Poll cooking season is officially in full swing. The headline today at the Washington Post reads: "Among likely voters, Obama-Romney close." Dan Balz and Jon Cohen report that in a September 7-9 poll, "the (presidential) race remains close among likely voters, with Obama at 49 percent and Romney at 48 percent, virtually unchanged from a poll taken just before the conventions." Ah, but Obama supposedly has a six-point lead among registered voters.
Based on pair's report, the easy choices on how to interpret the results are these: Either President Obama really didn't come out of the Democratic Convention with a polling bounce, or, if he did have a bounce, it disappeared after last Friday's dreadful employment news. There's a third and far more likely choice, which only becomes apparent once one sees the mix of respondents in the poll's final listed question.
Here is that question:
Well, no wonder it's close. The poll's mix of registered voters was 32-26 Democrat-GOP, while its likely voters were 33-27. There is no evidence that the pollsters adjusted the results to reflect a more accurate breakdown of the electorate. If they come back and say that they did, the obvious response is "Well, you should have said so in the first place."
Because of the poll's all too predictable skewing towards Democrats, it's likely that Romney is ahead among likely voters and barely behind among registered voters.
The most recent Rasmussion party-ID poll showed a breakdown of 37.6% Republicans and 33.3% Democrats. The most recent such attempt by Gallup shows 31% of Americans identifying as Democrats and 28% as Republicans.
The WaPo-ABC poll shows Democrats favoring Obama by 94-5 and 91-5 among likely and registered voters, respectively, and Republicans favoring Romney by 91-6 and 89-6. From there, it's relatively easy to estimate that if the pollsters had instead sampled equal percentages of Democrats and Republicans in both instances, Romney would have a four-point lead among likely voters of being a point behind, and would only be behind by one point instead of six with registered voters.
The poll's overall mix of respondents, including those not registered to vote, was 33-23 Democrat-GOP. The poll didn't present their results for presidential preference.
The poll is also marred by bizarre questions asked of registered voters which, if asked before expressing a presidential preference (the order in which the questions is listed is not necessarily the order in which they were asked), seem to have been designed to skew more favorability in Obama's direction. Examples:
- Who do you think would make a more loyal friend?
- On a ship in a storm, who would you rather have as the captain?
- Who would you rather invite to dinner at your home?
- Who would you prefer to have take care of you if you were sick?
Silly me. I always thought that the answer to the last listed question above was "a doctor, nurse, or loved one," not a presidential candidate.
What garbage. Since phone utilization is virtually free any more, the questions could also arguably be targeted at getting sensible people to hang up the phone in disgust while allowing those who have too much patience for this garbage and a lot of time on their hands to stick with the survey. In my view, those who remain end up being more than a little unrepresentative of the general population.
The only thing that surprises me about this fundamentally dishonest poll is that WaPo and ABC disclosed the party mix of those surveyed at all.
Cross-posted at BizzyBlog.com.
