Skip to Content

Support MinnPost

MinnPost logo 7th Anniversary

MinnPost’s online auction is now live!
Register and start bidding today

Is the McCain-Obama race really close in Minnesota?

Whoa, Nellie.

Has McCain really pulled even with Obama in Minnesota?

The Star Tribune stated it, this morning, as if it was an established fact — and in headline type, on Page 1A, above the fold.

The story was a lot more cautious. The Strib's page one editor acknowledged this morning that the headline lacked nuance. But the decision to put the poll story on page 1, with that headline, was a shock because — and I speak as a 30-year Strib veteran — the paper historically has had a huge bias against putting a poll about Minnesota, other than its own Minnesota Poll, on Page 1.

The headline is based on a Quinnipiac University poll taken in four "swing" states for the Washington Post and Wall Street Journal. Esteemed colleague David Brauer put up a Political Agenda item when the poll came out yesterday. And Brauer noted in today's Daily Glean that while the Strib found the poll to be a Page 1 story, the Pi-Press ran a wire story on it on page B5. Have I mentioned lately that journalism is not science?

The poll actually found Obama leading by 46-44 percent, but that's well within the margin of error. So if you were to base your entire understanding of the race for Minnesota's electoral votes on the Quinnipiac poll, you could call it even.

If it turns out that other pollsters, including the most reputable ones using the most trusted methods, confirm that the Minn prez race is now a toss-up, that will be a very big deal for politics-obsessed horserace watchers like me (and, admit it, you, too). It would even change our understanding of where the whole race stands nationally.

But, of course, that's one reason to take the Quinnipiac result with plenty of salt. There's nothing going on in the national polls or in Minnesota's neighboring states (including in this same Quinnipac poll, which has Obama leading in Wisconsin by 50-39) that would make a big Obama collapse in Minnesota seem likely.

When a single poll deviates from the consensus of many previous polls, I suggest that our attitude should be: Whoa, Nellie.

At the moment, the Quinnipiac poll is an outlier. Big time. Using the compilation of Minn prez polls by Pollster.com, you'll find that of the last 10 Minn prez polls taken, other than this Quinnipiac poll, Obama led by double-digits in eight of them. The biggest lead shown — 54-37 for 17 points — was in the previous poll by Quinnipiac itself one month earlier. That makes the whole thing mysteriouser, since the big swing toward McCain can't be explained by a difference between polling operations or methodologies.

When an outlier like that comes along, you need to stay calm. It may be the first one to pick up a trend. But there's an excellent chance it will not turn out that way and you should await developments before you treat it as the new reality.

The estimable Pollster.com, by the way, which is run by a group of poll-obsessed political scientists, uses a complicated formula to derive an average result when looking at many polls of the same race — except it isn't an average, it's a regression based on trend lines, whatever that means. Taking all recent polls, including the recent Quinnipiac, Pollster finds that the trend in Minnesota favors Obama by 50.2-37.7,

I wouldn't state that as a fact either, nor do the guys at Pollster make any such claim, but it has more going for it than one recent outlier.

Let's violate the code a little more and talk about the whole list of Minn prez polls. Most of them come from the same two pollsters, Survey USA (which polls locally for KSTP-TV) and Rasmussen Reports (which polls for Fox TV stations, which locally would be KMSP).

They poll very frequently, their polls are less expensive because SUSA and Rasmussen use the robotic interview method where respondents can punch their answers into their phone in response to questions from a recorded voice.

The polling establishment hasn't made up its mind about this methodology, and big news organizations tend to feel more comfortable relying on the older (but more expensive) methodology requiring actual human-to-human interviews.

My most poll-savvy advisers are undecided but a bit skeptical about the new methodology. They also tell me that SUSA has the better track record in Minnesota and does a better job disclosing and defending its methodology than does Rasmussen. Personally, I take that into account when the two firms continue pumping out frequent polls on both our Senate and prez races.

So, I assume that one thing going on in the background of the Strib's decision to go big with the Quinnipiac result is that it used the more trusted methodology. It also had a larger sample and a smaller margin for sampling error.

To his credit, my esteemed former colleague Bob von Sternberg, who wrote the Strib story (but not the headline), talks about a lot of these issues. He mentions in the second paragraph that Quinnipiac is "at odds" with other recent polls, including Quinnipiac's own June poll. He talks about different methodologies and sample sizes. He was not able - within the voice boundaries of mainstream newspaper journalism - to indicate which methodologies and organizations are more trusted, as I just did, but I would say he was trying to at least hint at it.

So, to summarize, reasons to make a Page 1 story out of the Quinnipiac poll:

• It's surprising.

• Large sample.

• Established methodology.

• Comes with the imprimatur Washington Post and Wall Street Journal.

Reasons to be leery of it:

• It's an outlier (which is why it's surprising).

• Same poll showed Obama still leading by 50-39 in Wisconsin. No obvious explanation for the disparity.

• Changes in national popular vote polls show some decline in Obama's lead but not consistent with a huge one-month swing in Minnesota.

Then there's this rather strange acknowledgment from Quinnipiac University's Clay Richards, as paraphrased by my esteemed Associated Press colleague Brian Bakst:

"Richards, the assistant director of [Quinnipiac's] polling institute, said the Obama slide probably isn¹t as dramatic as the raw numbers reflect."

Oh really?

So, returning briefly to the Strib's decision to front page the poll story, which, as I said above is at least historically strange, esteemed colleague Brauer spoke to esteemed former colleague and current Strib Page 1 editor Colleen Stoxen (hi Colleen, hope you're doing well).

She said that yes, the headline was too strong ("could have been more nuanced"), that no, the fact that the poll was good news for McCain had nothing to do with it ("absolutely not"), that she has not been front-page editor long enough to know how big a break with tradition it was (it's me saying that it was, but it was), and that the poll story got onto the front page late in the day, because of considerations of what Page 1 editors call "mix." ("It was just a day where that type of story fit in topically with everything else.")

Have I mentioned lately that journalism isn't science?

P.S. I haven't mentioned that the same Quinnipiac poll showed Norm Coleman leading Al Franken in the Senate race by 53-38 percent of likely voters. The Rasmussen poll, taken at about the same time, had it Coleman 44, Franken 43.

What think?

Get MinnPost's top stories in your inbox

Related Tags:

Comments (9)

Is anyone watching the odds in Vegas? I would trust this more.

I, too, appreciate your analysis. The multiple poll business is really adding to the chaos of this election. It's already nuts, making the 2004 election look as ancient as Truman v. Dewey.

I think Patty is dead right about the significance of cell-only households. Whether those voters are primarily younger may be less certain (I think so, but know many 60s+ folks who've opted for the savings), but I think it's safe to say they're much more comfortable with changing technology, more likely to frequent the blogosphere, and possibly more likely to get fed up with the whole mess. In 2004 I think the young let us down. In 2008 they may decide we've let them down.

Another aspect of this type of polling, that I've been reading on various blogs, is the discounting of the growing number of Americans who rely primarily on their cell phones, versus a home phone. Do these Americans have a particular political lean that isn't being factored in? Whether they do or do not, there's a growing concern that these voices aren't being heard when polls such as Quinnipiac's are conducted.

Polls are polls, after all, and even the best are subject to so many uncontrolled variables that reliability is minimal.

Perhaps what esteemed former colleage Colleen meant was "It was a pretty much boring day."

Meanwhile, the sun is shining, and I think I'll go get a bit of it.

Why are news organizations (the Strib is NOT unique in this) reporting single pollsters?

Have anyone thought of averaging out these polls?

Useful analysis, Eric.

What will also be interesting is to see whether the next Quinnipiac poll on the same subject is anything like the 17-point Obama lead seen in the poll prior to this statistical tie.

If the same poll and same methodology go from Obama +17, to Obama +2, back to Obama +high teens over the course of consecutive polls, what are we to make of the nature of the outlier? Super schizophrenic polity? Hanging chads on the surveys? Can't trust anyone anymore?

Also, would Obama's return to +17 then have to make the front page of the Strib above the fold? I can see the headline now, "For first time (since two polls ago), Obama takes commanding lead over McCain in MN"

From my own personal observations (make of them what you will), Survey USA conducts garbage polls. First of all, I've been conducted for both of the last two Senate/MN political issues Survey USA polls for KSTP. That makes me wonder if they're just using the same database of responders for every poll they conduct. The odds of being contacted twice in a row for a tiny sample size are pretty low.

Second, Survey USA phrases the questions so that every single question leads with the pro-Republican response as the first option. Survey writing 101 (which I actually took in grad school) tells you that you should never set up polls this way. People who can't be bothered to pay attention to the survey will just hit 1 on their phone for every single question. Survey USA sets up their phone polls so that you don't even have to listen to a word of the question before giving your response.

Third, there's no way to change a reponse once you've entered it. Again, this is likely to favor the first (or "Republican") option. Responders (like me) will hit one before hearing the whole choice, and then be unable to fix their mistake.

Finally, a note about underrepresentation of the cell phone-only crowd: I don't think it exists. They're a blip in the overall population, and aren't likely to be statistically different from the population as a whole. They didn't show up as predicted (and often blogged about) for Kerry, and they won't show up for Obama either.

I had been wondering when someone would address this disparity from the other polls; especially given that there have been no major developments in the Minnesota that would affect the numbers this way. Kudos for giving us some perspective.

And as someone who has abandoned a home phone line in favor of only a cell (along with almost all of my 20-something and 30-something friends), I do feel that this has some impact on the existing poll numbers. A whole segment of the population is being under-represented and in a campaign year where the youth vote is more energized than ever, I suggest the pollsters find some better way to gauge the public sentiment.

Keep hoping, Eric, but Franken and Obama are both done.