Given that we’re about neck-deep in the campaign for both parties’ Presidential nominations, I think a few words are owe towards how I’m going to handle polls on this (obviously rather new) blog.
First Principle: It’s just one poll. Basically, no poll should be considered to be gospel. This is what every single campaign I have ever seen does unless they -are- doing well in the polls and moving up in a bunch of them. Then they brag about all of them. Well, there’s a problem with isolated polls, especially those that tend to generate the bragging rights: Either those polls are way off or everybody else’s polls are way off.
Second Principle: The margin of error. So an interesting poll has come out…and it shows a different leader in a hotly contested race. Big story, right? Maybe. Depends on how much different the poll is from the others; if the difference is one or two points…congrats, you’ve found a poll that says just about the same thing as the other guy. If it’s off by 5-6 points, you might have a trend. And if it’s off by 20, refer to principle three.
Third Principle: This poll makes no sense. Alright, so an odd poll has come out. It shows something nobody else is showing…a landslide win for somebody who’s losing in all the other polls, shows something completely counter to a dominant trend at a given moment, or something else odd. So, this is probably news…if it makes sense. But what’s causing the difference? Sometimes, it’s a poorly-balanced sample. Sometimes, it’s odd methodology. Sometimes, it’s odd timing (there was a particularly interesting incident back when Ford pardoned Nixon and his numbers collapsed in the middle of a multiple day sample…it made for some interesting analysis since there was a change in the tide in the middle of a poll -and- it made sense). And sometimes it’s just dumb luck and the pollster should have known to throw this one out of the last twenty out.
Now that I’ve discussed a bit of how I’m going to look at poll results, let’s talk pollsters. I’ll admit that I’ve got mixed feelings on a lot of them…Zogby came up with that boner back in 2004 on the Presidential results (300-odd EVs for Kerry? Now, really…), but everyone’s entitled to screw up once in a while for whatever reason. We’ve all had our own moments of brilliance, and I suspect I’ll have my own in the future as well.
Rasmussen is one of the better pollsters for general elections, I think, though he does have a slight Republican tilt. In the primaries, he filters harder against independents. This makes sense in some states (Iowa comes to mind, as does New York, with its extremely strict party registration) but not in others (New Hampshire, for example, as does Virginia…and any other state that lacks registration). Nationally, this has given him the distinction of being the one and only pollster to put Fred Thompson in the lead…as well as giving Mike Huckabee his only lead nationally to date.
American Research Group (ARG) is something of the antithesis of Rasmussen, in that ARG probably doesn’t filter enough…or at least didn’t filter very well up until recently. This can be shown in a slew of nutty Iowa results that dogged ARG throughout the early part of the campaign, giving McCain far better showings there than he got in most other polls (at least according Real Clear Politics). ARG seems to be a bit more in line with everyone else now, so presumably either they got a long string of bad samples or changed their methodology.
The Partisan Pollsters. This is the term I’ll use for the groups that get the (R) or (D) next to their name over at Real Clear Politics. Strategic Vision is the big one on the Republican side, while the Dems have a handful of them; Democracy Corps is their big one, however. I question their accuracy, as they have an obvious advantage in producing results that give their people a lead. With that said, this probably isn’t as big in a primary (unless the party leadership is backing someone in said primary), but it still gives good cause for suspicion.
