Tuesday, September 09, 2008

Eight weeks out

One thing I've been rolling around in my mind is taking a weekly look at where the election stands, not from the minutiae of daily polls, but in a more comprehensive view. Clearly, we know that the national polling data is irrelevant except in how it frames the story of the election process. Right now, the GOP is riding on the crest of it's post-convention bounce (the democratic bounce got squished by the compression of the conventions as well as the Palin announcement, as McCain knew it would), and the scenarios are looking better for them now than they once did. We will see if that continues after the bounce settles next week.

What I've done is look at two websites doing projections - fivethirtyeight.com and electoral-vote.com. (Incidentally, 538 is run by Nate Silver, one of the wizards behind baseballprospectus.com, my favorite baseball analysis site.) E-V compiles polls and comes up with a composite spread (using a rather straightforward system). 538 is more complex, with a model they're still tweaking, and they assign not only a weighted average of polls (in part based on which ones tend to be reliable) but also a percentage likelihood of victory.

For example, right now, E-V gives Obama a 7 point win in Oregon, while 538 gives him a 7.4 point win and an 88% chance of winning. I put this data together and sorted it into five categories:

1. McCain will win barring something bizarre - McCain has a polling lead of at least 5 points in both models as well as a greater than 80% chance of winning.

2. McCain is likely to win - at least one of the three criteria above is met, but not all 3.

3. Obama is likely to win - at least one of the three criteria above is met, but not all 3.

4. Obama will win barring something bizarre - Obama has a polling lead of at least 5 points in both models as well as a greater than 80% chance of winning.

5. Tossup - the other ones.

To see if this makes any sense, let's first look at which states fall in which categories:

McCain win (in descending order of certainty): Utah, Wyoming, Oklahoma, Idaho, Tennessee, Alabama, Kentucky, Kansas, Nebraska, Arkansas, Louisiana, Mississippi, Alaska, South Carolina, Arizona, Texas, Georgia, West Virginia, Missouri. Total: 165 EV.

McCain likely: South Dakota, North Dakota, Indiana. Total: 17 EV.

Obama likely: Michigan, Pennsylvania, Washington. Total: 49 EV.

Obama win: DC, Vermont, Rhode Island, Hawaii, New York, Illinois, Massachusetts, Connecticut, Maryland, Maine, California, Delaware, New Jersey, Wisconsin, Iowa, Minnesota, Oregon, New Mexico. Total: 211 EV.

Tossup: North Carolina, Montana, Florida, Nevada, Ohio, Virginia, Colorado, New Hampshire. Total: 96 EV.

So, what do we think? It looks pretty good to me. It's a little odd seeing New Mexico in Obama's column, but that state has evolved recently, and other than that, there's nothing on the Dem side that surprises me. Obama will do very well in Iowa (where he has an excellent ground game) and states like Wisconsin and Minnesota, where they're sick of Bush, and despite McCain's efforts, aren't really in play. I also don't see a chance that Washington goes for McCain.

I'm going to work through two scenarios - one pretty balanced and one heavily tilted towards McCain.

Scenario 1: All the "likely" or "win" states go in the expected direction. Before the toss-ups, it's Obama 260, McCain 182, meaning that Obama needs 10 more for the win. What are the odds of that? Treating all the toss-up states as 50-50 to go each way, there are 2^8=256 possible combinations. The only combinations where Obama doesn't get to 10 EV's are as follows:

Obama wins 0 of the tossup states (1)
Obama wins only one of the following: MT, NH, NV, or CO (4)
Obama wins two of the following three: MT, NH, or NV (6)

That gives Obama a likelihood of winning of 245/256 = 95.7%.

Scenario 2: McCain gets all his likely states, plus stealing Michigan and Pennsylvania (and getting Ohio in the process). Obama does hold on to Washington. McCain 237, Obama 222. In this model, there are 128 combinations, and Obama would need 48 EV (of the available 76), and since Florida is worth 27, the only way he can win without Ohio, Florida, Michigan, and Pennsylvania is to win NC, MT, NV, VA, CO, and NH. Putting it all together, he'd have a 34/128 = 26.5% chance.

What does that tell us? At this point, McCain should be hitting the PA, MI, OH, and FL hard. Very hard. If he can grab all 4, or even 3 of 4, he's got a real shot. If he only takes two out of the 4 (say Florida and Ohio, two pretty big prizes), Obama needs only 10 more EV to win, and winning NH and CO (which would both be in his column right now, even with the GOP bounce) would put him over the top.

1 comment:

I'm Not Ned said...

Nice analysis.

Unfortunately not all of the states polling data has been updated since the conventions and may not represent any VP/convention bumps.

If PA and FL go extremist we're screwed.

I'm really scared.