September 26, 2012

The science of unskewed polls

Pollsters are ubiquitous these days.  Sure, they're always polling all the time, but right now their presidential polling is high profile media stuff.  That's true because (1) there happens to be a presidential election coming up, and (2) the polls seem to indicate that Obama is winning, and you know the mainstream media is loving that.  Of course they'd want to highlight that.    Conservatives have been right to question the veracity of the polling that in many cases is sampling voters using a partisan split that unfairly penalizes both Republicans and Independents in favor of a larger Democratic voter sample within the poll.  Intentionally or unintentionally there is a bias built into many of these polls.

Photo via Examiner.com
A poll for example that shows a partisan split among likely voters of Democrat/Republican/Independent of 40/33/28 is classified as a D+7 poll.  The exit polling in 2008 when Obama won in a landslide, was roughly +7, at the apex of his approval.  In the 2004 mid-term elections the D/R/I split was 35/35/29.  Therefore a poll based on a sample of D+7 is going to incorporate a bias into the results.  Of course Democrats are going to skew in favor of Obama. In fact many polls even are weeding out Independents, as they appear to be breaking for Mitt Romney this cycle.

There's a great post about it here (at Battleground Watch), and continued here that goes into more detail about the bias.The point is that the polling is biased, with intent - the intent is to demoralize the right into not voting, thereby raising the possibility of an Obama win in November.

Nevertheless, I'm just as skeptical about the reverse view of the landscape as well.  Yes, I do believe there is a lot of incorrect sampling going on, but just as this is not 2008, it's not exactly 2010 either.  It's likely the partisan voter split is somewhere in the middle, say about D +3.5%.  Yesterday I took a look at Unskewed Polls, and while I know the rationale behind it, bench marking everything using the weightings of Rasmussen is probably overly optimistic. That's putting all of your eggs in one basket, and if Rasmussen is wrong then the "unskewed" results are wrong, as much as I'd like them to be right.

In discussing the skewed polls, the website Examiner points out,
Democrats will have to have a four percent edge among the voting electorate, meaning 37 percent Democrats to 33 percent Republicans, for Obama to win the popular vote by a very narrow 50.40 percent to 49.60 percent majority.
That's close, and it's close to my back of the napkin estimate of D+ 3.5%. I think this is going to be a close election nationally, although on a state by state basis, and the electoral college level, just about anything can still happen.
Related Posts Plugin for WordPress, Blogger...

Share This