JWC InTrade-Based 2008 Presidential Election Oracle!

It's back, third time up to bat, and (perhaps) more interesting than ever.  That's because the 2008 U.S. Presidential Election promises to be complex on several fronts: the unusual rancor among the Democrats; the fairly wide field of plausible Republicans; and the ever-possible chance of a third party candidate emerging to complicate the situation for one or both of the mainstream parties.

For details of the model and its basic premise, please see its description on the 2004 Model Page.  The 2008 concept is the same, although the TradeSports betting site has shifted its political contracts trading to a website called InTrade; the latter is where I get the raw state contracts data, usually gathered after 6 PM EST.  You also might want to look at the model's predictions for the 2006 mid-term senatorial elections.

I've decided to start the model fairly early in 2008, just after the Iowa and New Hampshire primaries.  What may be worth watching for are "phase transitions" in the electoral votes predictions; these might result from wagering discontinuities just after important primaries, such as the "Super Tuesday" February 5th multiple state event.  See this website for a list of all the primary dates, and this website for all you'd ever want to know about the various state primaries.  Obviously, the nominating conventions (August 25-28 for the Democrats; September 1-4 for the Republicans) may generate some notable wagering shifts.

So forget those edgy "talking heads" and polls; just check the web-site!  Without further ado, here is my daily predicted outcome .

September 2008.  This has turned into a very interesting, complex election; it seems that a bit more analysis might be of interest.  In addition to the daily predicted outcome, I'm adding a plot to show how the so-called "18 Obama Battle Ground States" (AK, CO, FL, GA, IN, IA, MI, MO, MT, NV, NH, NM, NC, ND, OH, PA, VA, WI) compare with the non-battleground states.  Here is the battleground plot.

Happy Voting!

---

November 7, 2008
Well!  After its third exciting venture into the misty realms of political prediction, the wagering/market-based model has another win, predicting 353 to 364 electoral votes for Obama and 185 to 174 for McCain between the last 4 days up to the election.  We now know the final results of the Election, with Mr. Obama dealing Mr. McCain and several minor candidates a decisive defeat, as predicted.  Interestingly, the model never showed the Republican rising above the Democrat during its 300 day run, a trend that held well before either party identified the final candidates.

The 2008 Official Presidential General Election Results can be obtained from the Federal Election Commision website.  Here is a summary:
Obama:      69,456,897 votes (access to 51 states or districts), with 365 Electoral Votes
McCain:     59,934,814 votes (access to 51 states or districts), with 173 Electoral Votes
Nader:             738,475 votes (access to 46 states or districts) , with 0 Electoral Votes
Barr:                523,686 votes (access to 45 states or districts), with 0 Electoral Votes
McKinney:      161,603 votes (access to 32 states or districts), with 0 Electoral Votes
13 Others:        approx 164,000 votes, with 0 Electoral Votes.

See USA Today for a good map.  So the InTrade-based model again performed quite well, to the bitter end.  Of course I took plenty of data throughout November 4, and will post further analysis in the near future.

One final note.  During the early evening of the election I was very perplexed by substantial inconsistencies between certain state contracts (Virginia and Indiana) and the early returns being reported for those states.  For example, McCain appeared to have a real edge over Obama with over 50% of the Virginia precincts reporting, yet the InTrade contract stayed quite stable in favor of Obama.  However, after 9 PM EST, as results from the more urban precincts in Northern Virginia began to come in, the veracity of the trading model was borne out.