Economic theory suggests that performance pay may serve as an effective screening device to attract productive agents. The existing evidence on the self-selection of agents is largely limited to job tasks where performance is driven by routine, well-defined procedures. This study presents evidence for a creative task and studies how agents self-select into a tournament-based scheme vs. a fixed pay scheme. The experiment allows for the measurement of creative productivity, risk preferences, self-assessments, gender, and other socio-economic characteristics such as the Big Five personality traits. Results show that the two payment schemes systematically attract agents with different characteristics. However, results differ fundamentally from previously found patterns. Agents did not self-select into the tournament scheme according to their creative productivity, but only according to their risk attitudes and self-assessments. The reason for the absence of a selection of the most creative agents into the tournament is that there exist substantial misjudgments of relative creative productivity. Further evidence from a representative German survey data set provides additional support for the experimental results suggesting external validity.