Skip to content

Don’t trust Rasmussen polls!

Political scientist Alan Abramowitz brings us some news about the notorious pollster:

In the past 12 months, according to Real Clear Politics, there have been 72 national polls matching Clinton with Trump—16 polls conducted by Fox News or Rasmussen and 56 polls conducted by other polling organizations. Here are the results:

Trump has led or been tied with Clinton in 44 percent (7 of 16) of Fox and Rasmussen Polls: 3 of 5 Rasmussen Polls and 4 of 11 Fox News Polls.

Trump has led or been tied with Clinton in 7 percent (4 of 56) polls conducted by other polling organizations.

To put it another way, Fox and Rasmussen together have accounted for 22 percent of all national polls in the past year but they have accounted for 64 percent of the polls in which Trump has been leading or tied with Clinton.

Using Pollster’s tool that allows you to calculate polling averages with different types of polls and polling organizations excluded:

Current Pollster average: Clinton +2.7
Removing Rasmussen and Fox News: Clinton +7.7
Live Interview polls only: Clinton +8.8
Live interview polls without Fox News: Clinton +9.2

I find it remarkable that simply removing Rasmussen and Fox changes the average by 5 points.

Hey—I remember Rasmussen! They’re a bunch of clowns.

Here are a couple of old posts about Rasmussen.

From 2010:

Rasmussen polls are consistently to the right of other polls, and this is often explained in terms of legitimate differences in methodological minutiae. But there seems to be evidence that Rasmussen’s house effect is much larger when Republicans are behind, and that it appears and disappears quickly at different points in the election cycle.

From 2008:

I was looking up the governors’ popularity numbers on the web, and came across this page from Rasmussen Reports which shows Sarah Palin as the 3rd-most-popular governor. But then I looked more carefully. Janet Napolitano of Arizona is viewed as Excellent by 28% of respondents, Good by 27%, Fair by 26%, and Poor by 27%. That adds up to 108%! What’s going on? I’d think they would have a computer program to pipe the survey results directly into the spreadsheet. But I guess not, someone must be entering these numbers by hand. Weird.

I just checked that page again and it’s still wrong:

Screen Shot 2016-05-20 at 8.10.03 PM

What ever happened to good old American quality control?

But, hey, it’s a living. Produce crap numbers that disagree with everyone else and you’re gonna get headlines.

You’d think news organizations would eventually twig to this particular scam and stop reporting Rasmussen numbers as if they’re actually data, but I guess polls are the journalistic equivalent of crack cocaine.

Given that major news organizations are reporting whatever joke study gets released in PPNAS, I guess we shouldn’t be surprised they’ll fall for Rasmussen, time and time again. It’s inducing stat rage in me nonetheless.

If only science reporters and political reporters had the standards of sports reporters. We can only dream.


  1. Brad Wyble says:

    But if they’re reliably skewed by a roughly similar amount, isn’t it still useful as a predictor? You just need to adapt to its bias (which of course the papers will never do)

  2. Llewelyn Richards-Ward says:

    I don’t understand how a pollster who gets things wrong or who is biased is able to continue publishing opinions. Surely the only point of publishing this information is the appearance of accuracy in predicting the fickleness of public voting. Lose your accuracy and lose your newsworthiness?!

  3. Mark Hartgerink says:

    Rasmussen is the only poll I trust. Look at their record. They’re dead on every time.

  4. Natalie says:

    Those numbers are incorrect. Taking Rasmussen and Fox out of our trend results in a 44.9/41.4 race, Clinton +3.5.!minpct=25&maxpct=60&mindate=2014-01-01&maxdate=2016-10-02&smoothing=moderate&showpoints=yes&showsplines=yes&hiddenpollsters=rasmussen,fox&hiddensubpops=A – D,A – R,A – i,RV – D,RV – R,RV – i,LV – D,LV – R,LV – i&partisanship=S,P,N&parties=D,R,I,N&selected=clinton,trump&fudge=1&except_questions_with=johnson

    Adding Rasmussen back in makes it 44.5/41.4, Clinton +3.1. Adding Fox back (all the data) makes it 47.6/42.8, Clinton +4.8. So Fox is actually pro-Clinton (and their polling is editorially independent of the news network, so really they shouldn’t be in this category).

