New at Reason
Katherine Mangu-Ward looks at the Election 2006 shortcomings of the online futures markets.
Comments to "New at Reason":
I'm not so convinced the markets did that bad of a job. The markets were trading at about a 70% likelihood that GOPers would keep the Senate. That means that if we saw 10 markets with the same price, we would expect 3 of them to be "wrong" (i.e., the lower-probability outcome occurs). Now if InTrade was giving a 99% chance of a GOP Senate . . .
kebko | November 13, 2006, 3:39pm | #
Am I correct that the markets called all the individual contests correctly, but just missed on the aggregate?If that's the case, it seems like the markets just didn't have enough faith in their own accuracy. If arbatrage traders applied 100% certainty to each individual prediction, they would have been pushing the aggregate markets to favor the democrats, but because they applied some discount to each individual prediction, and democrats had to win every one in order to get control, then the aggregate prediction was that they would fall short. Maybe the lesson next time should be that the traders themselves should have more confidence in their individual predictions.
thoreau | November 13, 2006, 3:54pm | #
KMW, you say in the article that you were talking all about the markets before the election, but I don't recall any articles or blog posts on that topic. Or was that something you wrote elsewhere?Just curious.
One thing about the markets is that calling "almost all" of the races correctly isn't terribly relevant to predicting who will get a majority. The fact is that incumbents almost always win, and by significant margins. What really counts is whether the markets, or pundits, or any other source of predictions can do better than a coin toss on the handful of close races.
Brian Courts | November 13, 2006, 4:11pm | #
Tradesports did indeed get each individual senate race correct as of the night before the election - not a bad record. They were not clearly wrong in the aggregate either, since the mere fact that the Dems won the senate does not prove that, say 70%, was a bad estimate of the odds that they wouldn't. After all, had the GOP managed to win even one of VA, MT, MO, or MD races that one pundit site called "toss-ups" they would have kept control. In other words, if you were flipping 3 60% heads-weighted coins and Tradesports had each individual coin toss contract valued at 60 for heads and the aggregate all three heads contract priced at 21.6, would anyone claim the markets did a bad job when, after the fact, it turns out all three came up heads? Assessing the betting market's performance is going to take more than noticing that the Dems won the Senate despite the GOP price advantage.Paul | November 13, 2006, 7:20pm | #
Feh, so some short sellers made money. Moving on now...Futures markets are futures markets. They're bets. Bets go wrong. I envision a much more accurate poll. One where every person who wants to vote voluntarily picks his choices by marking them on some kind of form, be it electronic or paper. At the end of this poll, a bi-partisan institution counts those choices using the best methods available. Ah, to dream.
db_cooper | November 13, 2006, 7:37pm | #
McLaughlin![super fast]An additional benefit of being a McLaughlin Group viewer[/super fast] is that you get to lead a suuuurpriiiise freee exxiisstance.
[super fast]If you'll cast your mind back 6 weeks when I predicted that Vicente Fox would become president of Mexico ... [/super fast]
Cannot wait for this weeks show to hit the web.
Again | November 13, 2006, 8:47pm | #
Kebko,The problem is that it was difficult to determine to what extent the outcomes of those races were and were not independent of one another.
So I guess the range was one of complete independence (which is almost certainly wrong) and at the other end one where there were a strong link between the results of one race and another.
I dunno it's a tough problem to approach mathematically and I think the futures folks winged it for the most part and had the races as more independent than they actually were.
Allen, Talent, Burns and Chafee are all a relatively diverse set of Republicans and they all seemed to be done in by the same thing. Now Allen and Burns just _barely_ lost so I'm not sure there's any reason to go overboard on the markets.
Shem | November 13, 2006, 9:46pm | #
I envision a much more accurate poll. One where every person who wants to vote voluntarily picks his choices by marking them on some kind of form, be it electronic or paper. At the end of this poll, a bi-partisan institution counts those choices using the best methods available.You mean...voting?
Actually the arb guys got it right they had odds of winning of 70% in Montana, 50% in Missouri, 60% in VA, 70% in Maryland and 20% in Tenn. They needed 4 out of 5 for control. I don't feel like running the probability of a Dem victory, but 30% wouldn't surprise me. In this case the prediction market looked odd because people were able to "strip" it into its core components and trade the odds of each one separately.
mw | November 14, 2006, 10:30am | #
I was watching the Tradesports contracts pretty closely, and was also confused by the behavior of the GOP CONTROL SENATE contract. No matter what happened, it seemed to be nailed at 70% down the stretch. Prior to the election, I commented on the discrepancy between the Trade Contracts and the "100 year rule" on my blog in this post: Fifty-Fifty As Divided as it gets."That said, it is worth pointing out that the Tradesports GOP Control Senate contract may yet prove to be prescient, in a literal sense even if not for the bettors who went long:
Should Joe Lieberman decide to caucus with the Republicans, they would retain "control" of the Senate based on Cheney's tie-breaking vote.
There is another possibility, which falls in the tin-foil hat category. Let us say a canny Republican strategist determined that it was critical to maintain an optimistic view of the the Republican's prospects down the stretch in the mid-terms in order to ensure a large GOTV effort. Let us say that this strategist was also keenly aware that many well know pundits (Kudlow - for example) tracked and commented on the betting contracts. Might it not be worth spending some of hundred of millions raised in this election on propping up the GOP CONTROL Senate contract?
This is not the Chicago Mercantile Exchange we are talking about. This market is not protected by SEC regulations. It is am off-shore betting market of questionable legality in the US. I have not looked in detail at how much money was involved over the life of that contract, but I doubt that it would take much more that cost of a well placed TV ad to have a significant impact on the price.
mw | November 14, 2006, 12:05pm | #
PS - Just saw this "Lieberman won't rule out GOP caucusing"Perhaps that agggregate Senate trading contract will be proven right after all.
Brian Courts | November 14, 2006, 12:46pm | #
Perhaps that agggregate Senate trading contract will be proven right after all.But again, there is no reason so suspect it was wrong based on one event, and one event isn't going to prove it right either.