Polls vs Markets

My friend Alan Warms sold his small company Buzztracker to Yahoo! last year and has been doing some great work inside Yahoo! since then. Yesterday, he sent me a link to their new election dashboard service which I have bookmarked and will now visit everyday until the election.

I like the fact that I can get everything I need to know in one page view with links to things I care about like the most blogged about political stories of the day. It’s very well done although I sure wish they’d done it in ajax instead of flash.

The thing that pops out at me when I look at the election dashboard this morning is that the race is very tight in the polls (47% Obama, 44% McCain). And the polls will clearly tighten in the next week as McCain benefits from the limelight of the republican convention.

But the election markets are not tight at all. Intrade has it at 60% Obama, 40% McCain and it hasn’t really moved in the past three months.


Hub Dub has it at 70% Obama, 30% McCain and that also has not moved much in the past three months.


So why is it that the polls are tight and tightening and the markets are not? And are markets better predictors of news than polls? I suspect the answer to that is yes.

I’ve been placing a lot of bets lately on Hub Dub and I am having fun and engaging with the news in a way I’ve not done before. It’s the same effect playing fantasy baseball or football has on your enjoyment of the real game. And in the process, it may become a new way for all of us to think about the news.

Reblog this post [with Zemanta]
#Politics#VC & Technology

Comments (Archived):

  1. kidmercury

    any discussion of election polling and markets for elections would be incomplete without noting that the last two elections have been stolen. links to learn more:http://whatreallyhappened.chttp://liesofbush.com/2000e…blackboxvoting.org is a great site to get active.”Those who cast the votes decide nothing, those who count the votes decide everything.” – Joseph Stalin

    1. andyswan

      Curious if you are including the 2006 Congressional elections when you say “the last two elections”

      1. kidmercury

        i don’t know for sure, but where there are diebold machines there are going to be problems. the hacking of diebold machines has been a big problem — here is one good article among many others:http://www.engadget.com/200…it’s worth noting there was also a lot of controversy surrounding the 2008 new hampshire primary on both the rethuglican and dumbocrat side, with evidence suggesting both obama and ron paul were robbed of votes. there are plenty of people arguing both sides of the issue, so it is tough to prove with certainty. below is a google serp that may help people make their own opinion.http://www.google.com/searc…ultimately the diebold issue is very real. so i don’t know for sure but it certainly seems possible if not likely that there were issues with the 2006 congressional election as well.

        1. andyswan

          Gotcha. Agree integrity of voting is everyones issue.

  2. Boris

    I wrote up this exact post on my blog about 2 weeks ago noting the large disparity between election markets and the polls.I think it really comes down to the fact that the polls are overall popularity contests, whereas the markets take into account relative popularity in key swing states. Just because they’re getting closer together in national polls, doesn’t change the poll numbers out of states like Ohio, Pennsylvania, Florida, etc. I think a large part of the disparity comes out of our electoral system.Markets are definitely better predictors than polls. There’s a great research paper out of the University of Iowa which runs the Iowa Electronic Markets as a research into prediction markets. The abstract of their findings is here:http://www.biz.uiowa.edu/ie…”The market is closer to the eventual outcome 74% of the time. Further, the market significantly outperforms the polls in every election when forecasting more than 100 days in advance.”

  3. byrneseyeview

    So why is it that the polls are tight and tightening and the markets are not? The markets are probably discounting any predictable changes in the polls. If we all know that McCain will get a boost from the convention, who is going to sell us those underpriced McCain contracts to take advantage of this? And the odds are presumably less tight than the polls because there is only a 40% chance that McCain can overcome a three point gap in the vote. The best model is probably to look at the vote gap, and the margin of error for the polls — that should give you the probabilities of all outcomes (e.g 45-55, 46-54, etc. ) from which the final odds can be calculated.

  4. fnazeeri

    I follow a bunch of these political sites, and by far the best of them all is http://www.fivethirtyeight….. It’s not partisan as far as I can tell, instead they aggregate astonishing amounts of data on polls. It’s the data sets that the campaigns look at…state-by-state, up- and down-ballot. If you want to geek-out on numbers, it’s the place to go.

  5. jedc_mercury

    Yes, prediction markets show more information than polls, for the simple reason that it’s trader judgement in addition to polls.I’ve done research on prediction markets, and if constructed correctly are really quite accurate. The important thing to remember is that just because a contract is above 50% doesn’t mean it’s going to happen, and I’ve blogged about that here:http://blog.mercury-rac.com…What I find interesting is that the best polling stats site, run by a well-known sabermetrics figure (Nate Silver), has a winning percentage for Obama that seems to be tracking well with the prediction markets at around 55-65%. MUCH more data on that here:http://www.fivethirtyeight.comAnd I also enjoy Hubdub… it’s a well-implemented prediction market with a good community.Jed

  6. MartinEdic

    The polls cannot reach households that do not have a landline, the majority of which are under 35- or right in Obama;s sweet spot. in addition there is some thinking that independents are so sick of polls that they are either not participating or even lying to throw off their value. If these scenarios are true then the markets are probably indicating a truer direction. Read Surowiecki’s The Wisdom of Crowds for why markets have high accuracy rates (great read!).

  7. andyswan

    Well….these are two VERY different types of numbersI would argue that a three point lead in voter polling SHOULD lead to a 60-40 lead in wagering because with a three point polling lead you are statistically favored to win 60% of the time.I’d assume that if the polling was coming in at 55-40, the wagering odds would be around 95-5.

