The Big Joke

How do you put lipstick on a pig?

Eric Black tries.

Eric Black's "The Big Question" blog is one of the Strib's better efforts. In the latest post, Black interviews Rob Daves, the Strib staffer who runs the Strib's always-comical "Minnesota Poll".

You know - the poll that predicted Skip Humphrey and Roger Moe would be our governors, and that Ann Wynia and Walter Mondale would be our senators.

Black asks about those pesky allegations of gross incompetence and rampant pro-DFL bias (Black's questions in italics):

3. Republicans have mocked your Senate poll result. Kennedy aide Heidi Frederickson said it belongs on the comics page. Both the Minnesota GOP and the Kennedy campaign put out press releases alleging not only that the poll was ridiculous, but that it was a deliberate attempt to affect the election. How do you respond to that?

It’s been a long-held custom - since the 1700s, actually - for partisans to criticize polls that aren’t favorable to their candidate. [Especially when that poll is not only consistently slanted toward one party, even when the other party repeatedly wins. Or so one might think - because I'm not aware of another poll that's inaccuracies are so, oddly, predictable.] But the accusation of some sort of conspiracy to affect the election is ridiculous. We poll using standard, proven methods of sampling, questionnaire construction and analysis. Then we print the results. That’s it, pure and simple.

Well, that settles it, doesn't it!

So why does the "proven methodology" that, in the hands of other pollsters, at least delivers inconsistent, varied mistakes, yield such consisten results?

I work for the folks who print the news, not the editorial page editor who’s in the business of trying to change opinions.

In terms of lines on an organizational chart, that might be true.

In terms of how news is actually produced, that is transparent rubbish. The bias in the Strib's news coverage - or, rather, the omissions from the Strib's coverage - perfectly reflects the "Editorial page's" direction, and always have; the Strib's news pages run the most absurd stories about Republicans to death (Morgan Grams, anyone?) while Twin Cities bloggers have shown how stories that might reflect badly on Amy Klobuchar, Keith Ellison, Mark Dayton, Paul Wellstone and so on never see the light of day, Strib-wise. I showed myself how, once a pro-Mike-Hatch story turned ugly for the Attorney-General, the Strib and PiPress spiked it for good.

So the divide between editorial and news at the Strib doesn't seem all that divisive, to the mere consumer of news.

Back to Black:

4. As you know, the Republicans have been complaining about your polls for years. In 2004, they even called for you to be fired. Obviously, it isn’t just based on this latest poll but on a long-standing Republican complaint that the Minnesota Poll routinely understates Republican support. And they have a table of numbers that they say proves the pattern. Do you dispute their numbers? Do you dispute the pattern? How do you respond?

I saw that table a couple of years ago, and thought that someone must have a lot of time on their hands, Eric. [Ah. And they sit in their basements in their pajamas, too, right? Do these news people ever figure out that some of us just have a lot of passion and indignation on our hands?] But after I looked at it, it was clear that there were a couple of things going on. One is that they use faulty information. [The jokes here are too obvious. Fill them in yourself]

Our past practice has been to poll the week before the election and publish those results on the Sunday before the election. In the early- to mid-1990s, it became clear that there was a shift toward the Republican candidate in the waning hours before the election. So we started doing tracking polls the two or three days just before the election, then publishing a chart on Wednesday after the election showing how things changed. Why after the election? Editors didn’t want to appear as if they were influencing the election by publishing close to Election Day, as the national polling organizations do.

In 2002 - which is one of the elections the Minnesota Poll’s critics point to - our poll taken the week before the election and published on Sunday did, in fact, show a Walter Mondale lead. But the tracking poll, done on Sunday and Monday and published the day after the election, showed a dead heat at 45 percent for Mondale and 45 percent for Norm Coleman.

In 2004 I convinced the editors to publish right up to Election Day. And yep, the critics are right: Our poll published Sunday before Election Day showed a big Kerry lead - 8 points. But our tracking poll published online Monday night and in the paper on Election Day showed a shift to a 4-point Kerry lead - the conservative shift I mentioned; Kerry won by 3.5 points. (And FYI, those who keep up with such things found that the Minnesota Poll was one of the most accurate in the nation; check out www.surveyusa.com/Scorecards/2004PresGovSenOnly.xls)

So if the Minnesota Poll consistently shows a "last-minute surge" of conservative voters - a surge that no other polling organization consistently or widely shows, what does it mean?

Do us conservatives gather every two years and hide in the woods and snicker and giggle and underreport ourselves to the Strib's pollsters (but not Harris or Gallup or Quinnipiac or Ipsos or Rasmussen), and then go racing to the polls and vote for winning GOP candidates and then gather at Keegans for yet another big laugh at the "Minnesota Poll"?

Or is the "Minnesota Poll's" methodology fatally flawed toward underreporting conservatives?

So which poll results should that chart cite? I think it should be the one done closest to Election Day that shows how voters shift at the last minute toward the more conservative candidates. My guess is that the chart doesn’t reflect those numbers, just the ones that make the Republicans’ point about bias.

My "guess" is a little more pointed, inasmuch as Daves' explanation leans toward proving the critics' point.

The DFL-slanted polls in the months before the election serve the DFL - which reflects the Strib's editorial bias.

The magically-"accurate" last-minute polls are cover to provide the Strib some veneer of "accuracy".

Seems like a conspiracy theory, doesn't it?

Then why, in election after election, has the poll done nothing to correct this "error" in their polling? I don't know of any pollsters (besides Zogby) who sit by complacently, year after year, allowing their polls to continue displaying patently, ridiculously inaccurate information without raising serious questions about their own methodology.

Those are the two things: We’ve seen solid, empirical evidence that there’s a conservative shift in the electorate in the days immediately before the election, and we’ve seen that the poll’s final, final numbers often are ignored.

Is the "evidence" "empirical?" There's been an experiment with a fixed control sample?

That's not "empirical". It's "wishful".

Let's see how big the magical conservative surge is this time.

Posted by Mitch at September 19, 2006 06:51 AM | TrackBack