Tea Party Candidate Mitt Romney Has Negative Approval Rating

[Phatscotty]

Will they respond with their typical violence?

Hmmm, this looks like a guy a lot of dispossessed Democrat's might vote for....

btw David, I don't think it's very healthy for you to be talking about violence every single day. It worries me. It isn't very healthy for our community either.

Could you please take a break from the daily spamming about violence?

[rdsrds2120]

Is Romney's negative approval rating as pathetic as Obama's?

Just as I expected, only half of what is polled. Here is the full information. http://www.rasmussenreports.com/public_ ... ex_history Also, Scotty this is a thread about Mitt not Obama, so please stop derailing thread and posting half information.

Rasmussen polls on a gradient of options, but it uses only the strongest opinions when calculating the plus/minus of the favorability rating. The theory being that those who strongly approve or disapprove are less likely to change their minds, so if they are skewed one direction, it means that the person has less middle room to gain people to his side.

And besides my explanation above, showing all the numbers is actually a worse indicator for the president. Only 10% of those polled somewhat disapprove of his work while only 20% somewhat approve of his work. This means there are a lot more people who could be swayed against him than those who can be persuaded to move to him. And his total disapproval is already over 50%.

Oops, forgot to debunk this when it was posted.

While in Theory we can say "Hey, that makes sense kind of", the way that Rasmussen polls their population sample ignores a golden rule of polling, and that's that the Simple, Random Sample yields the best results the most consistently (there are small deviations that are acceptable, but an SRS will leave you best off).

Furthermore, Rasmussen Polls are automated phone polling. In this day and age, when most poll systems can't either can't get to cell phones, or when they do, aren't answered, it creates a skew in the population sample and usually means you can toss the numbers out.

The biggest problems here are cultural tendencies. For example, it can be conjectured that cell phone usage is more popular among the younger generation than the old, and that not many young voters even have a landline. As we all know, there is a high tendency for younger voters to be more left-leaning/liberal.

So, not accurately including those with cell phones + those with cell phones may tend to be left/leaning = skew in poll. If you think this sounds absurd, it really isn't. The most famous example in history of the non-random/non-stratified samples (which also used telephone sampling, btw) was the big ordeal with Dewey vs. Truman, where the Chicago Tribune declared Dewey as President (because their polls showed him winning in a landslide), then Truman mopped the floor with him in a final electoral vote of 301-189.

tl;dr -- Rasmussen doesn't poll accurately, leaves out a left-leaning group, creating a skew.

-rd

edit: submitted and accidentally left some stuff out. Not all telephone polls are bad. I'll get back to it later.

[Night Strike]

Rasmussen polls "likely voters" because when you're polling political issues, those are the people who actually matter. It doesn't make sense to poll a 100% random selection simply for the sake of selecting from a pool of all people, whether they are interested in political subjects or not. Most other political polls will poll from registered voters and not all adults.

The point about people who only use cell phones not being included in the poll is valid, but doesn't this apply to all other major polling groups? Even if you only poll landline individuals, you can still look into the data behind the poll numbers in order to analyze the political leanings of those who were polled.

[rdsrds2120]

Rasmussen polls "likely voters" because when you're polling political issues, those are the people who actually matter.

It doesn't make sense to poll a 100% random selection simply for the sake of selecting from a pool of all people, whether they are interested in political subjects or not. Most other political polls will poll from registered voters and not all adults.

Ok, I can level on those two. In my post, I wasn't clear what I was referring to when I said what should be randomized: the population sample.

Population in how I used it was semi-vague. It could have referred to the population of likely voters, people in the US, or even the people in Uzbeki-beki-beki-stan-stan-stan. I was referring to the population as being whichever target group (likely voters, in this case, sorry) and how they should be polled, not whether they're the right sample. Mybad.

The point about people who only use cell phones not being included in the poll is valid, but doesn't this apply to all other major polling groups?

That's what I was going to expand on when I got the chance. To an extent, yes. However, certain measures can be taken in a poll stratified for political alignment, race, class, etc. so that the effect is extremely low (Gallup does a lot of this, they actually hang up if they have too many people of a certain group, to make it more balanced). I looked around, and it didn't look like Rasmussen did this. It would pass for a high school or college freshmen level stats project, but not something I'd look to for accuracy.

Even if you only poll landline individuals, you can still look into the data behind the poll numbers in order to analyze the political leanings of those who were polled.

Well, yeah, I wasn't saying that you couldn't do that! I was saying if you did that or something that leaned towards landlines, you couldn't use it to extrapolate data onto others that aren't included in that group (which includes part of the likely voter in U.S. population). So, any conclusion from those numbers would have to state "In this poll blah blah, X +/- Y% of people with a landline...[rest]."

-rd

[Night Strike]

The point about people who only use cell phones not being included in the poll is valid, but doesn't this apply to all other major polling groups?

That's what I was going to expand on when I got the chance. To an extent, yes. However, certain measures can be taken in a poll stratified for political alignment, race, class, etc. so that the effect is extremely low (Gallup does a lot of this, they actually hang up if they have too many people of a certain group, to make it more balanced). I looked around, and it didn't look like Rasmussen did this. It would pass for a high school or college freshmen level stats project, but not something I'd look to for accuracy.

