Are the opinion polls biased on the US elections?
If Obama's supporters look for good news from the polls, they'll find it. Pradoxically so can Romney's. Can the polls be trusted?
As if it wasn't confusing enough for non-Americans trying to work out who's going to win the elections on Tuesday -- who outside the US really thinks they have a handle on what's happening in the swing states? -- the problem I'm having is interpreting what is supposed to be the easy stuff, the national opinion polls.
At least twice a day I check out the must-read Real Clear Politics national average of 10 top polling agencies. Here's what it is giving me right now (approx 0800 GMT, Saturday, Nov 3):
So, as you can see, Obama has a 0.1 percentage point lead when all of the polling is aggregated. Obviously, that's negligible given the margins of error, but it has to be said that Obama was a percentage point down just a few days ago.
That would suggest the momentum has moved a bit in the president's direction, though not enough for him to secure the White House.
But here's where I run into problems: can the opinion polls be trusted? Part of it is that two of them give five percentage point leads to both Obama and Romney, National Journal and Gallup respectively.
You could say that the best approach is to knock out the two polls on either side that are most at variance with the rest. All eight of the remaining agencies either give a one percentage point lead to one or other candidate, or record a tie. That essentially means it's a dead heat, as most of the headlines are suggesting.
But there is a school of thought that the polling, like most of the mainstream media, has been infected by a pro-Obama bias. This has seeped its way into the methodology giving Obama a boost in the polls that he won't get at the ballot box. A lot of this has to do with the sampling used among particular segments of the voters and assuming that they will turn out to vote in the same proportions they did in the last election.
It's complicated. But Karl Rove explains the point (in an article arguing that Romney will score an overall win of 51 percent to 48 percent) in the context of turnout patterns in the key swing state of Ohio:
"Desperate Democrats are now hanging their hopes," he says, "on a new Quinnipiac University/New York Times/CBS News poll showing the president with a five-point Ohio lead. But that survey gives Democrats a +8 advantage in turnout, the same advantage Democrats had in 2008. That assumption is, to put it gently, absurd."
Here's another fascinating extract from a piece in The Hill, explaining just how complex the situation really is, this time looking at voting patterns among ethnic groups:
"No one yet knows the makeup of the 2012 electorate.
"When the Census Bureau compared the 2008 electorate against the 2004 electorate, for instance, it found that in the more recent election 2 million more African-Americans and 2 million more Hispanics had voted, in addition to about 600,000 more Asian voters. 'The number of non-Hispanic white voters remained statistically unchanged,' the Census Bureau reported.
"That finding alone raises questions to which no one really knows the answer: Was the higher number of black and Hispanic voters in 2008 overwhelmingly a consequence of excitement about Obama’s historic candidacy that year, and if so, might that excitement have faded? Was the Hispanic increase instead a consequence of demographic changes that might have kept pace or even accelerated since then?
"Each campaign, naturally enough, is trying to spin the situation to its advantage..."
As the author of the piece, Niall Stanage, concludes with reference to Obama's and Romney's campaign pollsters and their different reading of the data:
"One man will be right and one wrong next Tuesday. Accuracy could mean the difference between winning and losing the election."
It could also seal the credibility of several polling agencies for many years to come.
We are wholly dependent on the kindness of our readers for our continued work. We thank you in advance for any support you can offer.