Predicting elections

The New York Times has published yet another article about “market” approaches to predicting outcomes: people essentially wager on an outcome (like “Barack Obama wins the 2008 presidential election”). The article points out that on the website Intrade.com, the “market” had decided that the Republicans would lose control of the U.S. Senate by about 2am on election night, even though the news and political pundits were reporting that the Republicans appeared to be holding on by one seat. The article goes on to say the usual good things about markets, and says:

N. Gregory Mankiw, a former adviser to President Bush, who has written about Intrade on his blog, explains it this way: “Everybody has information from their own little corner of the universe, and they’d like to know the information from every other corner of the universe. What these markets do is provide a vehicle that reflects all that information.”

I don’t disapprove of these “information markets” and I think they’re kinda fun, and potentially even useful. But these sites might not be illustrations of the power of the market to correctly take disparate sources of information into account in order to make a prediction. Instead, they might be illustrations of the fact that if you have information that is not available to other people, you can make a better prediction than they can. Even the name of the site that is mentioned, “Intrade.com”, carries the suggestion of “insider trading.” For example, suppose you were a poll worker in Virginia and you knew two things that were not known to the media or the public at midnight on election day: (1) most of the highly Democratic precincts have not reported yet, and (2) the ones that have, have been even more strongly Democratic than usual. Of course you would have good reason to think the Republicans would lose the Senate, and you could log into Intrade.com or some other similar site and put your money down. And the market price would indeed reflect your knowledge. But what is at work here isn’t the magical power of markets to use all of the available information, it’s the magical power of “insider” information to let you make a better prediction. If the same information were available to CNN…or to me, for that matter…we wouldn’t need “the market” to tell us the Democrats would win the Senate. I’m not sure the Intrade.com market was actually integrating information any better than the pundits, their bettors/investors might have just had better information.

8 thoughts on “Predicting elections

  1. But isn't that kind of the point? The unit that sways the "market" decision is the number of dollars, not the number of people, on each side. So if I have more information that everyone else, I realize that I can make a killing by putting a big pile of money on the the likely, but unpopular, side. And then my pile of money tips the scales back in my direction. And if you actually know something that I don't, you may react by upping your ante, and the scales swing back. Given a large number of us (assuming normal distributions of bettable income and risk aversion among us) eventually, without ever conversing or arguing or comparing information, the "market" finds the weighted average of our knowledge, measured by the number of dollars we're willing to back it.

    You're right that common wisdom, public information, and an "opt in" format relegate this dynamic to those in the know (or who think they're in the know!). But given the multidimensionality of some of these decisions, lots of people are kind of in the know. Any any centralized group would find explicitly incorporating all that individual knowledge very difficult.

  2. well that's the whole point,
    talking heads at cnn couldn't have this information
    because the poling guy just has no incentives to give it to them.

  3. Regarding markets vs. the TV pundits: another factor is incentives. A market player has an incentive to bet on whichever option is judged to be more likely, but TV networks have an incentive to be more hesitating in their judgments: the negative payoff for being wrong (in a "Dewey Beats Truman" sense) is so great. On TV, they're not going to want to call a surprise winner unless they're completely sure.

    It's similar to what they say about football coaches and baseball managers, regarding decisions such as going for it on 4th down: if you go by the book and win, great. If you go by the book and lose, too bad. If you do something risky and win, great. But if you do something risky and lose, you look really really bad. (This may be changing with sabermetrics and all, but that's the traditional view, at least.)

    With TV networks there's the additional concern that they'll be accused of bias if they call the wrong winner.

  4. Brent, I think there are two separate issues: (1) markets supposedly make use of the entire range of available information, whereas any single analyst only makes use of a portion of it, and (2) some analysts have information that is not available to other analysts. (I'm using 'analyst' here to refer to anyone who is trying to make a prediction.)

    Item (1) is what is usually emphasized in these articles about markets, and is probably what is most important in the stock market, but what I'm saying is that item (2) is perhaps more important in markets like the one discussed in the article. It's the "insider"-ness of the information that I think is significant here.

  5. It's certainly the case that some people who enter the market have better information than others — the so-called insiders.

    But there's perhaps a more general point to made that the best prediction is probably not the one made by the person with the best information, but some sort of aggregation of private information held by different insiders weighted by something like how much they're willing to bet. That is, even if you restrict the futures market to just the insiders it will do better than any one of them.

    And the same general point would go for incentives. Even taking out incentives to be risky or cautious, you could imagine it to be the case that betting pools that tap into private information are better than the performance of any member of the group with the appropriate incentives.

  6. This is probably a wrong example to use for either side in a general argument about markets and/or insider information. Most of that "insider info" was publicly available on the WWW. Having no more privileges than anyone else, I was able to follow the geographic distribution of votes coming in, and around midnight I was reasonably sure that Webb would win Virginia. I suspect I might have been able to make similar inferences about Montana if I had been more knowledgeable about its demographic patterns.

    The really interesting market story here is that the pundits were talking as if those data – free publicly accessible data – had not existed, which was appalling. It is a story of market failure, that such incompetent people get so highly paid.

    It may even be that some of them are not incompetent, but they willfully ignore the available information because their opinions are somehow perceived as more valuable if they don't rely on facts. That would be a sad and scary explanation, but I don't see that it can be easily dismissed.

  7. Greg, I'm right with you: "the best prediction is probably not the one made by the person with the best information, but some sort of aggregation of private information held by different insiders weighted by something like how much they're willing to bet." Well put.

    indeed, some companies have set up internal "markets" in which executives can bet on things like the cost of oil in six months—a good example of your point that a market of all-insiders might do a good job. Or at least, the companies that have set up these internal markets think they might. (There is even a company that sells software to enable corporations to set up this sort of market, so perhaps the demand is not tiny). But of course, although these guys are all insiders compared to _us_, they're presumably all on a level field relative to each other. If the Senior VP has access to info that the other executives don't, the market will eventually just reflect his views and the others will stop wagering.

    I think that in some cases — and the case of the Virginia election _might_ be one of them — you might have one or two insiders and a bunch of outsiders. Even if the insiders don't quite have things right, they might be so much better-informed than the rest of the market that "the market" doesn't really come into play, you're really just looking at the opinion of the one or two people who are in the know. So as far as betting on the outcome of the 2008 Presidential election goes, if I'm going to put $50,000 on the outcome, I won't do it in one of the online markets: the guy who has photos of Hillary Clinton, Barack Obama, and John McCain in a menage-a-trois can clean my clock before emailing the photos to CNN.

Comments are closed.