« BACK  |  PRINT

RS

FRONT PAGE CONTRIBUTOR

The Rasmussen Secret Sauce: Pennsylvania Governor Edition

By: Neil Stevens (Diary)  |  May 24th, 2010 at 01:13 PM  | 

RESIZE: AAA

Onorato Corbett

Are we tired of Pennsylvania yet? Of course not! Specifically we now check in on the Governor’s race. Rasmussen has released the first poll since the primary, but I will compare that with the last pre-primary Quinnipiac poll anyway. Tom Corbett and Dan Onorato were obvious likely nominees.

I believe we have as much to learn about Rasmussen’s distinctive modeling as we do about the race itself.

Both polls have Republican Corbett ahead, but the post-primary Rasmussen poll gives him a much larger lead over Democrat Onorato than the pre-primary Quinnipiac poll. Rasmussen shows Corbett 49, Onorato 36 (MoE 4.5) for a whopping 92% lead probability, while Quinnipiac has the race a bit closer at Corbett 43, Onorato 37 (MoE 2.9) and an 84% lead probability.

Both polls look very good for the Republican at this point, and show Onorato as being in need of a post-primary unity bounce as passions cool after the multi-way primary. But why does Rasmussen show so many more Republican supporters?

I think the key is that Quinnipiac has no secret sauce, and merely polls Registered Voters, while Rasmussen has his “House Effect,” as Nate Silver calls it, of likely voter filtering. He’s just assuming Republicans and right-leaning Tea Party voters will have more enthusiasm to vote this year, and that should be making his polls favor the Republican.

Quinnipiac gives Onorato a 34-13 favorability margin, and Corbett 49-13. Rasmussen shows 46-32 and 57-26 if we add strong and weak readings, or 15-14 and 27-11 if we only include the strong feelings. In other words, Onorato’s favorability ratings simply collapse in Rasmussen versus Quinnipiac.

Rasmussen’s final polls for the Democrats in their primaries for Senate and Governor were closer than both of Quinnipiac’s last readings, but I’m not sure how relevant that is to this issue. In theory Rasmussen could be the most accurate pollster on earth, but be entirely wrong in the profile of a likely voter in November 2010.

I have to think this election will bolster or destroy Rasmussen’s credibility as a public political pollster. The firm is going way, way out on a limb, and if it breaks, the wolves below are ready to kill. There’s no middle way here. You can’t substitute your own judgment for the 2006 and 2008 electoral history, and walk away unscathed if you’re wrong.

Crossposted from Unlikely Voter

Tags: , , , , , , , , , ,

COMMENTS

POSTING RULES  |  LOGIN TO COMMENT  |  REGISTER TODAY
  • victor_cocchia

    has proven over and over again to be the most accurate of pollsters. First of all his sample using Likely Voters definitely gives a more accurate picture of the race than a Registered voter, or the Dem face-saving All Adults poll (which are usually conducted to bail out the One when his approval ratings go lower than the MSM wants).

    As well, when comparing the polls, Quin. has a definite D slant and will overestimate the D vote totals in a majority of their polls.

    The biggest knock on Rass really has been that his one day 500 LV state polls tend to have some big swings because they are conducted on, just one day. When he does he three day polls you can usually take those to the bank.

    just my two cents.

  • http://www.hakubi.us/ Neil Stevens

    If the polls are right in November, Rasmussen will be fine.

    But people are watching and there will be a coordinated attack on Rasmussen in November if those polls are wrong.

  • Scope

    Rassmussen will continue to be seen by the left as being “biased.” That really makes me laugh, as I do believe there are alot more left leaning polls out there than there are right leaning ones. In the current political atmosphere, the lefties are trying to do their darndest to make things look better for them than they apparently are. If there are some elections that are truly stolen, which I don’t only think may happen, I believe it will happen in some areas. Hard to accurately poll stolen elections.

    The RCP PA Governor’s poll has shown Corbett ahead by double digits since March. He is currently averaged at +13.5. That is great for the RCP averages.

