Post

Yes. Here's one example: www.sciencedirect.com/science/arti... Non-paywalled version here: www.nber.org/system/files...

Interestingly, Manski shows there are problems with the conventional interpretation even if you assume risk neutrality.

I’ve often wondered about this. I’ve never bothered to sit down and really understand the microeconomics of prediction markets; I just get the vibe and assume that there is some model where perfect competition pushes the price to the actual expectation or something.

I should add, without under any circumstances “handing it to” Polymarket, that a high win probability for Bolsonaro is not incompatible with a narrow Lula victory. This exact thing is what drove Nate Silver to madness.

the model said he had a 75% chance to win, so it should have been a slam dunk. that's a 3-to-1 majority

A 75% chance of winning means that if you ran the election 100 times it would result in a loss 25% of the time. Not really sure anyone would consider that a "slam dunk".

People have done overall analyses of whether or not prediction markets are accurate, and they generally are.

Not really what I’m asking!

I think research into the mechanism by which prediction markets are inaccurate would be quite contingent on prediction markets being inaccurate.