Is Rasmussen the most inaccurate pollster out there?
Dsw4 | Wikimedia Commons
One of the most frequently cited pollsters in comments on this blog is Rasmussen. Not only is that company the most prolific pollster in the United States, it’s also one of the most controversial.
Nate Silver at Five Thirty Eight Blog recksons Rasmussen leans about +2 to Republicans candidates. Silver must be talking about in national polls. Because in the swing states, in the 2008 presidential election, Rasmussen leaned toward John McCain in 11 so-called “battleground states” by a wider margin.
The average Rasmussen lean for McCain was +4.08. Of the 11 contests, Rasmussen actually called the election the wrong way in 4. He had McCain ahead in Indiana, Florida and North Carolina, but Obama won all of those states by slender margins. Ramussen’s last Ohio poll called the 2008 race a tie. Obama won by 4.6
But don’t take my word for this. Look it up yourself.
Below is a list of the battleground states in 2008, followed by Obama’s actual winning percentage, Rasmussen’s closest-to-the-election prediction, and the difference between those two (the skew). You can click on the states to check my numbers, which come from Real Clear Politics.
Colorado result: Obama +9, Rasmussen: Obama +4 (skewed McCain by 5 points)
Florida result: Obama +2.8, Rasmussen: McCain +1 (skewed McCain by 3.8 points)
Indiana result: Obama +1.1, Rasmussen: McCain +3 (skewed McCain by 4.1 points)
Iowa result: Obama +9.5, Rasmussen: Obama +8 (skewed McCain by 1.5 points)
Nevada result: Obama +12.5, Rasmussen: Obama +4 (skewed McCain by 8.5 points)
New Hampshire result: Obama +9.6, Rasmussen: Obama +7 (skewed McCain by 2.6 points)
North Carolina result: Obama +0.3, Rasmussen: McCain +1 (skewed McCain by 1.3 points)
Ohio result: Obama +4.6, Rasmussen: tie (skewed McCain by 4.6 points)
Pennsylvania result: Obama +10.3, Rasmussen: Obama +6 (skewed McCain by 4.3 points)
Virginia result: Obama +6.3, Rasmussen: Obama +4 (skewed McCain by 2.3 points)
Wisconsin result: Obama +13.9, Rasmussen: Obama +7 (skewed McCain by 6.9 points)
The remarkable thing here is not that Rasmussen didn’t nail any of those elections. Pollsters only predict within a stated margin of error. If a poll predicts a candidate will get 46 percent, and the margin of error is 3 points, that means there’s a 95 percent probability a candidate’s “true” percentage will fall somewhere between 43 and 49 percent.
What IS remarkable are the following:
- In 5 of the 11 contests, Rasmussen’s final poll showed numbers that were OUTSIDE the margin of error. That’s bizarre. You’d expect them to have a 95 reliability. Instead it’s 55 percent.
- The next one is the biggie: The Rasmussen polls in battleground state ALL were skewed to McCain.
- And that’s how you know there’s an actual “Rasmussen skew.” If 6 of them leaned too much to McCain, and 5 leaned too much Obama’s way, it would be roughly even and there wouldn’t be any evidence of a skew.
FINALLY, because Rasmussen is the most prolific pollster, there’s a good chance that in any given state, that skew is reflected in the Real Clear Politics Average.
So stick that in your back pocket between now and Tuesday and consider the following the next time you read a Rasmussen poll. In the 2008 presidential battleground states:
- They had a only a 55 percent chance of being within the margin of error on the final state poll.
- They had the final outcome wrong in 4 of 11 states.
- And, they ALWAYS skewed to McCain.
October 31, 2012 - 11:00 AM
I cannot wait until this is over. I’d like to go to sleep and wake up on Wednesday.
@1 – Seriously. Just let me know who wins, if there is such a thing.
I agree, Kristen. I’m so tired of this election, I just want it over with.
Dan, I hope that came from somewhere and god forbid you wasted your time putting this together. All polls are BS!
