We show how the verifiability of signals affects their use in mitigating adverse selection. With verifiable signals, the observed practice of investigating agents before contracting is inferior to writing contingent contracts. This holds regardless of the agent's risk aversion, bounded penalties, or investigation costs. With risk-neutral agents, contingent contracts yield first-best outcomes. We further show that even unverifiable signals can be gainfully used in contingent contracts by removing a principal's incentives to distort signals or by removing an agent's incentives to demand verification.
Bibliographical noteFunding Information:
We acknowledge the helpful comments of Kashi Balachandran (the editor), Sugato Bhattacharyya, Jonathan Glover, Arnold Juster, Jack Kareken, Arijit Mukherji, Suresh Radhakrishnan, Stefan Reichelstein, Shyam Sunder, and the seminar participants at Carnegie Mellon University. We are grateful for financial support from the William Larimer Mellon Fund and the Margaret and Richard M. Cyert Family Fund. The first author also acknowledges the financial support of the Deloitte and Touche Foundation.