The Wisdom of Crowds?
Here's something fun. Point your browser to one of the websites at which people can trade "shares" in candidates' likelihood of winning the nomination, the Presidency, above a certain fraction of the Electoral College and so forth. The Iowa Electronic Markets are the best known of these "political markets," but my personal favorite is Intrade because it's the easiest to use. In any event, it's considered a sign of the wisdom of crowds (also the title of an excellent book by James Surowiecki) that these markets more accurately predict the outcomes of the Presidential race than do opinion polls.
Well, it may depend on your time frame. Over on Intrade, shares of Hillary Clinton to get the Democratic nomination were trading at 70 cents (meaning if you paid 70 cents you would earn a dollar if she got the nomination and nothing if she didn't) on January 1st, but as of 10:30 p.m. on January 7th, Clinton shares were down to 28 cents (with a bid/ask spread bringing their real value even lower). This sort of stock price crash seems pretty clearly to have been a reaction to the polls, rather than a leading indicator. Clinton's share price fell precipitously after the Iowa Caucuses, but then fell precipitously again after post-Iowa polls of New Hampshire voters began to show Obama opening and then widening a lead in the Granite State.
Somewhere in that observation about the political markets is a cautionary tale about the herd mentality. The wisdom of crowds works best---as Surowiecki shows---when many people actually know a little bit and their errors are random, so as to cancel one another out. That's why the average guess as to the number of jelly beans in a jar is so frequently better than any actual guess. But political marketplaces are guesses about what other people are likely to do, and as to these, it's not at all clear that errors are randomly distributed. Especially if most "investors" are getting their information from the same sources, any systematic biases in those sources will be reproduced in the investors' wisdom.
If this all seems like an esoteric amusement, it's worth recalling that serious academic and real-world proposals are afoot for important government decisions to be made using markets of this sort. The idea, despite its guilt by association with John Poindexter, is not inherently crazy, but it has definite limits.
Posted by Mike Dorf
Well, it may depend on your time frame. Over on Intrade, shares of Hillary Clinton to get the Democratic nomination were trading at 70 cents (meaning if you paid 70 cents you would earn a dollar if she got the nomination and nothing if she didn't) on January 1st, but as of 10:30 p.m. on January 7th, Clinton shares were down to 28 cents (with a bid/ask spread bringing their real value even lower). This sort of stock price crash seems pretty clearly to have been a reaction to the polls, rather than a leading indicator. Clinton's share price fell precipitously after the Iowa Caucuses, but then fell precipitously again after post-Iowa polls of New Hampshire voters began to show Obama opening and then widening a lead in the Granite State.
Somewhere in that observation about the political markets is a cautionary tale about the herd mentality. The wisdom of crowds works best---as Surowiecki shows---when many people actually know a little bit and their errors are random, so as to cancel one another out. That's why the average guess as to the number of jelly beans in a jar is so frequently better than any actual guess. But political marketplaces are guesses about what other people are likely to do, and as to these, it's not at all clear that errors are randomly distributed. Especially if most "investors" are getting their information from the same sources, any systematic biases in those sources will be reproduced in the investors' wisdom.
If this all seems like an esoteric amusement, it's worth recalling that serious academic and real-world proposals are afoot for important government decisions to be made using markets of this sort. The idea, despite its guilt by association with John Poindexter, is not inherently crazy, but it has definite limits.
Posted by Mike Dorf
posted by Michael C. Dorf @ 1:00 AM
3 Comments:
At 9:33 AM,
Mojave Joe said…
Reminds me of a term I heard for the first time recently, "Condorcet Jury Theorem," which is described as follows:
Suppose that there are n voters who must decide between two alternatives, one of which is correct and the other incorrect. Assume that the probability that any given voter will vote for the correct alternative is greater than 1/2 . Then the probability that a majority vote will select the correct alternative approaches 1 as the number of voters gets large.
At 11:13 AM,
Michael C. Dorf said…
That's right and one of the interesting things is that permitting deliberation can therefore decrease the accuracy of the result because the jurors are no longer acting independently. There is a burgeoning literature on when deliberation improves and when it undermines group decision making.
At 5:44 PM,
Jordan said…
I guess I am not quite sure the point of this post. Many analysts argue that information markets (or "policy markets" or "futures markets") are valuable predictive tools because they aggregate all the information available to traders and they are simply better than other tools like opinion polls. I don't see how the fact that the market for the winner of the presidential primary shifted dramatically in response to the results of Iowa or New Hampshire affects this debate.
The fact that information markets react to changes in the polls should not discredit them in relation to other tools. Of course information markets respond to polls; they respond to everything! Polls and the results of the early primaries are some of the best raw data for traders to use when trying to determine which "presidential stocks" are overpriced and which are underpriced. They are similar to financial reports in more traditional stock trading.
If there was a systematic bias in the response of the public to poll results, then savvy traders would be able to profit by trading in their shares of Obama for Clinton after Iowa, and making the opposite switch after New Hampshire. The effect of these trades would be to move the stock price closer to the true percentage chance that a given candidate will win.
It is possible to argue that the market of traders on Intrade is not robust, and there are not enough of these savvy traders. If that were true, then readers of this blog should enter the market and go make some easy money. If they did that, the market prices would be better for it.
I think the vast fluctuations in the price of the presidential stocks on Intrade in response to the early primaries is good evidence that it is difficult to predict the winners of elections. Or, more specifically, this election. However, I do not see how it is good evidence that information markets are not the best method for making that prediction.
A good source for publications about, and examples of, information markets can be found here: http://www.aei-brookings.org/pages/index.php?id=37
Post a Comment
<< Home