I don’t know..ask Roanoke College. They had a nice result from VA today.
Kristen.
Same here. The best news of all is that my power stayed on during the storm. As for the polling?
They are all just estimates. Just like th BCS. “Play the game”, already. Then we won’t have to speculate.
I don’t know about Rasmussen.
But…most of the polls we are all looking at have the Democrat turnout/advantage at a D+5 or better.
That is nothing short of CRAZY.
Please continue to believe that folks (liberals). Have you all forgotten 2010? Nothing has changed since then. If anything, it has only become WORSE for Obama.
Recent early voting polls show Romney up by 7 points. That is a 23 pt. TURNAROUND from 2008 that goes against Obama.
Gallup is less accurate
Of course Rasmussen is the most inaccurate poll out there. Certainly. No doubt.
Why is that? Well, in my opinion it’s because Rasmussen has been about the only poll which has consistently shown the race tightening over the past month, and also showing Romney going ahead of obuma in a bunch of swing states. That’s why.
In lib-world, “bad-polling” is the only possible explanation.
“I don’t know about Rasmussen.”
Well, LC, why don’t you bother to educate yourself? You can click on the link to each state and see how they muffed the 2008 state forecast in McCain’s favor each IN EVERY BATTELGROUND POLL.
That suggests there’s either a built-in bias to the GOP in their model or they’re deliberately skewing the results. For the record, it appears to be the former.
“Dem Early Vote Lead In Florida Down 70% Over 2008″
http://www.breitbart.com/Big-Government/2012/10/31/Dem-Early-Vote-Lead-Down-In-FL
We all agree with the cliche…”the only poll that matters is on election day”…
But, I remember 2008 for Obama (you do too).
This ain’t 2008. LOL.
LC, Obama most likely will lose Fla. But here’s the thing that’s killing you, and we all know it: That won’t matter. Obama doesn’t need Fla to win the election.
8
Dan, I don’t care about Rasmussen.
I’m looking at ALL the polling in comparison to 2008. This ain’t 2008. This will be much closer to 2010.
Do you really think there will be a D+5 turnout? Because most of the polls I am seeing out there are using that assumption (or higher).
It’s lunacy.
Kristen:
1.”I cannot wait until this is over. I’d like to go to sleep and wake up on Wednesday.”
If you haven’t voted yet, please do just that!
Here are some sites the predict the election using math instead of BS. I’m sure the conservatives will mentally delete “math instead of” when they see the word math. The RCP uses simple averages, so it’s probably the most conservative-friendly.
http://fivethirtyeight.blogs.nytimes.com
http://election.princeton.edu
http://votamatic.org
http://frontloading.blogspot.com
http://research.uvu.edu/DeSart/forecasting/october.html
http://www.realclearpolitics.com/epolls/2012/president/2012_elections_electoral_college_map_no_toss_ups_race_changes.html
“Here are some sites the predict the election using math instead of BS. I’m sure the conservatives will mentally delete “math instead of” when they see the word math. The RCP uses simple averages, so it’s probably the most conservative-friendly.”
Thank you, Dave. I’d like to note that ALL of the data in this blog post relies on polls published by Rasmussen AND actual 2008 election results, as complied by Real Clear Politics (which indeed is the most conservative in terms of making predictions. It’s also the one most subject to skewing by Rasmassen.)
Dan, does the headline “Dewey Defeats Truman” not mean anything to you media types?
Polls are irrelevant, as are early projections do not mean a thing Just ask the inventor of the internet, one Albert Gore. I’m sure he’d tell you that you can’t believe anything until the final vote is counted.
Rick H, that headline does mean something.
That wasn’t just a poll that caused that headline, though. It was veteran DC correspondent for the Chicago Tribune, who assured his bosses that Truman was going to lose.
That from-the-hip judgment was based in part on a Gallup poll. The problem was, Gallop stopped polling 2 weeks before the election with 14 percent of the electorate still undecided.
Polls are only irrelevant AFTER the election. And not always even then.
