Can election results be predicted? The researcher who examined a million bets and is willing to bet himself
Prof. Michael Gilead from Tel Aviv University published a study that examined nearly a million contracts in prediction markets and found that they are significantly accurate. However, he warns that in Israel, due to low trading volumes, they may not be superior to polls, and that close to elections they are easy to manipulate.
Prof. Michael Gilead from Tel Aviv University published a new study in the journal PNAS, together with doctoral students Hila Shinitzky and Maayan Pa'ar, which examined nearly a million contracts in prediction markets such as Polymarket. The study found that the predictions in these markets are significantly more accurate than a coin toss, and that they are well-calibrated – meaning, when the market estimates a 30% chance for an event, the event indeed occurs 30% of the time. However, Gilead emphasizes that in Israel it is not certain that prediction markets are superior to polls, partly due to lower trading volumes. He specifically warns against relying on prediction markets close to elections, since in relatively small markets it is easier to influence prices and create a false impression of political momentum. The study compared the accuracy of the markets to additional benchmarks, and in all cases the results were impressive.