2020 election: Prediction markets versus polling/modeling assessment and postmortem

Abstract

The author of the article analyzes the performance of prediction markets in forecasting the 2020 US presidential election, comparing them to polling data and statistical modeling. They argue that the markets performed poorly, particularly in the days leading up to and following the election, making overly confident predictions and failing to adjust their probabilities in the face of new information. The author concludes that Nate Silver’s model, based on polling data and statistical modeling, outperformed the prediction markets in almost every metric, except in predicting the final margin of victory. The author attributes the market’s poor performance to a number of factors, including the influence of partisan betting and the difficulty of accurately modeling the impact of factors like voter suppression and electoral fraud. – AI-generated abstract.