[EM] Markets and Information
alex_small2002 at yahoo.com
Thu Sep 30 14:56:40 PDT 2004
I'm sure some people here have heard of the Iowa Electronic Markets. People on those markets buy or sell contracts that will pay $1 if a given candidate wins. Some of the contracts also include stipulations that the candidate win by a certain amount. The market is run by the business school at the University of Iowa, to test and/or demonstrate certain theories of markets.
(FYI, to keep the transactions simple they base the trading on the votes received only by Kerry and Bush, not by any third-party candidates, and they only look at the 2-party vote, not the electoral college. In 2000 the people who bought shares in Gore got their pay-outs even though Bush won, or at least sort of won, however you want to look at it.)
The basic idea is that markets contain collective information that might not be recognizable to individuals. There's been a lot of mathematical work done on these sorts of things, and the idea is that when you collect information from lots of people the signals add up and the noise cancels out. If the race were a toss-up then a Kerry future would be trading at $0.50 because the expected value of that contract would be only $0.50. This approach is believed to be better than polling because the money aspect forces people to evaluate their information and decide how reliable it is.
And apparently it works. There is data out there supposedly showing that in most cases (no, not all, but a statistically significant fraction of the time) agricultural futures markets can predict droughts, cold snaps, and other weather patterns that will influence agricultural prices. It might seem crazy, but apparently there is a lot of good folk wisdom out there co-existing with a lot of just plain wrong ideas. The slightly bias that each person has from good information means that right guesses outweigh wrong guesses in the end, and so the prices are able to predict which crops will be adversely affected by weather. Likewise, the Iowa Electronic Markets have supposedly been good at predicting election outcomes over the years, supposedly doing at least as well as the pollsters.
Anyway, you can learn more about this from http://www.biz.uiowa.edu/iem/. Here's my question:
Does anybody know if work has been done (either mathematical or observational) on markets that affect the outcomes they're supposed to predict?
I'm not just talking about predictions on the actions of buyers and sellers, which are obviously both influenced and predicted by prices. I'm talking about, say, elections. On the one hand, stock prices in politically sensitive industries should reflect info on likely election outcomes. If it looks like candidate X will win, and X is promising to regulate the Widget industry, widget stock prices should reflect that. If it looks like Y will win, and Y is promising tax breaks for widget makers, the stock prices should also reflect that.
But at the same time, stock prices affect things like retirement accounts and even insurance prices (insurance companies often put some of their holdings into stocks). All of these things affect the overall economic picture, which affects elections.
Does anybody know if work has been done on the predictive power of markets that influence the events for which one is trying to get a prediction? The reason I ask is this: Say that mechanisms like the Iowa Electronic Markets were to get some publicity and develop a reputation outside of the small community of social science professionals and amateurs. If the predictions were received credibly, then those predictions might affect things like voter turnout, not to mention donors and activists deciding whether or not it's worth their time and money to help a candidate.
Anyway, just curious if anybody knows about work on the predictive power of markets with feedback.
(I realize that this isn't directly related to election methods, but it is a mathematically-oriented topic involving voting, so I hope people will forgive me for posting this.)
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Election-Methods