I'm No Pollster, So Please Help Me Out With This

[IGetItAlready]

I've read several items within the past couple of weeks which refer to an over sampling of Democrats in current polls.
I had simply brushed it off being that the claims are not getting a ton of attention. Being that I know little of the process by which these polls are compiled I suppose I figured there was a very good reason for it.

PEW POLL METHODOLOGY QUESTIONED

Polling this year typically includes more self-identified Democrats than Republicans.

The article seems to imply there is not some esoteric reasoning behind this but I've heard it mentioned so often despite virtually no collective outrage over it that I'm wondering if anyone can explain why pollsters might do this, aside from the obvious.

 

[theDoctorisIn]

Polling is done by randomly selected phone numbers. The percentage of "Democrats" and "Republicans" aren't predetermined, they're part of the polling.

 

[Publius1787]

I've read several items within the past couple of weeks which refer to an over sampling of Democrats in current polls.
I had simply brushed it off being that the claims are not getting a ton of attention. Being that I know little of the process by which these polls are compiled I suppose I figured there was a very good reason for it.

PEW POLL METHODOLOGY QUESTIONED

Polling this year typically includes more self-identified Democrats than Republicans.

The article seems to imply there is not some esoteric reasoning behind this but I've heard it mentioned so often despite virtually no collective outrage over it that I'm wondering if anyone can explain why pollsters might do this, aside from the obvious.

They do over sample Democrats. Which is why Reagan, Bush Sr., & Bush Jr. were all (according to the polls) supposed to lose their elections.

 

[theDoctorisIn]

Randomly selected people to be polled is, in fact, the only way to have a poll that means anything.

If you're determining who you're going to poll before you start polling, it defeats the purpose entirely.

 

[g5000]

I've read several items within the past couple of weeks which refer to an over sampling of Democrats in current polls.
I had simply brushed it off being that the claims are not getting a ton of attention. Being that I know little of the process by which these polls are compiled I suppose I figured there was a very good reason for it.

PEW POLL METHODOLOGY QUESTIONED

Polling this year typically includes more self-identified Democrats than Republicans.

The article seems to imply there is not some esoteric reasoning behind this but I've heard it mentioned so often despite virtually no collective outrage over it that I'm wondering if anyone can explain why pollsters might do this, aside from the obvious.

The reason is very simple. There are far more registered Democrats than there are registered Republicans. So if you call 1000 people, you are going to get more registered Democrats than Republicans.


Mystery solved!


.

 

[IGetItAlready]

Randomly selected people to be polled is, in fact, the only way to have a poll that means anything.

If you're determining who you're going to poll before you start polling, it defeats the purpose entirely.

So theoretically a poll could be conducted that excluded one party or the other entirely.

I suppose this is one more reason to put little faith in the polling, this and examples like those Publius mentioned. And it does make perfect sense that you'd not conduct a poll of 500 registered Republicans and 500 registered Democrats as that would only measure party support for each candidate.

Thanks guys.

 

[theDoctorisIn]

I've read several items within the past couple of weeks which refer to an over sampling of Democrats in current polls.
I had simply brushed it off being that the claims are not getting a ton of attention. Being that I know little of the process by which these polls are compiled I suppose I figured there was a very good reason for it.

PEW POLL METHODOLOGY QUESTIONED

Polling this year typically includes more self-identified Democrats than Republicans.

The article seems to imply there is not some esoteric reasoning behind this but I've heard it mentioned so often despite virtually no collective outrage over it that I'm wondering if anyone can explain why pollsters might do this, aside from the obvious.

The reason is very simple. There are far more registered Democrats than there are registered Republicans. So if you call 1000 people, you are going to get more registered Democrats than Republicans.


Mystery solved!


.

See, the thing is, it's not even about party registration - it's self-identification. It's one of the poll questions.

There's no way for Pew Research to actually know their party registration beforehand. Well, there is a way for them to know, but not if they want their polling to mean anything.

 

[kaz]

The reason is very simple. There are far more registered Democrats than there are registered Republicans. So if you call 1000 people, you are going to get more registered Democrats than Republicans.

Exit polling was projecting a sweeping Kerry victory, how did that work out for you?

 

[theDoctorisIn]

Randomly selected people to be polled is, in fact, the only way to have a poll that means anything.

If you're determining who you're going to poll before you start polling, it defeats the purpose entirely.

So theoretically a poll could be conducted that excluded one party or the other entirely.

I suppose this is one more reason to put little faith in the polling, this and examples like those Publius mentioned. And it does make perfect sense that you'd not conduct a poll of 500 registered Republicans and 500 registered Democrats as that would only measure party support for each candidate.

Thanks guys.

It's theoretically possible, yet incredibly unlikely.

Statistics is an interesting subject.

 

[Dick Tuck]

Party Affiliation and Election Polls | Pew Research Center for the People and the Press

 

[kaz]

It's theoretically possible, yet incredibly unlikely.

Statistics is an interesting subject.

