ABCs of principal-agent interactions: Accurate predictions, biased processes, and contrasts between working and delegating

We experimentally investigate people's evaluations of incentive pay contracts and people's predictions of others' evaluations of incentive pay contracts. We emphasize that the construction of evaluations and predictions often includes two substeps, involving likelihood judgment and likelihood weighting. Predictors appear to be biased at both substeps but in opposing directions. Accurate overall predictions thus sometimes reflect two errors that are of the same magnitude and thereby offset. Moreover, predictions can become more inaccurate if one step is debiased but the other is left untouched. Importantly, principals deciding whether to delegate a task are susceptible to just one of the biases. Delegation assessments are thus often flawed, reflecting a single error that is not offset.