|
Kay-Yut Chen, Bernardo A. Huberman and Basak Kalkanci
Information Dynamics Laboratory, HP Labs
Abstract
We study the agency problem experimentally focusing on two issues that are
central to its effectiveness. The first tests whether an incentive
compatible direct revelation mechanism performs well when human agents are
asked to report probabilistic information. The second addresses the
principal’s lack of knowledge as to how effort levels relate to the final
outcome. Our results reveal several behavioral effects that reduce the
efficiency of the principal-agent mechanism. We find out that human agents
underestimate low probabilities and overestimate high probabilities,
introducing errors into what should be a truth-telling mechanism.
Furthermore, principals were observed to underpay their agents by
substantial amounts. These behavioral issues may explain why contracts
designed through standard principal-agent models are seldom used in
practice.
Full paper in PDF format
|
 |
|
Printable version