I liken this election to a FSU-Miami football game.
I wish both of them could lose.
This election will come down to Hamilton County/Cincinnati southern area of Ohio. If the conservative base shows up in that region its over for Obama.
If they don’t turn out, its 4 more years of crap.
Hard to believe one area could decide the election
Sandi, they’re not irrelevant if they help you spot a pollster’s skew by comparing one pollster’s to the actual results!
Basing THIS election turnout on 2008 results is the problem.
Anyone who really believes the Democrats will have a D+7 advantage is smoking Whoopi’s dreadlocks.
Again, Dan, we’re talking ‘polls’ – and do polls count?
No, they do not.
I don’t see your point going after pollsters that do not support your position. There are left leaning polls out there, too.
People vote, and if people pay attention to polls and do not vote because of it . . . . then, perhaps they are not intelligent enough to have the privilege to vote.
LC, I’m not basing this election on the 2008 results. I’m judging the accuracy on Rasmussen polls based on their proven track record of INACCURACY in the 2008 battleground states. And that is pretty substantial.
Rick H, the point is to warn people that if you’re believing in an outcome because of some Rasmussen polls, you might want to be informed that that company has a not-so-stellar track record in the 2008 presidential contest.
The beauty of analyzing this AFTER that election is that the proof is right there in their own numbers.
Who do you think will win, Dan?
LC, all of my lunch/drink/charity-donation wagers are on Obama. I think he’ll win, but it will be very close. And Romney might win the popular vote, but not the EC.
I’ll have a blog post detailing those wagers later this week.
Here’s a nice historical fact for you: The Republican candidate has won every election held on November 6th since 1860. The streak began with the election of Honest Abe.
Its in the stars, baby.
Cell phone users and Hispanics have been under-polled. I still say it’s not as close as the media portrays it. We’ll see.
Here is a PPP poll for North Carolina that says Romney and Obama are EVEN at 49% each.
Q21 If you are a Democrat, press 1. If a Republican,
press 2. If you are an independent or identify
with another party, press 3.
Democrat ……………………………………………….. 48%
Republican………………………………………………. 36%
http://www.publicpolicypolling.com/pdf/2011/PPP_Release_NC_1031.pdf
They are basing their poll on a D+12!
I just think that is hilarious.
Things you won’t see in the Roanoke Times:
Rosy September Employment numbers revised….downward….by half
http://www.reuters.com/article/2012/10/31/us-usa-economy-employment-adp-idUSBRE89U1C620121031
Can’t do anything to hurt the White House this close to an election.
The Roanoke Times is leftist rag, Henry. Of course you’d never see it.
I was under the impression that polling might start to make more sense 10-14 days out; as I assumed the credibility factor might take effect.
However, it seems they are making less sense and leaving “pollsters” a CYA (they should be weathermen”).
Polls are oversampling D’s by at least 7 points (it seems). I understand it is not intentional but factual. They are also showing independents leaning to Romney by double digits and “enthusiasm” for R’s way ahead.
Should obama win, the pollsters will say that they “nailed it”. Should Romney win, the pollsters will say “obama could not make up the enthusiasm gap”. Either way they move forward to the next election getting gobs of money for something that proves useless.
I am with Kristen, let this just be over…
mike O and LC,
If I hear you correctly you are saying that equal numbers of the Repubs and Dems must be sampled to get a fair poll result?
Is that what you believe?
Henry,
I hope you realize those are not the official BLS numbers. Instead, they’re a payroll company’s numbers. Nothing official about them.
Sheeze Dan,
Try reading comprehension 101…
I specifically said that there is oversampling (not intentional).
Let me make it simple, if I “happened” to poll you, shrill, Kristen, gdad and LC, would I get accurate results???
I am not saying it is intentional, just saying that it is skewed.
They should find a way to measure based on realistic D, R, and Ind. numbers.
mike O
If you polled me, Kristen, gdad, shrill and LC the sample size would be WAY WAY WAY to small to assure any kind of accurate county.
