Abstract Privacy of consumers or citizens is often seen as an inefficient information asymmetry. We challenge this view by showing that privacy can increase welfare in an informational sense. It can also improve information aggregation and prevent inefficient statistical discrimination. We show how and when the different informational effects of privacy line up to make privacy efficient or even Pareto-optimal. Our theory can be applied to decide who should have which information and how privacy and information disclosure should be regulated. We discuss applications to online privacy, credit decisions and transparency in government.