Is Intrade Really a Valuable Predictor?

Intrade is an online trading platform where participants actually place (legal) bets on the outcomes of certain events. For close to a decade political pundits have been using it as a reference to predict election outcomes. Indeed it seems to have a good record, correctly predicting the outcomes of the 2004 and 2008 presidential elections, and getting all but two states correct in 2008. Currently, Intrade gives Barack Obama a 62.4% chance to win re-election.

So is Intrade a valuable resource that can be relied upon to accurately predict election outcomes? Not in the least.

This Business Insider article sums up several of the problems with Intrade, and hits upon the point that has bugged me the most about it, namely that all it does is distill current conventional wisdom. Take, for example, that 62.4% number above. Sure that looks good for Obama, but over a week ago that number was well over 80%. In other words, as Obama’s poll numbers moved down so did confidence by Intrade investors. As Joe Weisenthal put in when discussing the Republican primary:

So why ignore InTrade? Well, basically, because all it does is distill conventional wisdom. Seriously, what good is it to know that on InTrade Mitt Romney is far ahead, and that Hermain Cain doesn’t have a chance? All you have to do is read any DC-based political pundit, and they’ll tell you the exact same thing.

And when the conventional wisdom changes, so does the market.

Rick Perry is down in the dumps on InTrade now, but back in August — when everyone was talking about how he was the frontrunner — he was the frontrunner on InTrade as well.

Weisenthal then tracks Perry’s chances on InTrade, and notes how they basically just mirror Perry’s poll numbers.

Even the 2004 and 2008 results aren’t that impressive in retrospect. When people woke up on election day 2008, did anybody really doubt that Barack Obama would win, other than people who clung to fleeting hopes of a miracle McCain victory? And in 2004, Bush’s chances were just over 50% – meaning that the market as a collective was leaning the same way as most polls which, with a few exceptions, generally gave Bush a slight edge. In fact, if you look at Intrade activity on election day itself, Bush’s chances plummeted as early exit poll leaks suggested a Kerry victory, and then rose again as actual election results came in and a Bush victory became more apparent. In other words, Intrade just reflected the polls. And while the state predictions seem impressive, again, how many states were truly up for grabs? Intrade was therefore no more useful a guide than any reasonably informed individual with access to polling data.

Some fans of Intrade like to point out that participants literally have to put their money where their mouth is. I don’t really see how this makes the platform any more valuable as an index. Bookies all over the country would be the ones fearing having their legs broken if money induced wiser gambling behavior – and Intrade is, in essence, simply a gambling platform.

Long story short, Intrade offers no more insight into how the election will play out than some cranky guy writing on a blog who can look at the Real Clear Politics average of polls (which has Romney up by 1.3 percent, incidentally). So then why do pundits insist on citing it, and why do people continue to think it has any meaningful predictive value?


Share With Friends

Dante alighieri


  1. The question to ask is not whether Intrade is accurate, but whether it is more accurate than the alternatives. I just finished Nate Silver’s new book. In one of the early chapters he looks at the accuracy of predictions made by TV pundits. In most cases a pundit’s predictions turn out to be less accurate than random guessing. Intrade isn’t perfect, but the reality is that it’s a lot more accurate than that.

  2. After the election results are in it will be interesting to analyze why it provided relatively good or relatively bad predictive results for this particular case.

  3. Certainly, it’s basically distilled conventional wisdom, but when you’re dealing with moderately knowable events (or “given the way things are going” kind of predictions) it might be very slightly better than just looking at the polling data.

    So, for instance, on the question of prediction the presidential election: Right now the polls show Romney ahead, though there needs to be a little more state polling post debate before it’s clear whether he’s ahead enough in the right places to win the electoral college. InTrade still has Obama as the favorite, though by much less than before. If I was told to make the most conservative bet possible with a large amount of money, I might put the odds about where InTrade is right now. My reasoning would be that the Obama campaign has been fairly ruthlessly efficient at staying on top of things throughout the campaign, while the Romney campaign has been prone to unforced errors. Current polls suggest Romney slightly ahead, and a certainly hope that that’s because Romney has finally shattered the characterization of him that Obama has been running against so successfully. But if I really, really had to place money right now (and defend it to others rather than just pleasing myself) I’d probably put it on the assumption that Obama would edge ahead in a close election through constant negativity and Romney’s occasional errors and lackluster campaigning.

  4. The perils of something like Intrade are that you have a lot of people who have more money than knowledge about politics. This election will not be close and the result will be circa Romney 54 to Obama 45 with one percent split among the third party candidates. Why am I confident of this:

    1. Wretched economy.
    2. Drops in Democrat registration in key states.
    3. The enthusiasm gap between Republicans and Democrats.
    4. The results of 2010 which placed Republicans in control of more state houses and legislatures than since the 1920s. They put in place voter id laws in more than 30 states. The margin of fraud advantage that the Democrats have traditionally had is rapidly diminishing as a result, tied in with purging voter rolls of the dead, convicted felons, people who moved away decades ago,etc.
    5. Romney and the Republicans are on a par with Obama and the Democrats when it comes to campaign spending.
    6. Too many of the polls relied on by the rubes at Intrade have ludicrous oversampling of Democrats.
    7. Obama has a glass jaw as evidenced by the first debate.
    8. Republicans have a party id advantage over the Democrats. Prior to the 2010 landslide they were down 1.6 against the Democrats in Rasmussen party id.
    9. Obama’s war on the Church. Lots of Catholics in the swing states and Obama is going to be lucky to have his total among Catholics match the 45% I predict for his national total.
    10. Libya gate.

