Did CNN Rig Its Own Poll?
CNN is grabbing political headlines tonight with the release of its latest poll. It shows Obama surging to a 6-point lead over Romney, 52-46, among likely. Before the start of the Democrat convention, the candidates had been tied in the poll. Since it purportedly confirms a narrative the media is trying to build, i.e. that Obama is starting to pull away with the race, it is getting wide coverage. However, there are a couple of strange things within the poll that cast doubt on its veracity. And, at least one concern warrants a response from CNN.
First, this being a media poll, it has an obvious skew towards Democrats. The partisan breakdown is (D/R/I) 50/45/5. It perhaps isn't surprising that Obama is leading a D+5 poll by 6 points. Throughout the campaign season, Obama's margin usually is very close to the partisan skew in the sample. It is surprising, though, that Independents make up only 5% of the sample. Tellingly, Romney leads this group by 14 points.
I've seen others suggest that CNN pushes voters to identify with one party or the other, which may account for this. Although, CNN identifies the breakdown of sample numbers as X "registered Democrats" and Y "registered Republicans." Other polls that push Independents to identify show their work. CNN should as well.
There is another, more serious concern, however, The second question of the poll asks Obama and Romney voters whether their vote is "for" their candidate or "against" the other candidate. The sample size for this question is reported to be (top of page 3):
BASED ON 351 LIKELY VOTERS WHO PLAN TO VOTE FOR OBAMA -- SAMPLING ERROR: +/- 5 PERCENTAGE PT
BASED ON 340 LIKELY VOTERS WHO PLAN TO VOTE FOR ROMNEY -- SAMPLING ERROR: +/- 5.5 PERCENTAGE PTS
Those numbers didn't look right to me, considering the headline number reported in the poll. There were 709 Likely Voters in the sample. The 351 votes for Obama is 49.5%. The 340 votes for Romney is 47.9%. In other words, a 50-48 match-up. That's a lot different than the 52-46 reported in the headlines. Consider: the question to Romney supporters samples more Likely Voters than he received in the head-to-head result.
What could be happening here is that these are the raw counts from the poll interviews, which CNN then "weighted" to reflect certain demographics, etc. Pollster do this often, but it usually results in very minor changes to the overall numbers. If this discrepancy is the result of CNN "weighting" their poll, it would reflect a 4 point swing in the overall results. That's a very significant change in the outcome. Their weighting added 2 points to Obama's support and erased 2 points from Romney's. Combined with a D+5 sample, this is a 9-point edge for Obama, simply from the composition of the poll.
Again, it isn't unusual to weight a poll, for very legitimate reasons. But if the weighting produces this big of a swing in the final results, you really ought to show your work. CNN doesn't release its internal demos, but it absolutely needs to in this case. If they don't, then they've just been caught with their thumb on the scale.
Follow me on twitter.