On the Prudence of Rewarding A, While Hoping for B: Performance Measures and Effort Allocation Between Tasks

Does a performance measure which is a monotonous function of the benefit (a benefit measure) lead to a desirable allocation of effort between the tasks which create this benefit? We analyse two settings. If the mechanism designer knows the choice set of the agent, benefit measures generally do not result in the second-best effort allocation – unless restrictive assumptions about agent’s preferences hold. In this setting, it thus makes sense to reward according to a performance measure A, while hoping for benefit B. If the mechanism designer works with the wrong choice set, the designed mechanism is vulnerable to gaming. Then, benefit measures may inhibit damaging allocations and the gaming loss can be reduced. In this setting, rewarding B while hoping for B is reasonable.