On The (Relative) Merits of Money Burning For Optimal Contracting

We consider the following screening model of procurement. An agent (the seller) has private information on his cost parameter. A principal (the buyer) learns an ex post signal on this parameter. The signal is private information to the principal and proper incentives to reveal this signal must be designed. In related contexts, money burning, i.e., the ex post destruction of some of the gains from trade, has shown to be useful to provide such incentives. We demonstrate that money burning allows the principal to implement the first-best output with zero information rent for the agent; although it is never optimal to do so since output distortions are less costly. More generally, money burning is rarely optimal, and only used as a tool of last resort if output distortions are no longer feasible. In particular, when output must be chosen before the non-verifiable signal realizes, money burning becomes more attractive.