Persistent Distortionary Policies with Asymmetric Information

Why are distortionary policies used when seemingly Pareto improvements exist? According to a standard textbook argument, a Pareto improvement can be obtained by eliminating the distortions, compensating the losers with a lump sum transfer, and redistributing the gains that are left over. We relax the assumption that winners know the losses suffered by the losers and show that the informationally efficient method of compensating losers may involve the use of seemingly inefficient (but informationally efficient) distortionary policies. The risk of overcompensating losers may make distortions informationally efficient, as there are points on the Pareto frontier where distortions are used.