Pollster.com’s Mark Blumenthal rejects the contention from Asst. Prof. Todd Eberly, linked here, that the Newsweek poll is “cooked.” He concedes that Newsweek’s story and headline are fair game for criticism, but declares the accusation of book-cooking “nonsense.”
As I told Blumenthal in a message, Eberly may have been a bit vociferous in his criticism — “cooked” goes to motive, and I prefer not to argue about motive or what’s going on in somebody’s head — but I look at his defense of Newsweek and have a few bones to pick.
First, across most pollsters, we’re seeing a pretty consistent pattern: Republicans overwhelmingly support the Republican candidate, Democrats almost as overwhelmingly support the Democratic candidate, and the story of this year has been independents breaking more and more heavily for the Republicans. In most states, races, and polls, this adds up to a lead for the GOP candidate.
In Newsweek, they actually find the same phenomenon within the subgroups… but still end up with a Democratic tie; as laid out, the only way to get this result is to have a sample that is heavily Democratic. The Newsweek poll has one other wrinkle, a separate “other” category. I think a lot of folks in political circles use “independent” and third-party interchangably, and they probably shouldn’t. But I look at this “other” category, and it skewing in the precise opposite direction as the independents, and rub my chin. Are these folks Greens? Left-leaning registered members of the Libertarian party? I’m not saying it’s certainly wrong, I’m just saying it’s surprising, and as Blumenthal mentions, a lot of pollsters would toss out this category of 45 extra voters. With this result, I suspect they do so for a reason.
Another major quibble… these “other” folks are going to make up five percent of the voters who show up in a midterm? Really? I went to look back at the 2006 exit poll, and found CNN didn’t even have the category; voters were classified either Republicans, Democrats, or independents.
Also note this Newsweek poll is of registered voters. We’re about 60 days out from the election. From where I sit, polling organizations that can’t be bothered to try even a broad likely voter screen are wasting people’s time and money. Public Policy Polling applied a likely voter screen to their samples in Pennsylvania and Toomey jumped up from a tie to a 9 percentage point lead. A lot of registered voters are Democrats who only vote when it’s a presidential year and/or some particularly big or dramatic race. For Newsweek to not even bother… well, I said I don’t like arguing about other people’s motives, because Newsweek’s use of a registered voter sample suggests they’re not that interested in a realistic portrait of the November electorate, it’s easy to wonder if they just want to find a result that doesn’t look as bad for Democrats.
Finally, a party ID split of 36 percent Democrat, 32 percent Republican, 27 percent independent, 5 percent other/none/unsure may be a narrower partisan split than the 2008 exit poll numbers, but I think it’s probably understating GOP turnout this fall and perhaps by a lot. As we’re seeing in the Gallup generic ballot poll and elsewhere, the GOP base is mad as hell and can’t wait to vote.
UPDATE: A reader notices that the Gallup generic ballot poll is also registered voters, not likely voters. Elsewhere, others are asking if this means the GOP is more accurately described as a 14 percentage point margin.
It’s worth noting that Gallup does poll likely voters later in the election cycle. Newsweek’s article, touting a tie on the generic ballot (still historically bad for Democrats, BTW) never mentions the factor of registered vs. likely voters, and few of the optimistic Democrats want to even acknowledge that factor.