical predictions of reward divi-sion among the coalition members arc alsocomparcd. and some problems ofdata analysis are discussed. I n research on coalition formation (N-person cooperative games), two basic dependent variables have been used to compare and evaluate coalition theories: the frequency of occurrencc of each possible coalition and the reward division among the coalition members. Since a given theory may be accurate in predicting one of the response measures but inaccurate in predicting the other, both measures are important in evaluating coalition theories. Until recently, the choice of methods used to evaluate coalition theories has not been a serious problem because the vast majority of studies has been restricted to simple, three-person games. where only three pairwise coalitions are possible (in a simple game the prize is con-stant for all winning coalitions). However, when larger groups are used AUTHOR'S NOTE: The preparation of this manuscript was supported in part by research grant from the National Science Foundation (No. Bh'S 7749542). The author wishes to thank David Kravitz and David hkck for critical comments on an earlier draft of the manuscript