The new Rasmussen poll and the Horner factor

The most recent poll in the Minnesota guv race hasn’t gotten much attention. It was mentioned in yesterday’s Daily Glean. In case you missed it, it was a Rasmussen poll and it came out:

  • Dayton: 40
  • Emmer: 38
  • Horner: 15.

Compared to the two most recent polls by the Strib (Dayton ahead by 9) and the Humphrey Institute (Dayton by 11), it shows Emmer moving closer. But we are often warned against comparing across polling organizations with different methodologies.

Compared to the previous Rasmussen poll in Minnesota (Emmer up by 2), the new one shows Dayton moving ahead, but just barely and well within the margin for error. Still, the previous Rasmussen poll was the only public poll in the race since June that showed Emmer with any lead at all over Dayton. So the four organizations that are polling the race are all now in agreement about who is ahead. Not for attribution, Republican sources confirm that their internal polling shows Emmer trailing, although not by as much as the Strib and the Humphrey found. But this result has been steady for a long time and, with three weeks to Election Day, Repubs are plenty nervous.

Rasmussen’s sample of the Minnesota likely electorate (38 percent Repub; 37 percent Dem; 26 percent unaffiliated) is significantly better for the Repubs than the Strib or Humphrey Institute found.

For Horner, the new Rasmussen poll, whether compared to the previous Rasmussen or to the other polls, shows movement in the wrong direction.  Because of the nature of third-party candidacy, Horner desperately needs to show upward movement to overcome the wasted vote syndrome.

With this poll, Rasmussen completes a controversy that began with his previous poll, which I wrote about at the time.  And it was resolved in a way that is a weird sort of backhanded compliment to Horner and the Minnesota Independence Party.

As I wrote in the previous Rasmussen post, the Rasmussen organization — which is controversial in many ways —  has an unusual polling technique in which voters who favor a candidate the first time they are asked are then pressed to say whether they are sure they will vote for that candidate. Those that say they aren’t sure are then given a chance to change their minds. (Bear in mind, this is all happening in a relative few minutes in an interaction between a potential voter and a robotic voice.) Over time, Rasmussen claims, this technique has been able to anticipate the likely late drop in support for third-party candidates as their more tepid supporters decide to vote for one of the top two candidates in the race.

In the previous poll, Rasmussen found that 18 percent of likely voters expressed a preference for Horner when first asked, but that his support dropped to 9 percent when voters were given a chance to change their minds. I’m skeptical of this unusual procedure, but Scott Rasmussen says it has worked in previous cycles and he has used in all polls after Labor Day.

With this poll, Rasmussen decided to make an exception for Minnesota. Horner has not been dropping like other third-party candidates and Rasmussen has decided to treat Minnesota differently.

As of this most recent poll, Rasmussen also found Horner drawing more support from self-described Dems than from Repubs, although most Horner voters come from the group that describes itself as affiliated with neither the Dems nor the Repubs.

You can also learn about all our free newsletter options.

Comments (16)

  1. Anonymous Submitted by Anonymous on 10/12/2010 - 10:29 am.

    Two words: Cell phones.

  2. Submitted by John E Iacono on 10/12/2010 - 11:10 am.

    One question, which is not addressed:

    Has Rasmussen turned out to be more accurate in the end in previous Minnesota elections?

  3. Submitted by Brian Simon on 10/12/2010 - 11:41 am.

    “Rasmussen’s sample of the Minnesota likely electorate (38 percent Repub; 37 percent Dem; 26 percent unaffiliated) is significantly better for the Repubs than the Strib or Humphrey Institute found.”

    Do they come up with that distribution by taking what they get from the calls, or are they weighting the sample to achieve those numbers?

    We’ve seen, since the 2006 election cycle, shrinkage in the number of self-identified Republicans nationwide. This year, while self-IDed Repuds are reputed to be more ‘enthused’, the party itself still has remarkably low favorability ratings – lower than the Dems who are projected to suffer from a wave of voter rage.

    Something isn’t adding up here.

  4. Submitted by Kathlyn Stone on 10/12/2010 - 12:09 pm.

    Are polling companies like Rasmussen and Gallup required to disclose who commissioned their polls? Both are businesses. They don’t conduct costly polls as a public service like the U or as news gatherers like the Star Tribune and MPR.

    Can a news organization request info on who paid for a poll using the FOIA? It has always seemed odd that Rasmussen, Gallup and the other big polling companies are allowed to release poll results without disclosing the sponsor. Thanks.

