CNN Dead Heat Poll - Are The Results Skewed?

Conflicting information has made me flip flop as to whether the CNN poll is skewed or not. You folks can argue it out here, I give up!

Before I hit the submit button for this thread, I hesitated for some reason. And I should have heeded my gut feelings.

----------------------------------------------------------------------------------------------------

I found it amazing that CNN admitted that Obama and Romney were in a dead heat. And was actually quite happy to see them put this in print. (It is common knowledge that CNN leans left.) Until I noticed one thing missing from their article, the party affiliations. So I decided to go digging into their actual poll data. And I found that there are 11% more Democrats sampled than republicans!

Forty-nine percent of likely voters questioned say they support the president, with an equal amount saying they back the former Massachusetts governor.

CNN National Poll: Dead heat between Obama and Romney

The "likely voters" are what was used to calculate the poll results that offered up the "dead heat". Out of 1,010 adults interviewed, 693 were "likely voters". This gives you 284 democrats (41%) and 207 (30%) republicans sampled. A difference of a shocking 11%. So by utilizing 77 more Democrats, from 491 total democrats and republicans questioned, they were able to accomplish a dead heat!

Respondents who reported that they had already cast an absentee ballot or voted early were automatically classified as likely voters.

Among those likely voters, 41% described themselves as Democrats, 29% described themselves as Independents, and 30% described themselves as Republicans

CNN Poll Data

You would have a good point if 50% of the country was democrat and 50% of the country was republican. But that is not so. You have a little more digging to do.

reply to post by theconspirator


You have a very valid point. And you have sparked my interest. Thank you!

Edited to add... I just imagined that it would be somewhat close. But what concerns me more is that no one in our country has a clue what the accurate number of Democrats vs. Republicans is. How on earth can anyone do an accurate poll if folks are still arguing over this point?

reply to post by elouina

It is common knowledge that CNN leans left.

No it's not.

MSNBC, yes....CNN...nope.

CNN is the most centrist news channel we have.

So I decided to go digging into their actual poll data. And I found that there are 11% more Democrats sampled than republicans!

That's because there are more Democrats in America than there are Republicans. This is a fact that has been confirmed by poll after poll.

Do you know how sampling works for polls? They don't go out and look for a certain number of Democrats and Republicans, they call RANDOM people and ask how they identify themselves. Poll after poll comes back with more identifying as Democrats.

This whole crying about Democrats being "oversampled", to me it just shows that the person saying that doesn't understand polling, statistics, or random sampling.

I bet you really don't like that when they look the entire sample and not just "likely voters", Obama is up 2%. The desperation is just staggering.

What will the excuse be if after the election, that it turns out that the polls were all correct? Will the election have "oversampled" Democrats too?

reply to post by elouina


The electorate is not D+11, Not even D+8, it's more like R + 3.

Here's how Rasmussen had party affiliations from the 2008 elections vs 2012 elections.

Summary of Party Affiliation (Rasmussen)

September 30, 2012
Republican Democrat Other R - D Quarterly
Sep 36.8% 34.2% 29.0% 2.6% 2.6%

R+2.6%

October 31, 2008
Republican Democrat Other R - D Quarterly
Oct 33.3% 40.3% 26.4% -7.1%

R-7.1%

Summary of Party Affiliation

reply to post by elouina


Here's a couple good threads:

Cnn/Orc Poll: Unskewed...Romney Up 8

Polls Debunked: Democrats Oversampled

reply to post by alternateuniverse


But everything I looked at shows more Democrats. Doesn't anyone publish an accurate number?

Edited to add... Well I take this statement back. Not everything shows this. What I was looking at was outdated.

reply to post by alternateuniverse


Most pollsters, with the exception or Rasmussen and Gallup, project turnout similar to 2008. Rasmussen and Gallup are the only two major pollsters who make adjustments based on current party affiliation.

The 2012 turnout will not be similar to 2008.

reply to post by elouina


there are more democrats in the country then republicans. So to get accurate statistical data, the poll must have more democrats.

Yeah, but not D+11. Obama beat McCain in 2008 by 7.6. It's 2012.

Originally posted by alternateuniverse
reply to post by alternateuniverse

 

Most pollsters, with the exception or Rasmussen and Gallup, project turnout similar to 2008. Rasmussen and Gallup are the only two major pollsters who make adjustments based on current party affiliation.

The 2012 turnout will not be similar to 2008.

Rasmussen and Gallup poll land lines only and use robo-calling.

No surprise they get more Republicans with that method.

Bottom Line:
A dead heat poll, with D + 11. Good news for Romney.

Impossible that there are more Democrats.

Democrats, from what I have been informed, are leftist communist socialist marxist etc.

And I have been told matter of factly this country is center right...

I am starting to think the right wing folks may not have been telling the whole truth in their claims..which confuses me also, because they are Gods party..and since when is God a fibber?

-walks away all confused-
next someone will be telling be GWB wasn't the greatest president ever..my entire worldview is shaken.

reply to post by elouina


Stratified sampling is quite common when polling. Stratification looks like oversampling, but it is planned and the effect of the stratification can, and should, be removed when reporting the results. The reason this is done is to get more precise information on key, small subgroups within the population being sampled. What makes them "key" is that they might be of special interest or the client who paid for a poll may have asked for more information about specific subgroups.

You would have to get into the details of the analysis to determine if the reporting of the results was biased due to uncorrected stratification. That type of information is rarely reported by the media -- very few people would understand it. Actually, most people have difficulty seeing how a sample of 1100 people couple possibly represent the opinions of the entire country. The truth is, samples (even small ones) are frequently much more accurate than a census (asking everyone).

Anyway, if there was stratification and the raw results were reported, the impression provided to the reader would be biased. I haven't gone over the details of this particular poll, so I can't speak about it directly. I will say though that it is possible that a reputable polling organization could be pressured into reporting raw results through the threat of lawsuits. It's also possible, but very unliekly for a reputable firm, that they may simply want to assist a particular candidate by making him/her look better than they are. It has certainly been claimed that some, possibly all, major media are both manufactuirng and skewing news at the request of specific governments and political parties / politicians. In this particular poll? Who knows!

Reporting raw results only even when stratification corrected results are available would be a skewing of the news that might appear defendible to many people not familiar with statistical methods.

Originally posted by alternateuniverse
The electorate is not D+11, Not even D+8, it's more like R + 3.

I'm expecting something closer to +3 Democrat. Given that the polls are showing it a statistical tie, generally with +6 to +8 Democrat samples, obviously I expect Romney to outperform the polling data by a few points. Whether it will be enough or not, I don't know, but I am cautiously optimistic. I think Romney is going to win the popular vote something like 50-48 and the electoral college by a narrow margin as well.

Originally posted by Taiyed

Originally posted by alternateuniverse
reply to post by alternateuniverse

 

Most pollsters, with the exception or Rasmussen and Gallup, project turnout similar to 2008. Rasmussen and Gallup are the only two major pollsters who make adjustments based on current party affiliation.

The 2012 turnout will not be similar to 2008.

Rasmussen and Gallup poll land lines only and use robo-calling.

No surprise they get more Republicans with that method.

The last poll I read on Gallup used 50% cell phones and 50% landlines. And Rasmussen uses landlines and online polling.