Reassessing the Testing of Generic Utility Models for Mixed Gambles.

Miyamoto's generic utility theory (GUT) is a bilinear form that captures a diverse set of utility formulations. The present study, using mixed gambles, experimentally evaluates GUT in a fashion similar to the Chechile and Cooke and the Chechile and Butler studies, but employs a novel method of analysis. The reasons for the new method is to solve a fundamental flaw with the regression approach used in the earlier experiments and to solve a problem of model overfit. Several participants from the earlier Chechile and Butler experiment are now recognized as being consistent with the GUT representation, but many are not. A new experiment, with actual economic consequences for the participants, does not support GUT. Suggestions are provided for subsequent research studies assessing the GUT class of models. Copyright 2003 by Kluwer Academic Publishers