PREDICTION MARKETS VS. POLITICAL POLLS: FORECASTING ELECTION OUTCOMES

Abstract

This paper focuses on the race 2020 democratic presidential nomination and analyzes the effectiveness of polling scores and prediction market contract prices when responding to micro events and f orecasting the overall outcome of the nomination. Thirteen “mi cro events” are highlighted to track important milestones in the race for the 2020 democratic presidential nomination. When analyzing the effectiveness of polling scores and prediction market contract prices predictions, timeliness and accuracy are evaluat ed. There is evidence that changes in prediction market prices are positively correlated with changes in polling scores. There is also evidence that prediction market prices are influenced by potential reference prices which are a potential bias in their e ffectiveness of predicting future outcomes. Despite the potential for bias, this research supports prior claims that prediction market prices react more quickly to new information, in this case micro events, when compared to the associated polling scores. Overall, both prediction market contract prices and polling scores both sources accurately predicted the winner of the nomination within the last few days of the race. But the prediction market contract price data more quickly responded to key milestones allowed the market to predict a winner with larger during the race for the nomination which margins when compared to the polling scores.