(Inaccurate) Beliefs about Skill Decay

Abstract

For many skills, performance declines without practice, and accurate beliefs about the rate of this decline are important for a range of human capital investment decisions. Across five preregistered experiments (N=3,016) and three distinct tasks, we examine the accuracy of beliefs about one's own and others' skill decay. In the first three experiments, participants consistently underestimated their own skill decay, by 28% to 59% depending on task type, with little variation by break length. In a fourth experiment we study whether people's accuracy improves after skill decay has actually occurred, by eliciting predictions both before and after a two-week break. Although predictions become 60% more accurate, underestimation persists. In the fifth experiment, we randomized participants to different levels of task experience and elicited beliefs about others' skill decay. While participants had more accurate beliefs about others' skill decay, they still underestimated decay, on average, and task experience did not significantly affect predictions. Finally, leveraging methods from machine learning, we find robust evidence that age is an important factor in the underestimation of skill decay: belief errors increase with age because, while older participants exhibit greater declines in their skill level, their predictions fail to account for this difference. Taken together, our findings reveal a consistent underestimation of skill decay, suggesting a potential for errors in human capital investment decisions when making choices for oneself and for others.

Keywords: learning, beliefs, skill decay

JEL Classification: C91, D83, D91