Three experiments were designed to demonstrate that job performance inferences from personality inventories rely more on the agentic or communal value conveyed by the items compared with the Big-Five traits they are supposed to describe. In the first two experiments, the participants had to predict the job performances of fictitious job applicants based on their responses to a personality inventory. In Experiment 1, the information on personality was held constant, such that the applicants' responses varied solely on their agentic, communal, or purely descriptive orientation. In Experiment 2, the social value of the responses again varied as well as the information about the applicants' personality (agreeable vs. conscientious). The results showed that the agentic profiles were the most predictive of the performance, regardless of the personality factors. In Experiment 3, we reversed the procedure. The participants filled out a personality inventory in the place of a more or less successful employee. The results here showed that the information about the performance had the greatest impact on the agentic items, independent of the personality factors measured. These results confirm the relevance of social judgment models in personality research. Copyright © 2014 John Wiley & Sons, Ltd.