Don Thompson on prediction markets in political elections and business decisions

Who will win the election? Prediction markets are more accurate than the polls for prediction of the election results, according to Donald N. Thompson, an economist and professor of marketing, emeritus, at York University in Toronto, interviewed in this podcast: Let Your Employees Bet on the Company - HBR IdeaCast - Harvard Business Review (audio, 14 min).

"Dewey Defeats Truman", a famously inaccurate headline, Chicago Tribune,1948, from archives.gov

I found these 2 prediction markets for the US elections: Intrade presidential and senate race, and IEM presidential race.

Intrade prediction market: Obama vs. Romney

Note the large percentage differences between candidates in the predictions markets. This reflects the fact that these percentages are for predicting the election outcome, which can be virtually certain in some cases even if the poll results are close. For example, if candidate A is leading candidate B 53% to 47%, with a 1.5% margin of error, then the probability of candidate A winning is much higher than 53%.

IEM prediction market: Obama vs. Romney

I have not seen the supporting data yet, but according to the book Oracles: How Prediction Markets Turn Employees into Visionaries by Don Thompson (Chapter 4, p. 47), IEM prediction market was more accurate in predicting elections than than the polls 74% of the time, based on a total of 964 polls, conducted 1988-2004. This seems surprisingly more accurate. As they say, "prediction is very difficult, especially about the future".

[EDIT January 3, 2021]

After the widely publicized incorrect prediction of the US Presidential election results in 2016, prediction markets delivered correct predictions in 2020.