But coincidentally, that poll would accurate forecast the the OUTCOME of the election, even if the numbers were off.
I know this is wasted on idiots but what else is there but to keep leaving out facts and hope the light bulb eventually dawns?
“Within each contacted household reached via landline, an interview is sought with an adult 18 years of age or older living in the household who has had the most recent birthday. (This is a method pollsters commonly use to make a random selection within households without having to ask the respondent to provide a complete roster of adults living in the household.) Gallup does not use the same respondent selection procedure when making calls to cell phones because they are typically associated with one individual rather than shared among several members of a household.
When respondents to be interviewed are selected at random, every adult has an equal probability of falling into the sample. The typical sample size for a Gallup poll, either a traditional stand-alone poll or one night’s interviewing from Gallup’s Daily tracking, is 1,000 national adults with a margin of error of ±4 percentage points. Gallup’s Daily tracking process now allows Gallup analysts to aggregate larger groups of interviews for more detailed subgroup analysis. But the accuracy of the estimates derived only marginally improves with larger sample sizes.
After Gallup collects and processes survey data, each respondent is assigned a weight so that the demographic characteristics of the total weighted sample of respondents match the latest estimates of the demographic characteristics of the adult population available from the U.S. Census Bureau. Gallup weights data to census estimates for gender, race, age, educational attainment, and region.”
http://www.gallup.com/poll/101872/how-does-gallup-polling-work.aspx
http://www.washingtonpost.com/blogs/plum-line/post/the-morning-plum-no-pollsters-are-not-conspiring-to-destroy-romney/2012/09/25/cf6c2af4-06fb-11e2-a10c-fa5a255a9258_blog.html
http://blogs.wsj.com/numbersguy/polling-party-1173/
Oddly enough pistol pete predicts that “Hamilton County/Cincinnati southern area of Ohio” will be the key…and that is precisely where the Romney connected voting machines are….Tim where’s your white board?
“In Ohio, Romney needs Hamilton County, especially the Cincinnati suburbs, to offset Obama’s edge in Cleveland, Columbus and Dayton. Obama needs to trim Romney’s advantages in the suburbs and southeastern Ohio’s coal country.”
http://www.realclearpolitics.com/articles/2012/10/21/in_race_to_270_it_may_come_down_to_106_counties_115855-2.html
“…discover that the machines used in Hamilton County, Ohio—the county home of Cincinnati— are supplied by Hart Intercivic…”
http://news.yahoo.com/romney-family-investment-ties-voting-machine-company-could-003209723.html
Coincidence?
I thought that some right winger with super-secret inside info said Obama was coming to Roanoke Sunday. Looks like that one was wrong.
Amazing gdad….I guess their sources aren’t as good as billed.
Bill Clinton To Stump For Obama in Va. on Saturday
He will campaign in Chesapeake and Roanoke and then head to the event in Bristow, according to the Obama campaign.
http://www2.timesdispatch.com/news/2012/nov/01/bill-clinton-stump-obama-va-saturday-ar-2328993/
As usual misinformed RWers…
#40 “Oddly enough pistol pete predicts that “Hamilton County/Cincinnati southern area of Ohio” will be the key…and that is precisely where the Romney connected voting machines are…”
Gee, imagine that.
Sandi,
As usual you completely misunderstand/ignore the point. It has nothing to do with weighted estimates for gender, race, age, education and region. It has to do with samples of D’s and R’s and I’s.
I have seen interviews with gallup management who specifically stated they to not weight for political party, instead the poll the appropriate number and let the party affiliation fall where it may.
No Mike O, as usual, it is not me that chooses to completely misunderstand/ignore the point. From my link #3.
““I don’t believe that weighting by party identification is necessary or advisable,” said John Sides, a political scientist at George Washington University. “The overall balance of party identification in the electorate can fluctuate. There is no steady baseline against which to weight” — unlike Census data used to adjust surveys for demographic attributes such as age or gender.”
What you geniuses are asking for are skewed polls.