Are Prediction Markets Better At Predicting Election Outcomes Than Polls?

The morning after the folks in Scotland decided not to break away from Mother England, I read this article in The Upshot at the New York Times. The author noted that while most opinion polls about the Scotish vote correctly predicted the proposal would fail, they were volatile while prediction markets were more stable and suggested a resounding no vote, which is what happened. The author includes this comment:

My own research with Microsoft’s David Rothschild suggests that pollsters could do a better job if they learned from prediction markets. Instead of focusing on whom people say they plan to vote for, ask them instead to focus on who they think will win. Typically, asking people who they think will win yields better forecasts, possibly because it leads them to also reflect on the opinions of those around them, and perhaps also because it may yield more honest answers.

I decided to poke around and see if there were any prediction markets focused on U.S. politics. Intrade used to provide a predictive market for political races but sadly, they’re no longer around. I found a two that seemed active. I’m sure there are others out there and if you know of any, please let me know.

IPredict is a New Zealand based prediction market focused on politics. You can place bids on which Party will control the two Houses of Congress. As I type, they predict a 98.3% chance the GOP will hold the House and a 62.3% chance the GOP will take control of the Senate.

Another prediction market I found was run by the University of Iowa. The ask price as I type for the “GOP Gain” position in the House is .955, or a 95.5% chance the GOP picks up seats. The last actual price for this position was .88. The ask price for the “Democrat Lose” position in the US Senate is .835 or an 83.5% chance the GOP gains enough Senate seats to gain the majority. The most recent price for this position was .793.

Here is a graph of the “Democrat Lose” the Senate Market from the University of Iowa. As you can see, the purchasers in this market seem to be increasingly of the opinion the GOP will indeed win a majority in the Senate. The participants in this market seem more sure of this than most pollsters and those at websites which use polling to offer a percentage chance of a certain outcome.

Senate14

What do you folks think about this? Do prediction markets have a better read on what will happen on November 4th? Should pollsters ask people for what they think will happen and not just how they plan to vote?

18 comments

  1. dsean says:

    Intrade and similar markets were shuttered by the SEC (and/or the CFTC, I don’t recall which specifically). The government takes the position that these are unregulated commodities markets and are illegal.

    • Will Durant says:

      How is it that certain “commodities” on the regulated markets are still considered legal by the government? How does my purchasing an option on the perceived value of S&P 500 futures differ from wagering on a sports team? There really is no traditional underlying commodity involved and it suggests to me that the determinations of legality by the “government” vary greatly.

  2. Liquid prediction markets are pretty good. I don’t think the Iowa markets are very liquid. And the problem with the other markets is that Americans (those with the best knowledge) aren’t typically allowed to participate in them.

    • Bobloblaw says:

      It is essentially the market (betting sites like the old Intrade) versus Central Planning (one guy’s model). Ill take individuals trading with one another versus a model created by a college brain any day.

  3. George Chidi says:

    For the prediction markets to be considered authoritative, there needs to be enough liquidity to minimize arbitrage opportunities. I don’t think that’s what we’re seeing here.

    A pair of researchers examined anonymized Intrade data after the 2012 election, and discovered that one or more traders had been manipulating prices with the intent of making a Romney win appear more likely.

    One highly-determined trader with money to burn blew about $4 million in the final weeks of the campaign, making one-third of all bets on Romney during the two week period of the study — and about one out of seven of all bets during the election cycle, just to put a floor under Romney’s win-lose chances.

    The thinness of the trade at the University of Iowa and iPredict suggests that a trader could do the same in a state race or on the congressional contests with far, far less capital at stake.

    • Bobloblaw says:

      which is a stupid thing to do. It doesnt affect the outcome of the election and sets up others to make money. If you bid up Romney to 75%, Ill just place bets against it and drive him back down.

      • George Chidi says:

        … which didn’t happen until Election Day.

        We’re talking about the practical power of markets versus polling to predict political behavior. This isn’t an ideological question. There’s empirical evidence piling up that prediction markets — at least in their current, thinly-traded form — aren’t effective. Somehow, you’ve decided to interpret that as a call to abandon the free market and bust out the Little Red Book. It’s not. But markets are not always perfectly efficient mechanisms for evaluating information, and this is an example.

  4. George Chidi says:

    Now … were I an entrepreneurially-minded fellow (and I am) with financial training (guilty) and capital at my command (not quite …) I might look for ways to measure market sentiment in more liquid markets by bundling together the options on a set of companies with high sensitivity to political outcomes, then selling those options as a political market index.

    If, for example, one knew cold that a Hillary Clinton win probably takes $1 billion out of Halliburton’s cash flow over four years, one might put an appropriately-priced put option in the index. If the index manager were sufficiently cynical, one could do a regression analysis looking at the predictive value of campaign donations from corporate executives on the stock price of their firms, then craft a synthetic option weighting those predictions.

  5. Good discussion.

    One question for folks like Chris and others who actually do polling: Would it be beneficial to ask people who they think will win a given race in addition to, or perhaps even instead of, who they plan to vote for? Would that give you a better feel for who might win? It would take into account what the friends/neighbors of the person being polled are thinking right?

    • I have been asking both and maybe I will share after the elections are over! It is very interesting to see how they are sometimes the same and sometimes different.

    • On a more serious note, I think sometimes people forget that there’s a difference between media polling (who is winning and why) and most campaign polling (who is winning yes but what message should we use to try to persuade people to vote for us instead).

      In that regard, I don’t see most campaign pollsters moving to a focus group like who do you think will win methodology anytime soon, though it is interesting.

      But non-traditional polls like “Who do you think will win”, polls of children asking who their parents are voting for, even the 7/11 presidential cup poll are often excellent predictors. It’s one thing to tell a pollster in private who you are voting for, it’s another thing to show up at the work site with the loser’s coffee cup in your hand.

  6. elfiii says:

    The Iowa futures market might be thinly traded and not very liquid but it has closely tracked the outcome of most of the elections over the last 15 years. Call it luck, call it circumstantial. I call it voodoo mon.

Comments are closed.