Informational Separability and Entropy

This paper examines the relevance of entropy to economic analysis via informational separability, whereby suppressing information that differentiates a subset of states in each experiment reduces information in that experiment, relative to an uncertainty measure. I show that with three or more states, informational separability leads to entropy measures. This enriches the toolbox of information theory by providing an alternative perspective on entropy measures. Two applications are discussed: Segregation measurement and rational inattention. It shows that informational separability is a useful—and, in some contexts, perhaps even canonical—notion for thinking about the relevance of entropy measures