  5. Submitted by Peder DeFor on 10/12/2010 - 12:15 pm.

    According to Nate Silver at 538, Rasmussen has a small house effect that is pro-Rep. He also says that they have a better than average rating as a pollster, including a better rating than the StarTrib poll. This article here (which I highly recommend) talks quite a bit about cell phones and the problems they give pollsters. The take away is that it’s a known problem that they think they can address because it’s a stable demographic.
    Eric’s point about looking at the same polling agent for consistent views is a good one. It looks like Dayton has a small but very real lead right now.

  6. Submitted by Ray Schoch on 10/12/2010 - 01:02 pm.

    Mr. Iacono’s question is a good one.

  7. Submitted by Ed Stych on 10/12/2010 - 03:08 pm.

    Rasmussen tied with Pew for being the most accurate pollster in the 2008 national elections:

    I don’t know about Minnesota elections. But historically the Strib has been fairly inaccurate. Dems tend to poll stronger than actual votes. Even the best of the national polls overstated Dem strength in the 2008 elections.

    Not sure why Eric feels the need to slip in this phrase … Rasmussen “is controversial in many ways” … without explaining himself. Seems like an effort to discredit Rasmussen for some reason. Seems like accurate results are the only thing that matters.

  8. Submitted by Dan Hintz on 10/12/2010 - 03:33 pm.

    Ed, that link you cite to finding Rasumusen to be the most accurate is based on Obama winning by a 6.15 percent of the popular vote, which was his lead when that polling survey was taken right after the election. It says that the average poll predicted a 7.52 percent margin, overstating Obama’s support while Rasmussen got it right.

    The actual margin, after all the votes were counted, was 7.26 percent, which means that the polls did overstate Obama’s support by a quarter of a point, but that Rasmussen understated his support by a much greater margin. The linked site doesn’t have the numbers for everyone, but based on how close the numbers are, it is safe to say that Rasmussen was one of the worst pollsters. In any event, its an example of Rasmussen underestimating Democratic support which is consistent with the findings of poll watchers like 538, which base their findings on actual election results, and not bad information like your link did.

  9. Submitted by Thomas Swift on 10/12/2010 - 03:34 pm.

    “They don’t conduct costly polls as a public service like the U or as news gatherers like the Star Tribune [and MPR.]”

    I don’t care who you are, *that’s* some funny stuff, right there!

  10. Anonymous Submitted by Anonymous on 10/12/2010 - 04:21 pm.

    There is no “Minnesota Poll” anymore – it is contracted out, so it is absurd to look at its “record.” The real issue is, does this poll include cell-phone only citizens? They are close to 25 percent of the electorate now, and there is now way to “correct” for this – as a commenter suggested. In fact, those cell-phone only people in this election have tended to go for Dayton.

  11. Anonymous Submitted by Anonymous on 10/13/2010 - 09:21 am.

    There is NO formula for correcting for not polling 25 percent of the electorate. As David Brauer has pointed out on this very website, when cell phones have been included in polling the gubernatorial race Mark Dayton has done about five points better than non-cell phone polls. I personally could care less about who does better in which polls – I care about accuracy.

  12. Submitted by Peder DeFor on 10/13/2010 - 12:26 pm.

    Rob, add five points on then. I guess somehow that doesn’t count as correcting for a polling lack but it might make you happy.

  13. Anonymous Submitted by Anonymous on 10/13/2010 - 02:06 pm.

    It doesn’t “make me happy,” it makes me think the pollster is being more accurate by including cell phones. I’d say the same thing if the cell-phone polls showed a five point lead by Emmer.

  14. Submitted by Peder DeFor on 10/12/2010 - 09:24 pm.

    Rob, it sounds like Rasmussen actually did a pretty good job of correcting for lack of cell phone users in the past. Whether that continues to be the case is an open question of course. According to 538, they do an above average job at getting the polling correct.
    Look, it’s obvious that a lot of people want to throw out Rasmussen’s numbers because they don’t want to believe that this will be a bad year for Dems at the ballot box. That’s human nature. If you look back at conservative leaning sites in 2006 you’d find the exact same type of commentary, disregarding the bad polls and really pumping up anything that looked positive. If it makes you feel better to think that Dayton is ahead by five or six rather than two, then go ahead.
    The big news is that the difference between the two is much less than the stated support for Horner. If he collapses and his votes go one way or the other, that will be decisive.

  15. Anonymous Submitted by Anonymous on 10/14/2010 - 08:58 am.

    For anyone who cares about polling accuracy there is now a Pew Research Center report that says “Polls that don’t interview people on cellphones are producing potentially inaccurate results.”


  16. Submitted by Peder DeFor on 10/14/2010 - 04:48 pm.

    Rob, just for the sake of completeness, here is Nate Silver’s take on the Pew report:
    The short take is that Pew’s findings aren’t necessarily the same for other polling companies.
    We’ll find out how good the companies are on November 3rd.

Leave a Reply