(Dark Blue (253): Obama +7.0% or more
Lean Blue (96): Obama +2.6%-+6.9%
White / Toss-up (23): Obama +2.5% to McCain +2.5%
Lean Red (12): McCain +2.6%-+6.9%
Dark Red (154): McCain +7.0% or more)
Obama's position continues to be very strong in my presidential forecast. While his "solid" states have dropped below 270, he still goes over 270 in Minnesota, where he leads by 6.3%. Also, while Florida and Ohio are narrowing, Obama's leads of 3.5% and 3.3% respectively are still means that there is roughly a 90% chance that he is ahead in those states. Obama has also managed to push NE-02 into "Lean McCain," and North Dakota into "toss-up" status. So, things are looking good.
I am expecting the campaign to settle in at an Obama lead of 4-5%. I looked at Pollster.com's national trendlines today, excluding the VP and convention period from August 16th to September 15th. Without that period in the trendlines, the campaign has been extremely stable, hovering at right around 4-5%. Obama was a bit higher right after securing the nomination, and also at the height of the financial meltdown. McCain was a bit higher when he was able to outspend Obama on paid media in all but a few states. But really, this is a 4-5% campaign, with very little movement. Given the long-term stability excepting for major events, I expect that 4-5%, is where the campaign will end, with Obama winning between 325-350 electoral votes.
Also, on a methodological note, I have decided to exclude Zogby Internet polling from my averages for the remainder of the campaign season. While the rapid increase of cell phone only households require Internet polling to be the wave of the future, past Zogby interactive performances, combined with consistent outlier polls, just make me extremely uncomfortable with including them in my averages. The test I did on my methodology did not include any Zogby interactive polls, so I wouldn't even know if my method worked if Zogby was included.
* Maine's four electoral votes are awarded as follows: two for the statewide winner, and one for the winner of each congressional district. ME-01 is about 3.2% more Democratic than the state as a whole, while ME-02 is about 3.2% less Democratic than the state as a whole.
** Nebraska's five electoral votes are awarded as follows: two for the statewide winner, and one for the winner of each congressional district. NE-01 is about 6.0% more Democratic than the state as a whole, NE-02 is about 11.0% more Democratic than the state as a whole, and NE-03 is about 18.0% less Democratic than the state as a whole. NE-02 is different right now, since there is an actual poll of the district.
Methodology The forecast is updated whenever there are enough new polls to warrant an update. The methodology for the forecast is simple and straightforward.
For each state, all polls with the majority of their interviews condcuted within the last 7 days from publication of the forecast are included in the averages. There are some necessary exceptions to this rule.
Every state has at least three polls, so the first exception are states without three polls that meet the previous criteria. In the event that three polls taken within the past week are not available, the three most recent polls are used. (Note: "Most recent" is determined by the mid-point in the dates of the polls. For example, a poll taken from 9/15-9/17 has a mid-pint of 9/16. In the event that more than one poll shares the most recent mid-point (within 0.5), all polls with that mid-point are included in the average.)
The second exception comes with polling firms that have conducted more than two poll in a state over the last 7 days. Every polling firm only has one poll per average.
Almost no polling firm discrimination whatsoever. Polls are never excluded because the organization has a bad or partisan reputation. Non-random Internet sample polls are excluded, however.
No pre-September polls. I'm just sick of including them.
During the final two weeks of the election, only polls taken entirely within the past seven days will be included in the averages (with the same exceptions listed above, of course).
