A quick disclaimer. Although I think that this site is useful and serves a small niche, there are better sites available. In particular, I've drawn heavily from Sam Wang's Electoral College Meta Analysis.
The form below takes state-by-state inputs as percentages for Bush, such as that provided by Tradesports. (Look for Quote Board, then Politics, then the state-by-state Presidential section.) The state-by-state odds will incorporate the possibility of a national swing in the popular vote, plus state-by-state variation. Of course there is uncertainty associated with both of these. You choose how to allocate that uncertainty, by specifying the percentage that is national. My best estimate of the value is 60%. (It cannot be 100%, but it can be 0%. The overall prediction doesn't work consistently with values outside of 5% to 95%.) Then just plunk in the state-by-state odds. The cgi will do the permutations for various values of the national vote swing and predict the probability of a Kerry win. I've prepared another page with more details about the methodology (and the math).
A note about Maine's second congressional district: Maine is treated as three votes and the second congressional district is one. As Maine is slipping further out of play it matters less and less.
Comments: (11/3 12:00 PST) I didn't get a chance to update the web page yesterday during the election, but watching the shifts in the futures values was very revealing. The reaction to exit polling was pronounced, and the state-by-state aggregate probability tracked the direct contract probability quite well.
(11/2 6:37 PST) Kerry's advantage in the state-by-state was short-lived. (His state-by-state aggregate probability is back to about 48.36%, the direct-contract probability is still 54.9%.) Still, it does point out a rather sure betting strategy. Any time that the bid price for PRESIDENT.GWBUSH2004 is at least 1.6 more than the ask price for BUSH.FLORIDA, you can make a bet that is highly in your favor. If you were to buy BUSH.FLORIDA and sell PRESIDENT.GWBUSH2004, you would make the difference (less trading costs) in most scenarios. You would only lose your investment if Bush won the presidency without winning Florida. You would double your investment if Kerry won the presidency while losing Florida. Both of those are unlikely, but the second is much more likely than the first.
(11/2 2:22 AM PST) I am looking at the first state-by-state Tradesports results that show Kerry with a probability of more than 50%. The key values are FL at 50.1, OH at 51.1, WI at 41, PA at 28.9, MI at 11.1. Other broad-based movement in Kerry's direction has helped, too. This is also the first period that I've seen where the state-by-state prediction significantly out of sync with the direct contract in this direction. (The direct contract currently gives Kerry a 55.1% chance of victory.)
I've had historical results of this meta-analysis on the cgi page for a while. Here they are on the front page:
(11/1 5:20 PST) Well, I clearly don't understand where Tradesports gets their closing prices. Ohio was given a closing price of 60 this morning. It hasn't traded above 54 any time that I've watched, and I watched the close before going to bed. I've put a default value of 54 in for Ohio, but the historical data line reflects Tradesports' assigned closing price.
(10/31 23:00 PST) I've updated my algorithm for getting an overall probability from the bias analysis. It works very well. I've retroactively updated the historical numbers to use the new algorithm. In this range, it increased Kerry's probability by about 0.5% (because he's the underdog). There has been no more shift from the October surprise. The overall contract and the state-by-state are in sync, with both projecting about 46% probability of a Kerry victory.
(10/30 18:15 PDT) In light trading, both the overall contract value and the state-by-state aggregate have inched toward Bush by about 3%, effectively restoring the bump after October surprise. Kerry's predicted probability of victory stand at 44.5% by the direct contract and 42.2% by the state-by-state meta-analysis. It will be interesting to see how the bettors/investors do in their ability to predict this outcome. News reports are that Democratic turnout is through the roof during early voting. Is that accurate? Is it factored in adequately by the futures market?
(10/29 23:15 PDT) Well, with the release of the bin Laden tape, we had the October surprise, and the Tradesports futures markets took an immediate bump for Bush of about 3%. Remember, that's not a change of 3% in the margin of the vote—just a 3% change in the likelihood that he'd win. It suggests that the futures market saw less than a 0.5% change in the margin of the vote. Even that little bump has largely evaporated over the last six hours. Again, the contract PRESIDENT.GWBUSH2004 showed a much more rapid response to the changing events. The meta-prediction from the state-by-state futures shifted less.
(10/29 7:30 PDT) Last night, there was a strong shift in the future for the overall Presidential outcome at Tradesports. The contract "PRESIDENT.GWBUSH2004" dropped below 50% briefly for the first time since the summer. During this drop, that contract was out of sync with the aggregate probability computed from the state-by-state futures prices. The gap was about 5.6% at closing time this morning (48.9% vs. 43.3% for Kerry). They are back in sync as of this morning (Pacific Time), with a more typical difference of about 1%, both predicting a 48% probability of a Kerry victory.