Politics and Public Opinion

Yes, Intrade actually is that useful!

Jonah questions the usefulness of Intrade, doubting whether the “wisdom of crowds” actually generates much predictive value when it comes to political prognostication. Don’t betting markets just follow the consensus, which generally follows the poll? Well, a 2010 study looked at that very issue:

In keeping with similar research by Berg et al and others, this dissertation found that market prices tend to be better predictors of political event outcomes than corresponding polls. This superiority was observed over all time periods studied, with Intrade market prices correctly predicting the winner 36% of the time in the 2004 Democratic presidential nomination contest (compared to 13% for polls) and 54% of the time in the 2008 Democratic presidential nomination contest (compared to 45% for polls).

And this past October, the Washington Post took a look at Intrade and came to this conclusion:

For example, in the 2004 election, Intrade was predicting that George W. Bush would win by August. On the other hand, the market did not see Hillary Clinton’s New Hampshire victory in the 2008 primaries coming. Of course, no one else—including Clinton—did either.

And, Intrade has successfully predicted events for which there are no polls—for example, that former North Carolina Sen. John Edwards would be the 2004 Democratic vice-presidential nominee and that Donald Rumsfeld would resign as secretary of defense in 2006. (Politics aside, Intrade has shown a knack for predicting Oscar winners.)

And how about Intrade versus the pundits in 2008:

On the morning of Election Day, I printed out the expectations from the Dublin-based Intrade market as well as a roundup of predictions from nearly two dozen political consultants, journalists and academics that appeared at the Huffington Post. The Intrade bettors expected Mr. Obama to end up with 364 votes in the Electoral College — one less than he actually got. None of the pundits came so close. Alan Abramowitz, a political scientist at Emory, came closest with prediction of 361; all the rest were off by at least 12 votes. Nate Silver, the much-talked-about statistician at FiveThirtyEight.com, underestimated Mr. Obama’s tally by 18 votes. Many of the pundits underestimated Mr. Obama’s total by more than 25 votes, like Chris Matthews, Arianna Huffington, and the strategists Paul Begala, James Carville and Alex Castellanos.

I could go on and on. A 2004 study concluded “Prediction markets are remarkably accurate information aggregation mechanisms.” They are not perfect, of course. But I was made a believer in 2004 when the exit polls said Bush was a sure loser and Intrade gave a different story.

Believe, Jonah, believe!

3 thoughts on “Yes, Intrade actually is that useful!

  1. Isn’t the difference partly that polls are for measuring what people want to happen, and intrade isn’t necessarily about what people want but what most people think is going to happen. In respect of having predictive value it might be more accurate, but it’s essentially testing a completely different thing. On the other hand, it shows the value of putting some “skin in the game” instead of just offering an opinion when asked. Someone should be able to manipulate this into a lesson about capitalism or something…

  2. If I understand the numbers given, the logical conclusion is that – at best – Intrade, et al are right about half the time. That may be better than anyone else but it seems of limited value to those not betting but interested in information.

  3. We are all familiar with “push-polling,” which can be sculpted to produce a desired result by skewing the sample and framing the questions.

    I posit that a similar result can be had by loading, or over-betting, if you will, prediction markets. Depending on the size of the betting pool, a determined campaign operative might be able to force a desired upward trend in the odds for his candidate, and lower the odds for his opponent, by the simple expedient of “flooding the zone” — pumping a relatively large sum of money into the mix. If the pool is small, this maneuver would be easy and the market might thus be co-opted to reflect a favorable trend for a particular candidate.

    In the stock market, similar outcomes are made to happen by fund managers who want to pump up stock prices just before reporting quarterly results. In the case of prediction markets, a campaign manager seeking to galvanize a particular trend would not be concerned with losing money (it’s not his own money in any case) but rather with boosting his candidate’s profile with voters. Might even be cheaper than direct mail and TV.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>