Abstract: We study the general problem of public information design for a policy maker. We show that, under appropriate regularity conditions, optimal information design is always given by a finite partition of the state space, whereby the information designer announces the subset in the partition to which the state belongs. When the sensitivity of welfare to shocks is small, optimal partition converges to a linearized partition of the state space into convex polytopes. We derive several universal properties of linearized partitions in terms of principal information components, given by the eigenvectors of a special information matrix that we define explicitly. We then apply our general characterization to study optimal design of central bank announcements.
Written with Anna Cieslak and Andreas Schrimpf.
Finance Seminars at Caltech are funded through the generous support of The Ronald and Maxine Linde Institute of Economic and Management Sciences (lindeinstitute.caltech.edu).