Wednesday, October 31, 2012

Simpler Math

There's a national-state poll discrepancy. Mitt Romney leads by 0.75 points in the national average and trails by 1.11 points in the state averages. That's nearly 2 points. Right now there 6 states where Mitt Romney trails by 2-3 points in the RCP state averages. So a 2 point bias could be the difference between winning and losing. I've weighed in on it. Now Sean Trende and Nate Silver have.

The national total votes are the sum of the state total votes. It seems silly to remind people of that but some people ignore that when looking at the numbers. I think they do that because they only see the poll numbers, not the raw number of votes. So people grasp for explanations.

Romney is Doing Extraordinarily Well in States That Aren't Being Polled
The explanation for this is that Obama has an amazing campaign and ground game and that Romney will gain more where Obama isn't contesting. There are a few problems with this. First, it's anecdotal. No one knows if Obama's campaign is superior or how much that'd impact the numbers.

There are holes in this argument. As my chart shows, Romney has improved more in state polls for swing states than in other ones. There could be good reasons for this. As I noted, McCain maxed out the White vote in the South, limiting Romney's potential increases. Yes, Obama isn't competing in these states, but then neither is Romney. Romney needs to sell people he's a better alternative. There's only so much you can do without delivering a message. The big red and blue states (Texas, Georgia, California, New York, Illinois, New Jersey) are getting polled and those results are included. Delaware and Wyoming aren't swinging the election that much.

Exceptional Turn-out in Red States
The problems with this is that the pollsters aren't weighing red states heavier and no one is talking about red state turn-out being higher. It's likely to lag since evangelicals that have never warmed to Romney can leave their Presidential vote blank because they know Romney will win their state. Even increasing turn-out by 10% only in red states only moves the totals to Obama +0.82, a small percentage of the discrepancy.

The National Polls Are Wrong
The explanation for that is that state polls did better in 1996 and 2000. Yet in 2004 and 2004 the national polls were far more accurate, indicating that they changed methodology due to that weakness. The national pollsters tend to stay more stable. There are plenty of new state pollsters with unknown track records entering and leaving each cycle. The national polls could be wrong but no one has given an explanation as to a methodological flaw that doesn't exist in state polls.

That doesn't mean that the national polls can't be wrong, but there's just as much of a chance the state polls are wrong, or we meet in the middle. Remember that they all contain a margin of error, so the state polls and national polls can probably meet anywhere within Obama +2 to Romney +2 and there wouldn't be anything wrong with the polling.

What is wrong, however, is assuming Romney wins by roughly a point and the state polls being right. That's even simpler math.

No comments:

Post a Comment