    RCP also has Obama’s approval rating average at +1.5. Rassmussen has him at -11. The current polls, making up the current average go from 5/6 through 5/23. Out of the 7 polls making up the average, 5 of them polled only Adults. Ipsos/McClatchey has him at +9 on their 5/6-5/9 poll. Using 5 polls of only adults, which in most every case has him ahead consistently by at least 3 points, to me skews the results in Obama’s favor. RCP only ever had Obama’s number in the negative, and by less than a full percent only once, for only 1 day. Last week they had him ahead by more than 4 points.

    http://www.realclearpolitics.com/epolls/other/president_obama_job_approval_1044.html#polls

    What my question is, with respect to the RCP averages, is how can the numbers be anywhere near accurate when they do not always include the same polling outfits in every average, using the same methods (Adult, Likely Voter etc) and for the same number of days in the time period. Just as the media is in the bag for Obama, it seems so are many of the polling outfits. It’s easy to stay ahead when the averages include mostly left leaning polsters, and there are definately more of them out there than right leaning ones. It’s not enough to serve constant adulation in the news for the One, they are backing it up with crooked polling tactics. As I had asked before, has anyone ever called back all those polled to see if they really did get called, or if they did give those answers? At least Rassmussen has logs of phone numbers called, do the others?

  • gemimail

    Even Gallup admits that only likely voter polls of congressional generic ballots have any predictive ability for the final results in midterm elections. See: http://www.gallup.com/poll/128075/Vote-Congress-Remains-Tied-Among-Registered-Voters.aspx. It is just that Rasmussen uses likely voter polls all the time and Gallup does not switch to them until later in the year.

    Where likely voter polls do not do well is in special elections such as NY-23 and PA-12 because the voters turning out do not fit the norm. Still with a good likely voter model, such polls are the most accurate. A good rule is to compare Rasmussan to PPP and average them.

  • Adjoran

    You get the average of the latest polls – by various pollsters, using various methodologies, taken at different times.

    The theory is that any given poll, or even two, can be very flawed, but averaging a bunch together is likely to cover the flaws.

    No, it isn’t “scientific” polling and yes, an outlier can affect the average. If you don’t like their poll averages, RCP offers your money back, no questions asked.

  • tngal

    Given that everyone had the PA 12 SE tied, tossup, neck and neck right down to the day before? And I know SE’s are suppose to be difficult to poll, but man the numbers coming out were no better than guesses.

    Maybe people were lying to the machine or the person on the other end. Maybe it was too much or not enough secret sauce. Maybe they should have asked the question, how stupid are you on a scale of one to ten? Then anybody that admitted being over a three, you’d know they were voting dem.

  • http://www.hakubi.us/ Neil Stevens
  • http://www.hakubi.us/ Neil Stevens

    I’m not a fan of the RCP averages. The site overall is very nice. I visit it often and highly recommend it to people. I just see no basis for their averages, which are one tiny part of the site.

    I try to apply the actual statstical value of individual polls in my simulations. I’ll really be gearing up for that later, as the summer winds down and the general elections get into gear properly.

  • Scope

    the most of the polling outfits were Democrat leaning, and there was only a small representation of right leaning pollsters? Couldn’t the polls be skewed to show the results you want? Please look at who all is adding their polling data to the results, and who they are polling.

    Again, can YOU please tell me if anyone has ever checked on any polling results? It would be close to impossible. That’s my problem with “polling.” And, the leftists will stop at nothing, and, that has been proven.

  • tngal

    did skew waaay toward Critz, but it was difficult finding someone who actually believed it because it was so far off from everybody else. Also it was a one day poll with 400 likely voters, and turns out even S’hana was off by one and a half points.

    Its like we can’t believe anything. Look at Florida. Crist is up in the poll out by St. Petersburg Times done the 14-18. But in a poll done the 16th by Ras, its Rubio up by 8. That’s a pretty big diff.

  • http://www.hakubi.us/ Neil Stevens

    If races never changed shape then McCain would have won in 08 and Crist would be ahead of Rubio by 50. :-)

  • Mary Beth

    Obviously not related to PA since you didn’t grade them…but there’s a new poll out from SurveyUSA regarding the CA races.

    Are they good? I ask because their poll shows a pretty sharp change in favor of Fiorina and what I’m trying to figure out is if this is legitimate or if it’s an outlier or what.

    Any thoughts?

    Here’s a link btw: http://www.surveyusa.com/client/PollReport.aspx?g=203d9267-8869-45c0-b3de-95ed057a54f9

  • http://www.hakubi.us/ Neil Stevens

    Spoiler: SUSA has been too busy doing approval rating polls for us to get a good read on their performance this cycle.