Nate Silver of FiveThirtyEight.com confirms Rasmussen Bias...

[roughdraft274]

I checked FiveThirtyEight.com this morning for some quality analysis of all of the elections. Nate Silver made a post about Rasmussen polls that I thought was pretty interesting, and as it is I've always taken their polls with a grain of salt.

Live Blogging Election Night - NYTimes.com

Indeed, Rasmussen polls quite consistently turned out to overstate the standing of Republicans tonight. Of the roughly 100 polls released by Rasmussen or its subsidiary Pulse Opinion Research in the final 21 days of the campaign, roughly 70 to 75 percent overestimated the performance of Republican candidates, and on average they were biased against Democrats by 3 to 4 points.

I haven’t checked this in detail yet, but it appears as though the worst poll of the political cycle will be the Rasmussen Reports survey of Hawaii, which had the incumbent Daniel Inoyue defeating Cam Cavasso by just 13 points. Mr. Inouye is ahead by 55 points right now. If Mr. Inouye’s margin holds, the 42-point error would be by far the worst general election poll in FiveThirtyEight’s database, which includes all polls since 1998; the previous record was 29 points.

That last poll is particularly mind boggling to me. 42 point error? Just guess 50/50 and you're nearly guaranteed to be more accurate than that poll.

 

[Zyphlin]

I'm not going to even try to mess with this one in full because RightinNYC does a better job than anyone, over and over and over again, of showing why this notion is bunk.

Rasmussen has no inherent intentional bias. Their polling method, which focuses on LIKELY voteres rather than registered voters, tends to cause polls to be skewed more in favor of republicans as traditionally republicans are more likely to both self identify AND claim to be likely voters. Typically in polls involving REGSITERED voters you see the slant slightly going the other direction.

 

[roughdraft274]

I'm not going to even try to mess with this one in full because RightinNYC does a better job than anyone, over and over and over again, of showing why this notion is bunk.

Rasmussen has no inherent intentional bias. Their polling method, which focuses on LIKELY voteres rather than registered voters, tends to cause polls to be skewed more in favor of republicans as traditionally republicans are more likely to both self identify AND claim to be likely voters. Typically in polls involving REGSITERED voters you see the slant slightly going the other direction.

That's right, but most polls switch to liely voters when you are two weeks from an election, rasmussen does it all the time. Really no big deal to me. However, that doesn't explain that much of a bias in favor of Republicans.

I'm not just arguing that they are favored to republicans vs. other polls, the argument is that they are skewed right vs. the actual results. When it comes down to the actual results excuses shouldn't matter. The pollster should try to be as accurate as possible, period.

 

[DarkWizard12]

That's right, but most polls switch to liely voters when you are two weeks from an election, rasmussen does it all the time. Really no big deal to me. However, that doesn't explain that much of a bias in favor of Republicans.

I'm not just arguing that they are favored to republicans vs. other polls, the argument is that they are skewed right vs. the actual results. When it comes down to the actual results excuses shouldn't matter. The pollster should try to be as accurate as possible, period.

Do you major in statistics and/or political science?

 

[roughdraft274]

Do you major in statistics and/or political science?

No
3456789

 

[DarkWizard12]

No
3456789

Then, you must learn that this: "The pollster should try to be as accurate as possible" is virtually impossible, due to the type of polliing. It's not accurate measurements pollsters aim for, it's precise measurements.

 

[soccerboy22]

Then, you must learn that this: "The pollster should try to be as accurate as possible" is virtually impossible, due to the type of polliing. It's not accurate measurements pollsters aim for, it's precise measurements.

While that is true, political scientists look for polls that offer the best results. When you have a 42% MOE that is really bad. Not saying that I still won't look at the Rasmussen Polls, because I will, but 42% is not something you look for.

 

[roughdraft274]

Then, you must learn that this: "The pollster should try to be as accurate as possible" is virtually impossible, due to the type of polliing. It's not accurate measurements pollsters aim for, it's precise measurements.

Well, in the example of Rasmussen, if having actual screener do your telephone surveys and actually including cell phones into your list and many other things that Nate has talked about in detail in the past makes your polls more accurate then Rasmussen should consider including some new methods. I'd personally love to see how they arrive at their enthusiasm gap that they weave into their polls.

 

[DarkWizard12]

Well, in the example of Rasmussen, if having actual screener do your telephone surveys and actually including cell phones into your list and many other things that Nate has talked about in detail in the past makes your polls more accurate then Rasmussen should consider including some new methods. I'd personally love to see how they arrive at their enthusiasm gap that they weave into their polls.

there are many methods to conduct a poll, and just about each will yeild different results. Now, was rasmussen's polling off? Yes, there is no doubt about it, was it because of the method, or was it just an anomaly that happens with every poll, and more common this year due to the fact pollsters like to...you know...poll as many races as possible, as many times as possible?

