John Behan points out that Rasmussen has the results in from their latest Kaine/Kilgore poll and it’s…erm…a mystery.
Astute followers will recall that Mason-Dixon released their poll July 24, which showed Kaine with 38%, Kilgore with 37%, and Potts with 9%. It represented a big gain for Kaine, but it was logical, since the poll was conducted immediately after the first debate, the first real event in the race.
The Rasmussen poll released today shows Kilgore with 45%, Kaine with 39%, and “other” with 5%. Absolutely nothing has happened since the debate.
So what gives?
There are two possibilities. The first is that opinions changed quite a bit in the past month — something caused Kaine to shoot up and then slide back down again. The second is that one or more of the polls are not externally valid — they don’t represent the actual opinions of the whole of voting Virginians.
The first scenario is plausible, but I don’t buy that Kilgore rose by 8% in the past couple of weeks, any more that I’d buy that Kaine rose by 8% in the same period. Nothing’s happened.
The second scenario strikes me as more likely. Given that the Mason-Dixon poll is the outlier, it’s tempting to dismiss that entirely. But Mason-Dixon is a reputable firm — to accuse them (or Rasmussen, for that matter) of bad polling wouldn’t make sense.
There is one important difference between the Rasmussen polls and the one Mason-Dixon poll: the Mason-Dixon is the only one to mention, by name, Republican Sen. Russ Potts. This is important. We’re still at a stage in the race when I doubt most people could name Kilgore and Kaine, to say nothing of Potts. People must be prompted with the names of the candidates, much as they will be when they walk into the voting booth and they’re confronted with the names Kaine, Kilgore, and Potts.
Today’s Rasmussen poll has “other” at 5%, which is a pretty big drop for Potts, given that Mason-Dixon showed him at 9%. Now, I don’t have the crosstabs for this Rasmussen poll, only for the Mason-Dixon poll, but I do know that Mason-Dixon found that Potts is really sapping support from Kilgore, far moreso than Kaine. Factoring in that 4% of people who know that they like Potts but need to be prompted by name went with Kilgore (presumably described as “the Republican” in the survey), that would be enough to bring Kaine and Kilgore within the margin of error.
Another noteworthy difference is that Mason-Dixon found that 16% of people are undecided. Rasmussen found that just 11% of people are undecided. It’s possible that some of that 5% has started making up their mind, and they like Kilgore. It’s also possible that the methodology used by Rasmussen and Mason-Dixon is different. Perhaps Rasmussen “pushes leaners,” meaning that if somebody says “undecided,” they respond: “But if you had to pick somebody today, who would it be?” Researchers differ on whether or not it’s a good idea to push leaners — some political scientists argue that the name provided by a pushed leaner is who they’ll probably vote for, while others say that they’ll just say any name to answer the question and move on.
So, is this Rasmussen poll right, or wrong? I have no idea. If I had some crosstabs, I’d feel better about it. The only polling that I’ve done is in volunteering for campaigns to work the phones on voter ID (such as for the Kaine campaign last week). I have a single semester of Political Science Research Methodology under my belt, which largest consisted of playing with NES data and SPSS. Here’s hoping that Mason-Dixon does another poll, or another firm entirely gets into the game, so we’ll have some more information to work with.
I agree, its hard to compare two polls when one poll completely disregards one of the candidates.
To be fair, perhaps ignoring one of the candidates is the appropriate thing to do given that he’s only at the 9% (or is it 5%?) mark. I don’t know — I’m not a statistician. But I have to say that it makes a whole lot more sense to me to poll people on the people whose names they will be seeing on the ballot.
I’d be curious to see how Potts would do in a poll that asked people if they intend to vote for “Democrat Tim Kaine” or “Republican Russ Potts,” leaving off Kilgore entirely. I suspect he’d shoot up to at least 30%. That doesn’t mean that Kilgore doesn’t have any support — it would just be what would come of failing to mention one of the candidates.
Of course, you know how I feel about this situation. Still, I expect a couple more polls will give a more convincing conclusion that will be hard to deny by either side.
C’mon guy’s, 9 out of 10 polls is pretty hard to dispute. 5 sounds about right for “other”…lol
I wish I’d paid more attention in statistics class … it might be noteworthy that the Kaine and Potts/”Other” numbers appear to be within the margin of error for both polls. It’s the Kilgore number that isn’t. Why would that be? Could it be the order of the questions? Maybe if you’ve been asked if you’re ok with the tax increase and then are asked who your candidate is, you go with Potts or Kaine to sound consistent to the pollster. If you’re just asked straight up, maybe you go with Kilgore. Rasmussen doesn’t ask issue questions, but he does ask about popularity of Warner.
Could it be the order of the questions?
Generally, they’re as random as is possible. Some don’t make sense out of order, of course. It’s something that a good polling company (like Rasmussen; unlike Survey USA) will control for.
Maybe if you’ve been asked if you’re ok with the tax increase and then are asked who your candidate is, you go with Potts or Kaine to sound consistent to the pollster.
I think we’d be giving voters too much credit with that scenario. :) I think it’s altogether possible, if we knew how the questions were phrased. NES data show that voters are often really quite inconsistent between their beliefs and the votes that they cast. I’ll be curious to find out — if we ever find out — to what extent that will be true this November, particularly with regard to beliefs on budgetary practices and the candidate for whom people vote.
As I noted on John Behan’s blog, Rasmussen uses different polling methods than M/D. Specifically, Rasmussen uses an automated system whereas M/D uses actual people conducting the poll. Rasmussen thusly produces a system whereby everyone hears the *exact* same questions, free of accents or other verbal tics, while M/D not only suffers that but basic human error as well.
The problem with recorded polls is that people hang up on them a lot more often or push nonsense answers just to get off the phone. That means less and less of a true random sample, and more and more after the fact re-balancing. There are pluses and minuses to any methodology and little agreement on which is really the most accurate. Getting a true random sample, let alone one that’s actually of a properly defined goal for what you really want to measure, is still a real pain in the tookus.