The piecewise power value function proposed by Kahneman and Tversky is inconsistent with both theoretical and experimental work. We propose the expo-power function as an alternative to the power function and examine its ability to explain choices in Allais experiments.

We demonstrate theoretically and illustrate the implications of assuming power utility when the true function is of the expo-power form. Empirical results can appear to be consistent with cumulative prospect theory when they are in fact generated from a Markowitz model.