The Generalized Informativeness Principle

Abstract

The informativeness principle, as originally formulated by Holmstrom (1979), relies on the first-order approach. This paper extends the informativeness to settings in which the first-order approach cannot be applied. We introduce a generalized informativeness principle that takes into account non-local incentive constraints and holds generically. The provides a sufficient condition for a signal to have value when the first-order approach cannot be applied. This condition is also necessary for a signal to have value in the absence of restrictions on the utility function.