Jensen told NRO that the firm weights its samples only for gender, race, and age. “I think the 2008 vote is a better gauge of the representativeness of a poll than comparing the party ID numbers to the 2008 exit poll, but we don’t weight for that either; we just put it out there for people to see,” he said. “Last year, Democratic turnout was incredibly depressed in Virginia, but much less so in New Jersey. It would have been a mistake to weight those states to 2008 party ID or presidential horse-race numbers because the level of engagement is just different in different states.” Jensen can find lots of reasons why Democratic turnout may not be so bad in some states: “In places like Texas, where Democrats have a great candidate in Bill White, and Kentucky, where Rand Paul has a polarizing impact, we’re not seeing much falloff in Democratic interest from 2008. In Wisconsin, Pennsylvania, and Iowa, which all have very unpopular Democratic governors and where the president’s popularity has really declined, we are seeing big dropoffs in Democratic interest, similar to what we saw in Virginia at this time last year. As a whole, Democratic turnout will certainly be down from 2008, but that’s going to be true to different degrees in different states, and you can’t take a one-size-fits-all approach to state polling.”
Jensen says the only polls from the past year that he is “really unhappy with” were the ones for two special House elections — in New York’s 23rd Congressional District and Pennsyvania’s 12th. Jensen says the firm “seriously undermeasured Democratic interest and as a result overestimated Republican chances of winning. It’s fine with me if your readers want to not believe our polls, but we have no history of cooking the numbers for Democrats. In 2009, we were out of the gate, well before even Rasmussen, showing that folks like Blanche Lincoln, Michael Bennet, Jim Doyle before he retired, and Bill Ritter before he retired were in serious trouble.”
All fair points, but the experience of Research 2000’s impact in the Arkansas Democratic Senate primary indicates how much a poll or two can do to establish a powerful narrative. The polls, conducted for the Daily Kos, showed Bill Halter running ahead of incumbent Blanche Lincoln. This meant that he was treated as the favorite and she was covered as an embattled underdog, even though there is now no undisputed evidence that she ever trailed in the runoff.
Finally, a Republican consultant in North Carolina notes that PPP sometimes draws sweeping conclusions from tiny subsamples. For example, in August 2009, Jensen wrote about one of PPP’s North Carolina polls, “9 percent of voters in the state disapprove of both Kay Hagan and Richard Burr’s job performance. Those voters not surprisingly also give Barack Obama a 35 percent approval rating and only 12 percent like Bev Perdue. So these are the voters who basically don’t like anybody. They’re a Republican-leaning lot. 53 percent of them say they voted for John McCain last year to just 36 percent picking Barack Obama, and 45 percent are conservatives. . . . These folks are just mad in general and rather than taking it out on the party in control, they’re taking it out on the person they can control, which in this case just happens to be Burr.” Of course, this is all discussing 9 percent of a sample of 749, which amounts to roughly 67 voters. The margin for error on a sample that small is roughly 11 percent, raising the question of whether these numbers provide any actual analytical value.
Back in February, when Democrat Ken Lewis complained about PPP’s donations and past work for his rivals, Jensen responded that “every statewide campaign that has ever publicly attacked us — Richard Moore, Bob Orr, Elizabeth Dole — has ended up losing badly.”
Perhaps the latest candidate to scoff at PPP results, Richard Burr, will end up losing badly as well. Of course, that scenario is just what Public Policy Polling has been rooting for all along.
– Jim Geraghty writes the Campaign Spot on NRO.