Theory Workshop - Andrew Joshua Koh (MIT)

Title: Full Implementation with Dynamic Information, joint with Sivakorn Sanguanmoo and Kei Uzui

Abstract: 

We study dynamic information design for full implementation in dynamic supermodular games where players’ opportunities to revise their actions arrive stochastically. We show that noisy but conclusive bad news signals uniquely implement the largest equilibrium across all dynamic information structures. Such structures exhibit asymmetry, noise, and state-invariance: asymmetry skews future switching behavior—conditional on bad news not arriving—towards the designer-preferred action which incentivizes switching in the present; noise increases the probability that bad news does not arrive; state-invariance lifts the time-t optimality of continuation information structures to dynamic optimality. There is no multiplicity gap: the largest implementable equilibrium can be implemented uniquely. We discuss applications to macroeconomics, debt runs, and platform competition.