Thursday, November 01, 2012

Under a week out, updating my 2012 prediction

So, 9 months ago, I published this prediction on the 2012 election:

This was of course a best case map as far as I was concerned, but here's what I said at the time :
"I'm pretty confident that, presuming nothing insane, ridiculous, heinous etc... happens between now and november, that I will come within 20 votes, and I am moderately confident that I'll be within 10 (the only state I see as particularly iffy here is New Mexico)."
Well, now we're less than a week out; the election is next Tuesday... And I'm still pretty happy with my prediction from 9 months ago.

At this point, I'm unsure about New Mexico, and Pennsylvania; actually I'm counting them for Obama right now...

But I'm still pretty sure I've got it right within 20 or so electoral votes.

My current assumptions are:

  1. People who will actually vote  for Romney are generally underrepresented in polling (for various reasons)
  2. Pollsters "likely voter" assumptions are significantly off. Most polls have a +4 to +8 democrat bias and, I believe the opposite will be true; with at least a +2 Republican bias, and possibly as high as +10
  3. "Independent" voters are breaking at a +8 to +20 for Romney (depending on the poll)
  4. Almost all "undecided" voters are actually going Romney; unless undecideds are more than 6%, or the margin of error is over 4%
  5. Any state polling under 50% Obama, is going to break Romney, unless undecideds are more than 6% or the margin of error is over 4%
  6. Any state with Romney at 49% or higher is going Romney, unless undecideds are more than 6% or margin of error is more than 4%
  7. Any poll with a margin of error significantly over 4% is basically garbage data, and should probably not be considered, unless the polling margin is higher than the MOE plus 2%. 

Now, as to the polls themselves... well, take a look.

Several national polls now have Romney up over Obama by as much as +5. No credible poll has Obama at higher than 50% nationally, and most have him at 48% or below... And that's BEFORE correcting for democrat biased turnout models.

Still, it comes down to the individual states... and that data is VERY interesting:

I still think 270 to win  is the best poll indexer out there (you can click on the individual states to get multiple results from all major polls in the state).

The most critical states right now are Florida, Ohio, and Pennsylvania.

...Not surprising, since they've been the critical states in the last three elections; and probably will be for the next three elections...

That's part of the reason I am calling Ohio and Florida to break Romney; because NO credible polls with a margin of error less than 4% put Obama at over 50% in either state, and only a few even have him reaching 50% (and again, those are not corrected for turnout model bias).

Although the Obama campaign is representing that they are still competitive in Florida, it's pretty obvious at this point that he's done there. His support has topped out, and even with the bad turnout models, he's well under 50%. Not one single poll has Obama even at 50%, even with the biased turnout model. Most have him at 47% or under.

Ohio is still competitive for Obama, with a couple of polls having him hit 50%; but none are showing over 50%, most are 48% or under, and those that show 50% are higher margin of error polls.

So yes, Obama is still competitive in Ohio, but I'm confident in calling it for Mitt.

The only real question is Pennsylvania.

Only four major polls currently show Romney over 50% in Pennsylvania. Of those, one shows him at 52% and three are 51%. All the polls showing Obama at 48% or over, have margins of error over 2.5%. All the polls showing Obama at 50% or over have margins of error over 4%.

So, what's the big picture?

Note: The 8 states with 1 million population or less are generally going to have a very high margin of error in polls; both because the polling companies spend less money on those states (they have less electoral votes), and because the sampling methods polling organizations use are actually LESS accurate, for very small population.

Also, for cultural reasons, most of those states (5 of the 8 are very conservative or libertarian) generally
 have poorer response rates to polls; and tend to have a higher number of "independent" voters, and a higher number of "undecided" voters... ; all of which tend to increase the overall margin of error. 

Given the best poll data we have right now...

That set of assumptions above gives Obama the following states:

New York
Rhode Island
New Jersey

That set of assumptions gives Romney the following states:

North Dakota
West Virginia
South Dakota
North Carolina

These states have a margins under 2%, margins of error over 4%, or undecided over 6%, but I think will break Obama:

New Mexico (New Mexico polls are incredibly inconsistent with huge margins of error and undecided)

These states have a margins under 2%, margins of error over 4%, or undecided over 6% and I think will break Romney:

South Carolina
New Hampshire (the margins of error are concerning me, but the trend is Romney)

That scores out as 291 Romney, 247 Obama

Note, Romney can lose both Pennsylvania and Ohio; so long as he wins Florida.

He can lose Florida, if he wins Pennsylvania and Ohio (and Nevada, New Hampshire, and Iowa as I've projected. Some people still think Iowa and Nevada are competitive states for Obama. I'm not one of them).

If Romney takes Wisconsin or Minnesota (I don't think it will happen, but it's possible), it's basically over.

In fact, right now, there are far more ways for Obama to lose, than there are for him to win... and only a couple realistic ways in which he would win.