I cant post the picture I want but here is my favorite poll super tracker.

http://www.pollster.com/polls/us/08-us-pres-ge-mvo.php

Despite all the pundit talk about McCain headed for victory a month ago or Obama surging because of the economy a basic pattern seems clear:

McCain has a fundamental support level at 43. He tends to return to bounce off of this. If he breaks support levels in the next week only then will we have seen a fundamental change.

Obama on the other hand has been in a bull market since March, with what looks like a smooth correction durring the summer and another sharp one after the Republican convention. Otherwise he has steady gains as undecdided seem to slowly decide for him.

Obama is slightly above his long run trend in the last few days and so we should expect some slight pull back. A failure to pull back would mean that Obama has broken resistance and is going to a landslide.

Overall, this looks vastly less dramatic than the media is palying it, though they have clear incentives to make things seem crazier than they really are.

[Edited on October 15, 2008 at 7:48 PM. Reason : .]

10/15/2008 7:47:45 PM

http://www.intrade.com/

10/15/2008 8:25:53 PM

While useful Intrade and other betting sites don't provide information of the basic dynamics of the race.

A 0-1 futures contract will trade much differently than the underlying polls for a variety of reasons. The most obvious of course is time value. Suppose Obama held a steady 4 point lead over McCain from Jan 1st to Nov 4th. There is no change in the dynamics but the futures contract will trade steadily higher as the time value for McCain runs out.

More interesting would be a contract that paid out at the percentage of votes for McCain or Obama, but even there time value will effect the trade.

10/16/2008 12:56:01 PM

I'd short McCain something fierce.

10/16/2008 11:57:28 PM

^^ While intrade does tend to over-estimate the value of a consistent lead, the 0-1 payoff is not inappropriate given that it is the same as the possible outcomes of the election. It also makes sense if you're looking at the probability as a function of time and take the limit as the time approaches the end of the election. A 4 point lead held as the remaining time goes to 0 is a win, so the probability of a win as a function of time would increase as the days went on. This is basically the market expression of the idea that swings in opinion take either extraordinary events or a lot of little ones. The little ones take time to add up and the big ones take time for their relative small chances of occurring to be significant.

I agree that the all or nothing aspect would mean that the intrade value does not equate to an actual percentage chance. Using it as that would be a mistake. It does resemble an actual probability more so than poll numbers do, since poll numbers do not collapse to 0/1 as the end of the election approaches. Perhaps a better way of tracking momentum and chance would be to examine both the futures values (on intrade and cnn et al.) and a wide variety of polls. The polls give you a better idea of trends than the futures markets do; the futures markets on the other hand give a better total predictive value and seem more accurate in choosing victors.

A contract that paid out as a percentage of votes could be misleading but interesting. Say a future were to be overvalued. Polls dropped from 58-39% down to ranging 49-42%, but shifts in the electoral college gave the leading candidate a better probability of victory none the less. Their futures contract value would plummit, but this would not in any way indicate an actual lowering in their chances of winning in said scenario. For reasons like this, predictive futures markets work best as binary propositions. They can be used to get somewhat accurate guesses on "yes or no" but can lose their usefulness when you try to ask by how much.

That's actually one of the cool things about using them on a state to state basis. Our much maligned electoral system actually creates a situation where the futures markets are able to predict the electoral results using multiple binary contracts- and somewhat succesfuly compared to polls the last couple of years. Still, you're right to distrust them. If the US stock markets have shown us anything, it's that markets aren't always governed by rationality.

10/17/2008 4:11:06 AM

McCain's got this locked up.

10/17/2008 10:14:39 AM

http://www.intrade.com/ is really interesting to watch during the debates.

10/17/2008 1:06:34 PM

10/17/2008 1:10:20 PM