Election Projection, Final Version

(Basically, I updated the graphic with the new poll data and backed out on leaning Missouri to Obama)

Yesterday Karl Rove spoke at Washington University. I wanted to go but didn't. But as I remarked to my friend, "How appropriate it would be to go see Karl Rove speak at the end of his era, and the next day vote in the dawn of the next era."

It's not news, but Barack Obama will become the 44th President of the United States.

There will be more polls so I won't call this my last update, though I may not have time for another. Here's the analysis.

I maintain that Barack Obama's core of 311 EVs (Kerry+IA+NM+CO+VA+NV+OH) is safe, even though Ohio has tightened a bit recently. I'd have to say I can't count McCain's core as being any larger than 132 EVs. It is unlikely that he will lose any of ND, GA, AZ, or MT, but the numbers just aren't there to call those EVs safe for McCain.

The remaining (non-safe) states are:
Florida (27)
North Carolina (15)
Missouri (11)
Indiana (11)
North Dakota (3)
Georgia (15)
Arizona (10)
Montana (3)

Florida - I would love to call Florida for Obama. It's just a whisper away, but I just can't do it. Four days ago it would have been all but locked up, but there has also been tightening here. Rasmussen, one of the best pollsters, went from Obama 51 McCain 47 on 10/26 to McCain 50 Obama 49 on 11/2. That's just one poll, but to really look for trends it's best to watch how numbers move within a pollster. Strong Obama.

North Carolina - More than any other state, North Carolina's result will be determined by turnout. There's apparently some rain there today, which is apparently bad for Democrats. I show Obama with a significant but very slight lead. The latest poll in NC is from American Research Group, whom I don't trust too much. The four polls before that show McCain ahead, albeit by 1, 1, 1, and 3 pts. Weak Obama.

Missouri - The show-me state has become the ultimate swing state. As I'm sure you've heard before, Missouri is (I believe) 25 for the last 26 elections in going with the winner (went against Eisenhower's re-election in '56) but this year runs a little redder than the median, though when Obama surged a month ago, Missouri moved bluer faster than most states. That trend has settled down, and of the last 6 polls in the state, there are four ties, one Obama 1 pt lead, and one McCain 1 pt lead. There are few areas in the country as poor as St. Louis City in running an election (in 2004 they had to extend voting hours to 10:00 PM from 7:00 PM) and the results always come in late, so MO will be among the later states called. I just don't know what to do with this state. Due to the confusion in the city, voter purges might be possible, but on the other hand I expect turnout to be very high. My precinct is largely African-American, and we were literally lined up by the hundreds before the polls even opened. My gut tells me Missouri might stay red, but all my objective indicators give an ever-so-slight advantage to Obama. (Very) Weak Obama. Weak McCain.

Indiana - For all the talk of Indiana's historical red-ness, Bill Clinton only lost the state by 4 in '92 and 5.5 in '96. Popular Senator Evan Bayh was on the VP shortlist, and the state overperformed for Obama in the primaries. On the other hand, Obama's only led in 2 of the last 9 polls (in fairness, McCain only has in 4 but his leads have been larger) and the state never really shifted too far after the economic crisis. Since the state has been underpolled I tried turning down the sensitivity, but it still just shows a big swing state mess. I may very well come back to this one, but for now I'll say that the first indicator of the evening won't lead to a 6:30 concession speech. (Very) Weak McCain.

North Dakota - This state, part of the rural west, has suffered from a mighty dearth of polling. Indicators show it will be surprisingly close, but Obama withdrew from the state during the Palin bounce and he underperformed during the primaries. Weak McCain.

Georgia - Georgia, I feel, is just being a tease. The race there has tightened very dramatically, but McCain continues to average over 50% - a key threshold. While reports indicate that early voting turnout has been exceptionally high, Georgia would be an upset for the ages. Strong McCain.

Arizona - There are more undecideds in John McCain's home state than in Georgia, so McCain is not averaging 50% here even though his lead is about the same. But if you live in Arizona and haven't decided who you're voting for, I can't help but feel like that's not good for John McCain. Still, Obama's never led a poll here and underperformed during the primaries. Strong McCain.

Montana - This state has been a big tease for Dems all year. It also suffers from a dearth of polling; nonetheless, Obama's only lead in four months came from a university-comissioned poll, which are notoriously untrustworthy. Still, McCain isn't at 50% here either. One other issue is that Ron Paul has made the ballot in MT. Apparently I'm the only person in the country who thinks that this factor is likely to pull more votes from Obama than McCain, as Obama/Paul (now there's a ticket) shared the same young, energized voter demographics in the primaries (and their fundraising). As in North Dakota, Obama also pulled his resources here after the Palin bounce, which you may remember was particularly strong in the rural west. Strong McCain.

Final Analysis - I hate to call MO and IN one way or the other. The other states here could very easily be up for grabs, but also definitely have a lean. I'm really growing irritated by all these sites that use real statistics in their models, but then won't make any calls, or are very conservative. Who does it help to call 105 EVs "pure tossups" on the day of the election? Of course, I don't have much to risk if I'm wrong, but a four point lead on the basis of a weighted average of 30 polls in the last month is not a statistical tie.

Anyway, the way it's currently added up shows Obama 364, McCain 174 Obama 353, McCain 185 with Obama 353 McCain 185 Obama 364, McCain 174 and Obama 375 McCain 163 being the second- and third-most likely scenarios.

I project the popular vote to be Obama 53.1% to McCain 45.9%. Obama has currently led 278 consecutive polls in my model dating back to September 15. He is up 371-32 in the signs test and 15.8-0.0 in the weighted signs test. The polling averages are Obama 50.9% McCain 43.9%, which projected forward give you the bold numbers above.

The best guesses I have for state-by-state vote projections are given in the following table. I'm sorry it's messier than ever. There are no fudge factors in the following projections (i.e., I can't look at one state and say "hey, that should be 0.6 pts higher") but there is still subjectivity in terms of how sensitive the model is to new information (currently I give a poll a half-life of 5 days, down from 30 in June).

Note: I called Indiana above for McCain even though the math predicts (an insignificantly small) Obama victory, so my 353-185 differs from the graphic's 364-174.

Also, if you'd like a copy of my spreadsheet with all the data, I'd be happy to email it to you.