Wisdom of Crowd: Comparison of the CWM, Simple Average and Surprisingly Popular Answer Method

On a daily basis people and institutions are forced to make optimal choices, provide correct solutions to many times very complex tasks and predict future outcomes. Since individual answers tend to suffer from random errors in judgement, it is frequently better to rely on mathematical combination of people's opinions in order to make the best decision. This approach, which is academically called the wisdom of crowd, lowers the random error and delivers an accurate answer that is most of the times better than any individual guess. In this thesis I compare the performance of three aggregation methods withing the crowd wisdom concept: a simple average (SA), a surprisingly popular answer algorithm (SPA) and a contribution weighted model (CWM). Besides that I also conduct a supplementary research including the BTS score performance comparison, regression of the CWM score on the BTS score and an analysis of cheating behavior. By means of paper and online quiz about general knowledge questions I collect data, compare the outcomes of the models and conduct several robustness checks. The main findings of the paper show that both the SPA and the CWM perform significantly better than the SA for large samples, however, the results are more robust for the SPA/SA case. Moreover, there is no difference in the performance of the SPA and the CWM. Therefore, I conclude that the SPA is the most appropriate for general use as it has less requirements regarding the type and the number of the questions than the CWM and its good performance is more robust. Finally, further research should be focused on better incentivization of subjects so that more questions can be used for the analysis.