  5. To elaborate on my last comment a bit: people will often dismiss Intrade by saying that “it just tracks the polls.” It’s true that there is a correlation between what the polls are saying and the implied predictions on Intrade. How could there not be? At the same time, they don’t always track each other. For example, in the Business Insider article Paul links to, it mentions that in early November Herman Cain had a 4% implied chance of winning the Republican nomination on Intrade, whereas Romney had about a 70% chance. At the time, Cain and Romney were roughly tied in the polls.

    More importantly, even where the polls and Intrade do track each other, Intrade will often get there first. Paul mentions that the Intrade contract for Obama winning re-election has fallen as the polling for Romney has improved. Yet the Intrade contract started falling first. It was only later that polling caught up.

    The same goes for the idea that Intrade merely reflects conventional wisdom. If everybody agrees that, say, Obama did really horribly at last week’s debate, then it’s not surprising that this is also reflected at Intrade (though I would note that the Obama Intrade contract fell substantially before the debate was even over). Most of the time, though, you are not going to find such unanimity about the effect of some new political development on the campaign. When the 47% video came out, for example, a lot of folks said it would hurt Romney and a lot of folks (myself included) said it wouldn’t matter. In that case there really was no conventional wisdom for Intrade to reflect. Intrade, of course, reacted as if the news did hurt Romney, and my sense is that in the subsequent weeks opinion has shifted more towards this view as well.

    If you want to compare the accuracy of Intrade to polling or conventional wisdom, you can’t do it by looking at cases where they all agree. You have to look at cases where they disagree and see what it more likely to be right in those cases. If you do that, it’s clear that Intrade, while far from perfect, is indeed more accurate.

  6. (though I would note that the Obama Intrade contract fell substantially before the debate was even over).

    I only saw a few minutes of the first part, over my husband’s shoulder. Even I could tell Obama was sucking silt.

    If the drop had happened before he was eyeballing his belt buckle while Mitt talked, that’d be interesting.

    It’s also possible that Intrade isn’t tracking the same group of people; only folks who have money to spare are going to be risking it on betting about politics, and that’s a smaller group than it use to be, possibly with a bias towards college kids and gov’t employees.

  7. For what it’s worth, I agree with Donald’s forecast. If I had money to spare I’d buy a Romney future not just because I think he will win, but because he’d bring more money back in return, at least if I purchased a contract now.

    Perhaps what irks me is not Intrade itself, but the weight that people give to it. What would show me that it is of some value is if it had a regular history of not only outperforming the polls, but contradicting the polls and predicting winners early on. All that the Intrade results show me is that a few more people think Obama is going to win, but that a couple of more good Romney polls will probably bring the president’s numbers down even further. To me such a tool is of little or no value.

  8. Sorry to resurrect an old thread, but I end up seeing InTrade numbers all the time because I’m always refreshing the RealClearPolitics poll page. Throughout the election cycle thus far, Romney’s chances had been stuck under 40%. However, this last weekend he started to consistently creep over the 40% line and this morning I noticed Romney was above 44%.

    This would seem to suggest that the consensus that Obama’s victory is relatively certain is starting to recede as Romney’s (admittedly slim) lead in the polls holds steady just two weeks from the election. It’ll be interesting to see how that continues to develop.

    I wish I were as optimistic about the election as a whole as Paul and Donald. Right now it seems likely that things will all come down to Ohio, where I will definitely be at the polls first thing in the morning. I’d like to think that my new state is solid for Romney, but I don’t feel confident yet.

  9. I have been looking at the Intrade numbers as well and have noticed their almost obstinate refusal to budge until the last couple of days. Of course Rasmussen now has your state tied, so we’ll see.

    Something has to give with these polls. I know that it has been suggested that Romney could win the popular vote and lose the electoral college, but if that happens Romney’s pv margin would have to be fairly small. There is no way he can simultaneously win by 5% or more AND lose the electoral college – or at least it is very, very, very unlikely. So either Gallup and Rasmussen, who have had Romney up by 5 are off, or the polls that have Ohio a dead-heat or Obama up are wrong.

    FWIW, the Republican candidate has never done worse in Ohio than he has done nationally. Even when Republicans lose the state they do so by smaller margins than they do nationally.

  10. The author has analyzed this well. InTrade and the polls reflect the conventional wisdom that’s circulating in the culture. They automatically become more accurate as the election approaches, because the general societal thinking must be expressed during the election. As for predictive value, InTrade does not set static odds well in advance of the event. If it did that, then IT would be putting its money where its mouth is. While I understand the innate human desire to know the future, I do find it distressing that too many think that gambling is a way to do that. InTrade makes money no matter what the actual results are — just like the house in a casino.

Comments are closed.