Michael Giberson
The IOC recently selected Rio de Janerio over three competing bids to host the 2016 summer Olympic games. The Chicago bid was favored in public prediction markets, with prices at Intrade between 50 and 60 at the time of decision and prices at Betfair implying about a 50 percent chance. Did the prediction markets fail to predict well?
At Midas Oracle, Chris Masse has been asserting that prediction markets for IOC selections are fundamentally flawed, saying that the IOC is a small, secretive committee that doesn’t leak information and therefore no information is “out there” available to be aggregated by a prediction market. He was saying this before the IOC vote, too; this is not just after-the-fact speculation, it was his before-the-fact speculation. (Also posts here, here, here, here, here, here and, from April 2007, this post.)
I think the “small, secretive committee” explanation is weak, so I’ve been poking back a little in the comments. Chris, as is his style, has been elevating my comments into new posts in order to re-assert his views.
But a more fundamental question is whether or not it can be said that the prediction markets got it wrong. At Sabernomics, J.C. Bradbury reports watching Intrade closely the morning of the IOC decision:
Around 9 AM … the odds show Chicago to be the favorite with a 53% chance of winning, closely followed by Rio at 46%, Tokyo at 3%, and Madrid at 2%. Like all the pundits following the selection were saying, it was a race between Chicago and Rio, but was very close to call. These odds also show something else, Chicago was trending down and Rio was trending up. The trend would continue for the next few hours.
… Looks like useful information was leaking out from knowledgeable parties just before the vote. This is evidence for, not against, the strong-form of efficient markets hypothesis.
Bradbury does an excellent job sifting through the shifting coalitions revealed in the three rounds of IOC voting. Neither Madrid nor Toyko showed any significant ability to attract votes as the rounds proceeded. It was going to be Rio or Chicago all along, but Chicago was weakest in the four-way vote and lost early, leaving the games to go to Brazil.
Based on Bradbury’s analyis, I’m convinced that the decision was pretty much a toss up between Chicago and Rio. That conclusion was also implied in the prediction market prices just before the decision. Sure, the prediction markets favored Chicago, slightly, over Rio; I don’t think you can call it a miss given the closeness of the decision.
[Related: Market Design and Marginal Revolution both have brief notes; Infectious Greed provides related discussion.]
UPDATE: Chris Masse doesn’t like my analysis: Who has the best analysis for Chicago’s failed bid for the Olympics?; neither does Paul Hewitt: “Michael Giberson is wrong to imply that the prediction was accurate on the basis that Chicago and Rio were fairly close.” See also here.
Pingback: | Midas Oracle .ORG
If it was a toss up between Chicago and Rio, why was Chicago eliminated first? even before Madrid and Tokyo?
Best Regards
Seems like a fools errand to argue if the market some how got it wrong. Even with risk neutral investors and an efficient market Madrid should have won 2% of the time. With unique events like this, how can you ever infer from ex-post outcomes what the proper ex-ante probabilities were.
Every time an option expires worthless that isn’t an indictment of markets, that a recognition that something can have a positive price even if it has no value in many states of the world.
Amschel: I think it reasonable to believe that Toyko and Madrid were stuck at 1/4 of the voters max, making it impossible for either to win. If Rio is eliminated in the first round instead of Chicago, the Rio votes likely go to Chicago and so Madrid or Toyko is eliminated. Assume Tokyo eliminated as in actual second round. The Toyko voters don’t shift to Madrid, so Madrid is eliminated as in actual third round. Chicago remains.
Speculative, sure, but some support in the voting record.
OneEyedMan: in principle you are right, but you can compare a series of markets against outcomes and if the prices are good, then things with prices around 10 should happen 10 percent of the time, etc. Hard to judge with relatively few outcomes to judge against.
Maybe they really did gag on The One and his Consort.
Hi, Mike.
I have to agree with you here, and particularly agree with the Sabernomics post. It appears from the voting patterns that Rio and Chicago were the strongest competitors, which was reflected in the percentages.
However, I do agree with Chris that the overall quality of a prediction market where information and decisions are made in a close-hold group will likely be low. I would include IOC voting, Supreme Court nominations (but not confirmations), cabinet secretary nominations (but not confirmations), etc. As with all probabilistic markets we need more observations to properly evaluate success, and by their nature these types are few and far between.
Fundamentally, these markets are inherently risky. But the market with a 45% chance of winning won. So Chris is (once again) making a mountain out of a molehill to drive traffic.
Jed: So I take it you’ll not be betting on the Nobel prize announcements, and I’d have to agree that for Nobel prizes none of the betting houses or prediction markets are much use. The potential numbers of answers is large and rather undefined – which of the thousands of writers will be selected for the literature prize? – unlike, maybe, the Booker prize with a shortlist to work with, the Nobel prizes are too open ended.
I’m not convinced that IOC falls into this same category. The group is large and the membership is public – over 100 members from all over the world. They are making a highly public and newsworthy decision, from an announced shortlist of possibilities.
Mike: There is most definitely a spectrum. Nobel prizes are far to the side that are almost totally inscrutable. IOC and Supreme Court nominations somewhere less so.
As you mentioned, a shortlist helps, a larger voting pool helps, visibility in the process and criteria helps, and the potential for leaks helps. Essentially we just don’t know enough about all this to make accurate assessments.
Re: “we just don’t know enough about all this”
I agree that “we just don’t know enough about all this to make accurate assessments,” and so I hope the prediction markets keep offering these kinds of contracts and we have a chance to learn.