    1. fredwilson

      That’s a really good explanation AndySo obvious it makes me feel stupid actually

      1. randomwalker

        In addition to what andy said, intrade *has* been getting tighter in the last few days — it’s now around 55-45 obama, not 60-40.BTW, Andy — a three point lead by itself doesn’t mean much, you need to know confidence intervals to turn that into probabilities.

    2. Jake

      Great explanation Andy. I think that one simple way to look at the disparity is to ask what the two different numbers are getting at.Polls ask the question, “Who will you vote for” whereas the markets ask the question “Who will get more votes overall”

    3. LukeG

      In the obvious/stupid vein, how does that math break down?

  8. Bryan

    Interesting take Fred. Pretty hard to argue against it.Too bad on the Romney call.

  9. gyardley

    Hmmm – the HubDub market leans way more Democrat than the Intrade market. That’s probably because the HubDub markets use play money while the Intrade markets use real money – in the HubDub market, it’s all for fun, so you can make emotion-based trades.Of course, people trade emotionally in real-money markets as well. Do we think the InTrade community slants Democrat or Republican? I personally suspect there’s at least a few percentage points of wishful thinking in the InTrade numbers, and a lot of the trades are of the ‘let’s bet on my favorite sports team for fun’ variety instead of the ‘okay, let’s rationally make some money’ variety.

  10. steveray

    Every year when the entire college football community argues about the national rankings (which determine the allocation of $millions for bowls) I make the point that the people who make the polls should have money riding on them. Every week’s ranking should in effect be a bet. You’d get much more accurate rankings that way.

  11. iWonder

    How did these markets do at predicting McCain as the nominee, back when everyone counted him out?

    1. MattKane

      His price (for nominee) on Betfair spiked to around 35 last summer, which would imply he was being written off by the punters too. You can see the graph here: http://bit.ly/2LFcVq though for some reason it doesn’t show the dates on the axis. To give you an idea, I backed him at 7.8 in late December. I then covered that bet by laying at 1.07 in April, so I’m in the money even if he drops dead or something.

    2. Nigel Eccles

      McCain bombed last summer,, in the polls and on the markets. When he hit rock bottom in September he was trading at about 2-3%.Prediction markets respond to sentiment in the same way that stock markets do. If external data is saying the company isn’t performing well then the price will drop. Last September, John McCain was the about the only one on the Straight Talk Express. He worked his way back and his price rose with it.

  12. andyswan

    I was dead wrong about Romney being the pick….but this one is even better. Newspapers saying “Obama who?” all weekend after his amazing speech….and “diversity” on both sides of the ticket with a kick-ass woman that Biden can’t assault.Maybe McCain isn’t as lame as I originally suspected…

    1. fredwilson

      A 72 yr old man should pick someone who is ready for the white house as hisVP

      1. andyswan

        I’m sure the McCain camp is looking forward to hearing exactly how the dems define “ready” and “experienced”… Brilliant trapan inspiring feminist story forwomen, moms and fathers of daughters.

        1. fredwilson

          Obama has been tested by a year long primary campaign where he had to beheadthe king and queenShe has not

          1. andyswan

            And to think all that time, she was merely running a state. :)Of course, she isn’t on the top of the ticket.

          2. fredwilson

            Andy, I’ve rewarded you for being a thorn in my side today with the commentof the dayhttp://www.avc.com/a_vc/200…

  13. MattKane

    I think it’s undoubtedly true that people are lot more candid when they’ve got money on the line. Another interesting place to watch (though you can’t place bets if you’re in the US) is the politics markets on betfair.com which is a betting exchange, and spreadfair.com, which a spread betting site. I’ve already made some nice beer money on various US election markets on there. Politcalbetting.com is the UK’s most popular political blog and covers a lot of this kind of stuff. Much of it is inevitably UK-specific, but there’s obviously a lot of US coverage at the moment, as well as worldwide.

  14. tpurves

    Polls don’t necessarily consider turn out. This election is likely to be a landslide based on turnout. Obama was smashing democratic turnout records in the primaries.It doesn’t matter how many supporters you have, it’s how many get to the voting booth (and how many of those votes are counted but that’s another story).

  15. Nigel Eccles

    Interestingly Intrade’s forecast for Obama is actually at the low end for the betting markets. If you take the prices from Oddschecker (http://www.oddschecker.com/…, strip out the overround (that’s the bookmaker’s margin) and convert to percentages, you will see that Obama is trading at between 61% and 65%.Around the question if the betting markets are likely to be politically biased, this is highly unlikely. Many years ago you could make money by, for example, betting against England when betting with a English bookmaker (and then potentially arb it by laying off with a Swedish bookmaker) however the betting markets are so transparent and so many bettors use bots that systematic bias doesn’t really occur (or if it does it is fleeting).The Hubdub presidential market has been consistently about 5% more in favor of Obama than the betting markets. We believe part of the reason for this is the automated market maker we use but we are still investigating. We believe it isn’t because of emotion based trades (as someone suggested above). The price the market settles on is mostly controlled by our heavy traders. These users spend up to several hours a day on the site competing with friends and trying to climb the leaderboards. They tend to treat their trades as they would real money. The emotion based trades cause volatility but shouldn’t set the long term price.

  16. howardlindzon

    More important, who should Jim cramer pick to be his replacement on mad money.

    1. fredwilson

      You of courseBooyah

  17. Tom Royce

    It looks like things are tightening up…Intrade Real Time Quotes 48% McCain – 53% ObamaThat is a big swing amongst the betting folks out there.

    1. fredwilson

      It sure is a big move. The sarah palin factor is at work

  18. shassinger

    I really like the work being done on fivethirtyeight.com — heard Nate Silver, the force behind the site, on NPR and was attracted to the transparency and rationality of his methodology.