In a laboratory experiment, we investigate behavior in a principal–agent situation with moral hazard. We evaluate the predictive success of two theories. One is the standard agency theory, which assumes that the agent will accept any contract offer that satisfies his participation constraint, typically requiring zero expected utility. The other is the “fair-offer” theory suggested by Keser and Willinger [2000. Principals’ principles when agents’ actions are hidden. International Journal of Industrial Organization 18 (1), 163–185], which requires that the principal provide full insurance against losses to the agent and leave him a share of at most 50% of the generated surplus. The treatment variable of our experiment is the cost of effort. As effort costs increase, expected net surplus of a contract decreases. We observe that fair-offer theory generally predicts observed contract offers better than standard agency theory. However, the predictive success of the fair-offer theory decreases, while the one of standard agency theory increases with decreasing expected net surplus.