Discussion about this post

User's avatar
Overcoming Bias Commenter's avatar

Strange coincidence: I wrote the previous comment yesterday and then I opened a book I had received in the mail earlier that day, and it happens to discuss exactly the issue I raised: polls and averaging versus prediction markets. The book is The Difference, by Scott E. Page, and in chapter 8 he goes into when and why prediction markets would be expected to do better than simple averaging. The results were quite surprising to me, and I need to try harder to understand them. I hope to post a summary soon.

Expand full comment
Overcoming Bias Commenter's avatar

I am curious about studies that compare betting markets to other forms of information aggregation, in particular straightforward polling and averaging. I have been hunting through Gallup poll results and other places to try to find data about public predictions that could be compared with existing markets, either financial markets or prediction markets. So far I haven't found any good examples. Gallup does not very often ask people to make predictions; they more often ask how people think and feel about issues. They ask who are you going to vote for rather than who do you think will win the election.

I did find some time series of predictions about whether the economy is going to get better or worse, but I can't think of a futures market to compare that to.

Betting markets have the property that people who have stronger opinions and/or more money play a larger role. The question is how well correlated this is to correctness. Betting markets also encourage honesty, but for many issues an anonymous poll won't give people much incentive to lie. Betting markets reward discovery of new information, which polls do not, but it's not clear that that will by itself cause more accurate forecasts. Overall it is not clear to me how much better in practice a betting market will be than a simple poll.

Expand full comment
16 more comments...

No posts