Igor Karpov

University of Chicago · Economics · 2026–27 job market

Fields

Theory, Applied Micro

Job market paper

Competitive Persuasion with a Noisy Judge

Can a firm choosing between two agents gain from judging them with error? I show that she can when the agents design the information she receives. Noiseless judging makes even a narrow lead decisive, so agents conceal bad news not to lose; noise makes a narrow lead worth little, so revealing pays more. For a broad class of noise and binary types, the equilibrium is unique: each agent sometimes admits being bad and otherwise pools, and if agents are likely enough to be good, both reveal fully, delivering the first best, unlike under noiseless judging. Moderate noise is best: it curbs concealment without reversing the ranking of posteriors, and the right amount beats noiseless judging at every prior. With a continuum of types, the case for noise survives: under regularity conditions, the firm prefers to judge with error unless the prior density is nondecreasing, and if it is nonincreasing, her best rule tosses a coin only among low posterior means.