Today the Romney campaign is trumpeting the latest CNN/ORC poll which shows President Obama with a three-point lead over Mitt Romney among likely voters, 50%-47%. The Romney campaign likes this result, apparently, because it represents a significant improvement over the last CNN/ORC poll, which came out three weeks ago. In that survey, CNN/ORC found a six-point Obama lead.
I wrote about the earlier poll here, pointing out that it obviously over-sampled Democrats. A reader calculated that, given other data in the survey, the six-point difference was consistent with a breakdown of D-38%, R-26% and I-36%. I wrote that the most significant point in the CNN poll results was that independents favored Romney over Obama by a remarkable 14 points.
So today’s CNN/ORC poll has Obama’s lead cut in half, to three points. Does that mean that Romney is surging? Not really. It means they didn’t call quite as many Democrats. This time, the pollsters gave us the partisan breakdown of their sample: D-37%, R-29% and I-34%. So when the partisan gap in the sample narrowed from D +12 to D +8, Romney did three points better. Well, yeah.
Is either of these sample compositions plausible as a representation of the 2012 electorate? Plainly not. Currently, Scott Rasmussen finds that more voters identify themselves as Republicans than Democrats, and all surveys report that intensity is at least as high among Republicans as among Democrats this year. So who could possibly believe that there will be 8% more Democrats than Republicans at the polls in November?
Polls like the ones conducted by CNN and ORC are useful primarily as an indicator of which party’s members are more likely to answer their telephones when they are called by strangers.