NYT Overstates Prediction Markets Efficacy

The NYT David Leonhardt overstates the case for Prediction Markets in Monday’s paper:

"A kind of futures exchange where participants bet on real-world events,
these markets could have told somebody last fall to ignore the talk of a late
surge by John Kerry and stick with President Bush
. Based on the odds posted at
InTrade, one of the biggest exchanges, you could have correctly forecast the
presidential winner of all 50 states, and 33 of 34 Senate races." (emphasis added)

That’s simply a false statement — actually, these markets paniced and got it wrong (at least for a few hours). As we saw right after early exit polls in Florida and Ohio gave Kerry a slight edge, these markets actually raced to bet on that surge — Kerry Futures spiked as the Bush’s collapsed in response to exit polls. As Dan Gross so clearly elucidated:

“the political futures traders freaked out and rushed to dump Bush and
buy Kerry. By 4:30 p.m. ET, IEM’s Kerry winner-takes-all contract had
risen from below 50 in the morning to more than 70. The vote-share
contracts on IEM also shifted in Kerry’s favor. TradeSports’ traders,
who had shown Bush winning for the entire campaign, suddenly bid the
Kerry election contract up to 67—meaning they thought he had a
two-thirds chance at winning. By 5:44 p.m. ET, TradeSports gave Bush
just a 39 percent chance of winning Ohio and a 43 percent chance at
winning Florida. . . "

"The furious and seemingly irrational Election Day market action stands
as evidence that the traders are more poll-followers than poll-beaters.”

As we have discussed ad nauseum, enthusiasm for prediction markets should be tempered with a dose of reality. They are hardly infallible, as we saw in the Jackson and Purcell cases, as well as Howard Dean’s Iowa Primary, as well as the pre-Election and Election day trading frenzy.

As we’ve seen, Electronic futures traders don’t so much predict political events as they do track all the various polls, efficiently synthesizing the entire spectrum of them into a coherent expectation.

As we have said too many times, one does not derive wisdom from collective ignorance; Large liquid markets are most predictive when making probablistic detemrinations about the behavior of that same crowd itself.


Online Bettors Find a New Love: Real Estate
NYTimes, August 22, 2005

Bettors for Bush
Daniel Gross
Slate, Wednesday, Nov. 3, 2004, at 12:15 PM PT

Print Friendly, PDF & Email

What's been said:

Discussions found on the web:
  1. mh497 commented on Aug 22

    Soros was trying a little last minute alchemy there on Intrade, wasn’t he?

    (BR notes: The Ignorance was as originally posted; It is unedited by me — Do we now have our very own trolls? Cool!)

  2. Anonymous commented on Aug 22

    Don’t know about Intrade or Iowa Markets, but I traded a few election contracts on World Sports Exchange before and during Election Day. I can confirm for a fact that the Bush contract was trading at 33 on the *ask* side (i.e. bid 28 ask 33) at 3:49pm PST on Election Night (2-Nov-2004). This means the Kerry contract was trading at bid 67 ask 72 at the same moment.

    I bought additional Bush contracts at $33, which was the intraday low or very close to it. In fact, they were so low that I was instantly second-guessing myself, thinking someone had to know something I didn’t.

    By 6:30pm PST, the Bush contracts had risen to 45 50. This means Kerry was still the favorite at 50 55.

    At 6:34pm PST, the Bush contracts were still 46 50. But at that moment on CNN, Bush was leading the electoral count 170-112.

    At 6:37pm PST, the two contracts were a dead heat (both 48 52). CNN was still calling it 170-112.

    At 6:57pm PST, CNN was still stuck at 170-112, but by then Bush contracts were 61 65.

    At 7:27pm PST, the Bush contracts were 70 74, and all the major news networks had the Electoral count at 193-112.

    We all know how the story ends. But the fact is that on one exchange at least (WSEX), Bush contracts ranged from 33 – 50 on the ask side for at least 2 hrs and 45 minutes on Election Day 2004. It is simply incorrect for the media to revise this history after the fact.

  3. JD commented on Aug 23

    I used TradeSports “recreationally” leading up to election day last year as a way of tracking the polls (i.e. “synthesizing the entire spectrum”) and getting some market-based insight as to how the election would actually unfold.

    You are correct, it could be argued that the futures markets flubbed the end game by misreading the exit polls. I think, though, that if you look a little closer you’ll find that they didn’t perform as poorly as your analysis implies.


    Leading up to Election Day (i.e. prior to initial exit polling data being released), TradeSports was correctly predicting a Bush win. (I’ve lost the Excel sheet I was using to track this so I don’t have state-by-state figures — i.e. I’m not sure how accurate it was for each state’s electoral counts.) It was on Election Day, when the traders in these markets caught wind of exit poll data, that major shifts and fluctuations in trading prices occurred and they started calling for Kerry.

    But is it fair to criticize the futures markets for incorporating historically accurate exit poll data into their predictions? How would the bond and stock markets react if certain members of the Fed, prior to the Fed’s usual 2 p.m. announcement of interest rate changes, anonymously shared their thoughts with members of the press and the results of these interviews were leaked to the public? I’m sure these rumors would wildly roil the markets if the “poll data” pointed to a decision other than what was expected. Naturally, the markets would correct themselves once the facts were clarified in the actual announcement. The question is whether it would be accurate to say the markets “got it wrong” simply because they responded to these rumors.

    So, were the futures markets “wrong” about the election? One answer is, yes and no. A better answer is that at the end of the night, traders holding Kerry wins were “wrong” and traders holding Bush wins were “right.” But what about the market itself? The market is not an individual entity capable of being right or wrong. All the market can do is summarize the guesses its traders are making in probability terms.

    If we take, for example, the snapshot of the Ohio contract at 5:49 p.m. as described by Mr. Gross in the quote above, the market was “saying” in essence: George Bush has a 39% chance of winning Ohio, John Kerry has a 61% chance. This is a probability statement. In plain English, it’s a way of saying that if we could repeatedly run the election event again and again, then at that exact moment in time with the exact same set of information at our disposal, the final result would be that George Bush wins Ohio 39% of the time and John Kerry wins it 61% of the time. Sounds like a pretty reasonable statement to me. After all, if you had to make similar bets in future elections, how comfortable would you feel betting against the exit polls at 5:50 p.m. on election night? I certainly wouldn’t be.

    In my opinion, calling the futures markets “wrong” — even on Election night — is like complaining the weatherman is wrong when he calls for a 70% chance of showers on Saturday and not a drop of rain falls. One cannot assess the accuracy of such a prediction in a single instance; all one can do is to look at the accuracy of such predictions across a range of similar events. (I’ve never seen such data for future markets, so I can’t speak to the overall track record of these markets at making election predictions. My point is only that proof of their accuracy lies in that track record.)

    Personally, my concern for the futures markets is not their inability to make accurate predictions, but rather the potential for traders with political rather than financial (i.e. money-making) agendas to unduly influence these markets using relatively small amounts of capital. These markets only work when everyone’s in it for the money.

Posted Under