When are Experts Wiser than the Crowd?

This paper replicates the study of Prelec, Seung and McCoy (2013) and extents it with a more complex question. Their expert (LST) answer is based on how well individuals predict other people's answers. In their study, it outperforms the wisdom of the crowd (group) answer on questions about US state capitals. In mine, the LST and group answer perform similarly. Besides, for a more complex question (based on a bean-jar experiment), the LST answer underperforms the group answer. This corresponds to literature that finds less expertise with more complexity. Subsequently, I identify experts with another method to control for expertise and test whether the performance difference is the result of the LST answer's failure to properly identify experts (based on experts' ability to predict other people's answers). This is not the case. In fact, experts seem more able to predict other people's answers for the more complex questions. I confirm this finding by directly comparing the predictions of subjects I identified as experts and non-experts.