Abstract

This article reports a comparative test of the central-union theory vis-à-vis several other game-theoretic solution concepts in 3-person sidepayment games. Based on a laboratory experiment, this comparison utilizes nine games in characteristic function form. The solution concepts under test include the equal excess model, the Myerson–Shapley solution, the kernel, and two variants of the central-union theory. With regard to the player's payoffs, results show that the CU-1, CU-2, kernel, and equal excess theories have essentially equal predictive accuracy and that all of these are more accurate than Myerson–Shapley. When the solution concepts are extended and coalition structure probability predictions are incorporated in the test, one version of the central-union theory is overall more accurate than the other solutions.