Ian Jewitt, University of Oxford

"Two robust informativenes principles"

Abstract

Holmstrom's Informativeness Principle is the most celebrated result to have emerged from the moral hazard principal agent literature. But, it is only established for the very narrow class of seperable expected utility preferences where attitude to risk is independent of agent action. We introduce the notion of an informationally non atomic family of probability distributions and show that under this (relatively benign) condition, informativeness principles hold which are robust to a much wider preference class. For instance, when there are only two actions the principle extends not only to non separable EU, but also to not necessarily risk averse EU (e.g. Friedman Savage) and non EU preferences such as Yaari and Loss Averse.

Contact person: Peter Norman Sørensen