Instrumental Variables (IV) methods identify internally valid causal effects for individuals whose treatment status is manipulable by the instrument at hand. Inference for other populations inevitably requires some sort of homogeneity assumption. I develop a simple theoretical framework that...