In July, I started to notice something funky about the media polls. Almost all of them were using polling samples that assumed a dramatic increase in Democrat turnout over 2008. Even though the liberal group Third Way had found that Democrats lost around 1 million voters in 8 battleground states, media polls were implying an even bigger Democrat wave this year. It doesn't make sense.
Since I started this lonely crusade, I've been attacked by the left for daring to question the validity of these polls. Now that many other voices have chimed in, the debate has reached fever pitch. Even individual pollsters have chimed in to defend their polls, arguing, basically, "we're just reporting what we're hearing in interviews. If it's D+3,000,000 that's just what it is." In other words, they don't really poll in the statistical sense; they just do lots of interviews and the chips fall wherever they may.
The whiff of defensiveness is palpable. What I've yet to hear is any cogent argument for why Democrat turnout would be so much higher this year over 2008. Generally, it defies common sense. Specifically, it defies the terrible economic news and the growing crisis in foreign policy.
I read recently that something like 70% of first-time voters from 2008 have subsequently moved--probably to their parents' basement. And over 60% of these don't know where they are registered to vote. The youth wave of 2008 isn't going to happen again.
The most consistent critique to my analysis is the notion that pollsters seek accuracy over everything. "Why would they juice the polls?" is what I consistently hear. Apparently, pollsters are some deified profession that always seek total accuracy in everything. Dispassionate observers of the American political scene.
So, I decided to look back at the accuracy of polls in 2008.
There is such an assessment and it can be found here. The grading looks at 2 things; the accuracy of the final poll before the election and the consistency of its results to that outcome throughout October. Only one pollster earns an A grade for 2008: Rasmussen. It was, by far, the most accurate and consistent poll of the '08 cycle.
A look at other media polls' performance in 08 is especially illuminating, given the current questions surrounding their polls.
- NBC/WSJ C
- Marist D
- ABC/WaP0 D+
- Gallup D
- CBS/NYT D-
- Reuters F
So, if these polls were so off-the-mark in 2008, why should I pay attention to them now? It isn't surprising that the worst-performing polls from '08 all show Obama with a comfortable lead. The best-performing poll from 08 shows the race tied.
If you want to continue to ignore the problems in these polls' samples, feel free. Just understand that you've hitched your wagon to falling polling stars.
Follow me on Twitter.