    Our model is a Bayesian Kalman filter that does not allow any single poll (or pollster) to pull on the model.

  5. Timothy says:

    Actually don’t trust any pollsters or polls. The polls can be skewed to get the answer they want by how and when questions are asked. Penn and Teller did a great segment in Bullshit. Just go to you tube and search bullshit polls. Eye opening

  6. Billy Armstrong says:

    I know you leftists don’t want to believe Rasmussen. The best thing to do is to wait and see which poll was the most accurate. Let the facts be the judge. Then make a mental note of it so you don’t write this same article again in 2020.

  7. urbanix says:

    Well, most of the time. You could get percentages to sum to 99 or 101 if you’re looking at rounded numbers, and there are three or more categories. Getting to 108 is a whole other deal, though.

  8. David J. Harris says:

    For reference, 538 gives Rasmussen a C+ and Fox an A. If I’m reading things correctly, they nudge Rasmussen polls by about 2 points and Fox polls by about 1 point toward Clinton to control for house effects

  9. Gaurav says:

    Dug up some old data/graphs for entertainment:

    The data etc. underlying this posted here:

  10. Helen says:

    Whew, thanks for posting this story, I heard Breitbart radio boasting this morning that Rasmussen has Trump swinging up 9 points since his “strong” performance at the last debate. I needed to talk myself off the ledge

  11. Swart says:

    I would like to hear your follow up this this post. You were wrong, Rasmussen was right.

  12. Cynthia says:

    President Donald Trump is doing a Great job to make America America again.
    I’m always curious how in 60 years of life I have never been questioned by any poll or do I personally know anyone who has been either. So… I regard your findings as Useless Non Important BS.
    You belong to Fake News. Your math is plain Stupid.

    • Andrew says:


      If you have a land line, pollsters should be calling you from time to time, just from random digit dialing. So if you’ve never been questioned by a poll, I’m guessing that’s because you’ve declined to participate. Lots of people are too busy to respond to polls; that’s why we adjust the data. Unadjusted poll data would indeed be close to useless in many cases.

  13. Throughout this year 2017 Rasmussen’s polls have showed Trump’s approval over 44% Wich it’s not credible, because it differs from the other polls around 7% The other polls have had Trump all year on or below 40%. 15 different polls can’t possibly be wrong or be altering the facts. Right now Rasmussen has Trump on 44% Wich is the highest on all polls, Reuters poll show Trump at a low 35%. If we calculate all polls we have 39.5 % approval for Trump.

    RCP Average 10/26 – 11/9 38.3 56.9 -18.6
    Gallup 11/7 – 11/9 37 58 -21
    Rasmussen 11/7 – 11/9 44 55 -11
    Economist 11/5 – 11/7 40 55 -15
    Reuters 11/3 – 11/7 35 60 -25
    CNN 11/2 – 11/5 38 57 -19
    IBD/TIPP 10/26 – 11/3 36 58 -22
    ABC/WP 10/29 – 11/1 38 58 -20
    CBS News 10/27 – 10/30 39 55 -16
    PPP (D) 10/27 – 10/29 38 56 -18
    All President Trump Job Approval Polling Data

  14. Cecil says:

    …which just goes to show that the polls are $#17!!!

    This post did NOT age well:

    Maybe you should quit your day job, Andrew Gelman. Just sayin’…

    • Andrew says:

      Hey, it’s great to see that people are still reading some of these old posts! I hope someone at Rasmussen is also reading our old posts and will fix whatever it is in their process that allowed them to post numbers that added up to 108%. It’s the least they can do if they want to call themselves the most accurate pollster.

Leave a Reply