In your previous post, you stated the a simple, random sample set would provide the most consistent data. How can you do that AND select the participants based on race, class, etc.?

Even if you only poll landline individuals, you can still look into the data behind the poll numbers in order to analyze the political leanings of those who were polled.

Well, yeah, I wasn't saying that you couldn't do that! I was saying if you did that or something that leaned towards landlines, you couldn't use it to extrapolate data onto others that aren't included in that group (which includes part of the likely voter in U.S. population). So, any conclusion from those numbers would have to state "In this poll blah blah, X +/- Y% of people with a landline...[rest]."

Maybe that's why so many polls over-sample Democrats: to account for most young people being Democrats (although I don't buy it). However, when it comes to actual "likely voters" pool in the Rasmussen polling, I'd say that you're currently at a "push" (draw) when it comes to young voters (the ones who don't have landlines). The first election where they came out in major (and influential) numbers was in 2008 for Obama. They didn't vote in the 2010 mid-terms. We don't currently know for sure how young people will vote or if they even will come out to vote this year, so I don't think the polling methods (especially for "likely voters") needs changed at this time. Until there is an adequate set of data to indicate that the actual votes are significantly different from the polling numbers due to people without landlines not being sample, I think the standard margin of error adequately accounts for that specific uncertainty.

[rdsrds2120]

The point about people who only use cell phones not being included in the poll is valid, but doesn't this apply to all other major polling groups?

That's what I was going to expand on when I got the chance. To an extent, yes. However, certain measures can be taken in a poll stratified for political alignment, race, class, etc. so that the effect is extremely low (Gallup does a lot of this, they actually hang up if they have too many people of a certain group, to make it more balanced). I looked around, and it didn't look like Rasmussen did this. It would pass for a high school or college freshmen level stats project, but not something I'd look to for accuracy.

In your previous post, you stated the a simple, random sample set would provide the most consistent data. How can you do that AND select the participants based on race, class, etc.?

This is a good question. Earlier when I said that, it was in reference to just one group of people not divided by anything other than accessibility by the person polling. However, sometimes we need more info, or we want to narrow down our info. To do so, we create subgroups to become a part of the polling process so that they can be accurately reflected in the summary/conclusion, and conduct a different type of sampling known as stratified random sampling.

So, one generic group of people to reach an opinion: Simple random sample.
Many subgroups that form a group, and you want to know information from each subgroup: stratified sampling.

Here's a very basic example of stratified sampling:
http://en.wikipedia.org/wiki/Stratified ... al_example (it was the first one to popup when I googled for an example, because I'm lazy

Even if you only poll landline individuals, you can still look into the data behind the poll numbers in order to analyze the political leanings of those who were polled.

Well, yeah, I wasn't saying that you couldn't do that! I was saying if you did that or something that leaned towards landlines, you couldn't use it to extrapolate data onto others that aren't included in that group (which includes part of the likely voter in U.S. population). So, any conclusion from those numbers would have to state "In this poll blah blah, X +/- Y% of people with a landline...[rest]."

Maybe that's why so many polls over-sample Democrats: to account for most young people being Democrats (although I don't buy it). However, when it comes to actual "likely voters" pool in the Rasmussen polling, I'd say that you're currently at a "push" (draw) when it comes to young voters (the ones who don't have landlines). The first election where they came out in major (and influential) numbers was in 2008 for Obama. They didn't vote in the 2010 mid-terms. We don't currently know for sure how young people will vote or if they even will come out to vote this year, so I don't think the polling methods (especially for "likely voters") needs changed at this time. Until there is an adequate set of data to indicate that the actual votes are significantly different from the polling numbers due to people without landlines not being sample, I think the standard margin of error adequately accounts for that specific uncertainty.

"Open this box with the crowbar you will find inside"

The only way we'd be able to know if it was wrong is if we did it the right way, but you're saying we shouldn't, because there's no evidence to suggest it's wrong....

-rd

[Night Strike]

"Open this box with the crowbar you will find inside"

The only way we'd be able to know if it was wrong is if we did it the right way, but you're saying we shouldn't, because there's no evidence to suggest it's wrong....

-rd

I'm pretty sure that most polling entities are barred from calling cell phones, so it's hard (illegal) to even factor that into the polling sample to begin with. And until there are actual votes that lend credence to the current polling methods being inadequate (and that will happen eventually), I don't think we need to change the laws or add additional disclaimers.

[rdsrds2120]

"Open this box with the crowbar you will find inside"

The only way we'd be able to know if it was wrong is if we did it the right way, but you're saying we shouldn't, because there's no evidence to suggest it's wrong....

-rd

I'm pretty sure that most polling entities are barred from calling cell phones, so it's hard (illegal) to even factor that into the polling sample to begin with. And until there are actual votes that lend credence to the current polling methods being inadequate (and that will happen eventually), I don't think we need to change the laws or add additional disclaimers.

I didn't suggest we do. It's just an inconvenient uncertainty.

-rd

[Phatscotty]