Lies, Big Lies and Polls

Tuesday, April 10, AD 2012


Ed Morrissey at Hot Air for years has done yeoman work in examing polls minutely and he does this well today in examing an ABC Washington Post poll with purports to show Obama leading Romney 51-44.

I love the Washington Post/ABC poll.  It’s a great object lesson in how to manufacture news.  Need a story that the incumbent President’s fortunes are looking up?  Well, just adjust the sample a bit and voila, he takes a seven point lead over his presumed rival in the fall election!  Besides, it gives me fodder for snarky material every few weeks.

Let’s get down to cases, shall we?

With the general-election campaign beginning to take shape, President Obama holds clear advantages over Mitt Romney on personal attributes and a number of key issues, but remains vulnerable to discontent with the pace of the economic recovery, according to a new Washington Post-ABC News poll.

Obama has double-digit leads over the likely Republican presidential nominee on who would do a better job of protecting the middle class, addressing women’s issues, handling international affairs and dealing with health care.

You know where else Obama got a double-digit lead?  In the polling sample.  In 2008, when Democrats surged to the polls after eight years of George W. Bush, CNN’s exit polls showed a seven-point advantage for Democrats, 39/32, which mirrored Obama’s seven-point victory in the popular vote.  In 2010?s midterms, CNN exit polls showed a 35/35/30 split.  By contrast, the previous WaPo/ABC poll in March had a D/R/I of 31/27/36, which undersampled both parties relative to independents but left Democrats with a 4-point advantage — perhaps an arguable model for 2012 turnout.  Today’s has a D/R/I of 34/23/34, adding seven points to that Democratic advantage and presenting a completely unrepresentative, absurd model for the 2012 turnout.

Go here to read the brilliant rest.  In any poll the crosstabs, which give you information about the makeup of the people polled, is always the most important part.  If the crosstabs do not give an accurate snapshot of the electorate, the poll is meaningless.  One of the most important aspects of the polling sample is the partisan affilation of those polled.  That can be difficult to factor correctly as the political strength of the parties shifts from year to year.  Rasmussen, who nailed the 2008 presidential outcome, pays close attention to party affialiation and polls monthly on this question with a huge sample of 15,000 adults.  His latest poll indicates 36.4% Republicans and 33.4% Democrats.   His tracking poll for today shows a deadheat between Obama and Romney at 45-45.  There will be a huge number of polls this year, and most of them, as reflected in their crosstabs, will be junk.

One Response to Lies, Big Lies and Polls