September 08, 2003
the Iowa political futures market
A well-known experiment, run by Iowa Electronic Markets, allows traders to place bets on the outcome of political elections, including the current California governor's race. According to a paper by Joyce Berg and others, the Iowa Political Market has outperformed polls in predicting 9 out of 15 elections. Its average error in predicting election results is about 1.5%, compared to about 2% for an average poll. In some past elections, the Market avoided major errors that marred all the major national surveys, whereas it has never made a gross mistake itself. The apparently uncanny ability of the Iowa Electronic Market to predict the future was one of the reasons that the Defense Department recently floated the grisly idea of a futures market in terrorism.
I'm struggling to understand the theoretical explanation for this phenomenon. I realize that markets efficiently aggegrate the knowledge of investors (who must try to make honest predictions, since their money is on the line). But where do the investors in a political futures market get their knowledge? They cannot simply ask themselves how they intend to vote. As Berg et al. note, traders are "not a representative sample of likely voters; they are overwhelmingly male, well-educated, high income, and young" (p. 2). Some are not even US residents. Thus their own choices in the real election, assuming they vote at all, will be very different from those of the American people. Yet they seem to be able to predict the actual result more accurately than a random-digit telephone poll.
One clue is that a relatively small number of "marginal traders" drive the market; they make many more trades than other people and are less prone to sticking with an unlikely bet out of loyalty. I would guess that these "marginal traders" are political junkies: people who have no sentimental attachment to any of the candidates but love to prognosticate about elections. We can assume that they have seen all the polls—but that still doesn't explain how they outperform surveys on average. Could it be that they instinctively recognize a consistent error in polling, and adjust accordingly? For example, maybe polls tend to pick the real winner but predict a larger margin of victory than actually occurs. (Races tend to "tighten" right at the end.) Or maybe polls tend to make inflated predictions for the Democrats' share of the vote, because they count too many low-income people as "likely voters." It's also possible that the marginal traders rely on one or two polls that are better than the average. (Then we would find that the market outperformed polls in general, but was no more accurate than the best of the polls.)
These are hypotheses backed with no evidence. But if one of them turns out to be true, then we don't need a market to improve on surveys. We just need to make the same adjustment to poll results that the marginal traders (a.k.a., the political junkies) are making. Likewise, we would not benefit from a futures market in terrorism, but we should strive to understand how the best informed and least sentimental observers of terrorism make their predictions.
Posted by peterlevine at September 8, 2003 12:19 PM
Post a comment
This blog is under attack from comment spammers, who are causing a problem for the server. I believe I can block them by upgrading to a recent version of MoveableType. However, I do not have time to do that until late December. Therefore, I have temporarily disabled comments. Please feel free to email me feedback at email@example.com.