If the poll is 51-49 and people never change their mind and everyone is honest, the prediction market would settle close to 100-0.
Discussion
I see your point but obviously we don't live in that ideal scenario.
Yes of course that's not realistic, it's designed only to point out the difference in an obvious way.
In the real world, if the polls are say 60-40 for an extended period close to the election, the pred market is likely to be 95-5 or similar, *if* the market finds the polls trustworthy. It is not going to be anywhere near 60-40. Again, they are not measuring the same thing!
Right, and polling is designed (IMO) to perpetually make the race look close to keep people engaged.
Well, a poll doesn't *have* to be close. Also is it bad if people are engaged? I guess that heavily depends on your worldview :)
Engagement is good but deceptive polling can lead people to believe there's broad support for otherwise unpopular policies. IMO this is why people distrust polls, and thus the prediction markets struggle to predict the outcome with >X% certainty.
Ah, I see now the emphasis you were going for in your original note. Polls estimate voter sentiment, prediction markets estimate odds of winning.
Yeah to be fair there is something implicit, i.e. that the prediction is 'who would be president', or more generally, a binary outcome, not 'what will be the final distribution of votes as a percentage'. I don't know how prediction markets for outcomes like that work, it's obviously more complicated then.