So instead, look at Rasmussen as a whole. I believe it was the 2000/2004 presidential elections that they got almost 100% in each state or...something like that. and was THE MOST accurate in popular vote results. I would hardly disqualify them as pollsters because they got ONE wrong.

In fact, if you look at Rasmussen's polling method, it's pretty much the only one of it's kind. I don't know many others that poll likely voters, much less have a reputation as showing the best case scenario for republicans, which, as we learned in 2004, the best case scenario does sometimes happen.

 

[haymarket]

Rasmussen has no inherent intentional bias.

And you know this because.....? A three to four point consistent bias in a majority of cases is still bias. the word INTENTIONAL that you deem important is impossible to prove either way.

 

[DarkWizard12]

And you know this because.....? A three to four point consistent bias in a majority of cases is still bias. the word INTENTIONAL that you deem important is impossible to prove either way.

that's why he specified "intentional" bias. Rasmussen is bound to have bias, either way, because of it's polling method, which always is skewed toward republicans by the simple fact that republicans vote in higher numbers. This, however, is a completely legitmate and unambiguous method, and for most years, will have the most accurate results (AKA: 2004), AS OPPOSED to other pollsters we know....Dailykos...

 

[haymarket]

And nobody outside of a trusted few in the Rasmussen organization could ever step forward and prove to the righties here that there was this thing you keep calling INTENTIONAL bias. You cannot prove it either way. Give it up already because it means nothing. The guys firm shows bias in favor of Republican candidates. In an of itself that should finish his credibility.

ooooh - unless of course you want bias for some other reason ... like establishing a national narrative picked up by the media as some trend ... but who would do something like that????

 

[Whovian]

funny how the only time I ever see anyone whine about how unfair and biased polls are, is when it appears to favor the GOP, and the whining comes from the Democrats. Can't recall the last time I saw ANY poll which 'favored' or was 'biased' in favor of Democrats where the GOPers whined.

Just saying.

 

[Redbeard]

funny how the only time I ever see anyone whine about how unfair and biased polls are, is when it appears to favor the GOP, and the whining comes from the Democrats. Can't recall the last time I saw ANY poll which 'favored' or was 'biased' in favor of Democrats where the GOPers whined.

Example: Polls favoring Coons over O'Donnell..............No whine...........

 

[haymarket]

Lets see the evidence presented on this issue in the same manner that Nate Silver presented his evidence about Rasmussen and his bias towards the GOP.

 

[RightinNYC]

I've never really understood the argument that you can measure a pollster's accuracy by comparing the results on election day to a poll that was taken 2 or 3 weeks prior to the election.

 

[haymarket]

RightinNYC

your comment seems to ignore the history of the last six months regarding Rasmussen. It was alleged in many quarters - mainly progressive or Democratic ones to be sure - that Rasmussens numbers were outliers. It was suspected by some that this was done to produce a narrative which would be picked up by other persons in the media and help create the basic theme that the GOP candidates were doing better than they actually were. And it was also suspected that there was political motivation behind it. Many of those same people also said repeatedly that the numbers for Rasmussen would begin to fall in line with other pollsters as the election day drew near and he could then fall back on the claim that his polls were accurate.

 

[RightinNYC]

RightinNYC

your comment seems to ignore the history of the last six months regarding Rasmussen. It was alleged in many quarters - mainly progressive or Democratic ones to be sure - that Rasmussens numbers were outliers. It was suspected by some that this was done to produce a narrative which would be picked up by other persons in the media and help create the basic theme that the GOP candidates were doing better than they actually were. And it was also suspected that there was political motivation behind it. Many of those same people also said repeatedly that the numbers for Rasmussen would begin to fall in line with other pollsters as the election day drew near and he could then fall back on the claim that his polls were accurate.

I don't understand how this is a response to my comment.

I'm pointing out one of the difficulties in evaluating the accuracy of polling for any pollster. You're discussing how some people on the left have bitched about Rasmussen without evidence.

 

[haymarket]

But Rasmussen provided the evidence all along.

 

[RightinNYC]

But Rasmussen provided the evidence all along.

Oh, well if you say so.

 

[DarkWizard12]

But Rasmussen provided the evidence all along.

Prove that person x murdered person y
person y is dead
thus, person x murdered him.

Yea, thats your logic at work there.

 

[American]

I've never really understood the argument that you can measure a pollster's accuracy by comparing the results on election day to a poll that was taken 2 or 3 weeks prior to the election.

You can't, because polls can't account for dead Democrats coming out to vote. :lol: