Logconcavity versus logconvexity: A complete characterization

This paper provides a complete characterization of logconcavity, an increasingly popular concept in the economics of uncertainty and information. New and known results are proven without assuming that density functions are differentiable. A systematic comparison between logconcavity and logconvexity is made and the source of the asymmetries between the two is investigated. The key difference is that logconcavity is preserved under one-sided integrations regardless of the types of distribution supports. This property, however, does not hold for logconvexity. Logconcavity for multivariate distributions is also discussed. (C) 1998 Academic Press.