I was a math major and also a GE Six Sigma Blackbelt, so statistics and polling are something I'm pretty experienced with. The first rule of polling is that the pollster cannot have a stake in the outcome because through intent or just bias they can skew the results. Too many political polls are run by liberals, which is a big reason why as was accurately pointed out election results tend to go better for the Republican than the polls leading to the election. The most basic issue is "likely voter." Liberal pollsters want to believe that people who pick the Democrat are going to vote, but the reality is that it's hard to predict, which is why bias that's not necessarily intentional can come in to play.

A statistician is sitting with one foot in scalding hot water and the other in freezing ice water. Someone asks him how he's doing, he says on average, I'm comfortable...

 

[healthmyths]

I've read several items within the past couple of weeks which refer to an over sampling of Democrats in current polls.
I had simply brushed it off being that the claims are not getting a ton of attention. Being that I know little of the process by which these polls are compiled I suppose I figured there was a very good reason for it.

PEW POLL METHODOLOGY QUESTIONED

Polling this year typically includes more self-identified Democrats than Republicans.

The article seems to imply there is not some esoteric reasoning behind this but I've heard it mentioned so often despite virtually no collective outrage over it that I'm wondering if anyone can explain why pollsters might do this, aside from the obvious.

Here is a CBS poll
CBS News State of the Union Poll - CBS News

An overwhelming majority of Americans approved of the overall message in President Obama's State of the Union speech on Tuesday night, according to a CBS News poll of speech watchers.
91 percent of those who watched the speech approved of the proposals Mr. Obama put forth during his remarks. Only nine percent disapproved.

The above CBS poll of 659 people.

44% or 290 were Democrats
31% or 204 were Independent
25% or 165 were GOP

 

[IGetItAlready]

Party Affiliation and Election Polls | Pew Research Center for the People and the Press

Very cool Dick, thanks.

I've heard first hand how many older voters STILL consider the Democrat party to be the party of JFK which is clearly not the case. That may account for some of the difference in registered Dems to Reps.

With all these variables and all the instances we've seen in which the polling information simply didn't hold true on election day, it's odd that we put any faith in these polls at all.

 

[theDoctorisIn]

It's theoretically possible, yet incredibly unlikely.

Statistics is an interesting subject.

I was a math major and also a GE Six Sigma Blackbelt, so statistics and polling are something I'm pretty experienced with. The first rule of polling is that the pollster cannot have a stake in the outcome because through intent or just bias they can skew the results. Too many political polls are run by liberals, which is a big reason why as was accurately pointed out election results tend to go better for the Republican than the polls leading to the election. The most basic issue is "likely voter." Liberal pollsters want to believe that people who pick the Democrat are going to vote, but the reality is that it's hard to predict, which is why bias that's not necessarily intentional can come in to play.

A statistician is sitting with one foot in scalding hot water and the other in freezing ice water. Someone asks him how he's doing, he says on average, I'm comfortable...

"Likely voters" generally means randomly selected people from a list of people who've voted in 2 of the last 4 elections.

"Likely voters" generally returns results that are slightly biased towards Republicans over Democrats. For instance, in 2008, "likely voter" polls were skewed towards McCain, due to the large number of new voters in 2008.

It's not a simple thing, bias in polling.

 

[kaz]

It's theoretically possible, yet incredibly unlikely.

Statistics is an interesting subject.

I was a math major and also a GE Six Sigma Blackbelt, so statistics and polling are something I'm pretty experienced with. The first rule of polling is that the pollster cannot have a stake in the outcome because through intent or just bias they can skew the results. Too many political polls are run by liberals, which is a big reason why as was accurately pointed out election results tend to go better for the Republican than the polls leading to the election. The most basic issue is "likely voter." Liberal pollsters want to believe that people who pick the Democrat are going to vote, but the reality is that it's hard to predict, which is why bias that's not necessarily intentional can come in to play.

A statistician is sitting with one foot in scalding hot water and the other in freezing ice water. Someone asks him how he's doing, he says on average, I'm comfortable...

"Likely voters" generally means randomly selected people from a list of people who've voted in 2 of the last 4 elections.

"Likely voters" generally returns results that are slightly biased towards Republicans over Democrats. For instance, in 2008, "likely voter" polls were skewed towards McCain, due to the large number of new voters in 2008.

It's not a simple thing, bias in polling.

No, it's not simple. The polls leading into the election though did still have McCain losing. It was nothing like the Kerry exit polls. Though those apparently were rigged.

 

[theDoctorisIn]

I was a math major and also a GE Six Sigma Blackbelt, so statistics and polling are something I'm pretty experienced with. The first rule of polling is that the pollster cannot have a stake in the outcome because through intent or just bias they can skew the results. Too many political polls are run by liberals, which is a big reason why as was accurately pointed out election results tend to go better for the Republican than the polls leading to the election. The most basic issue is "likely voter." Liberal pollsters want to believe that people who pick the Democrat are going to vote, but the reality is that it's hard to predict, which is why bias that's not necessarily intentional can come in to play.

A statistician is sitting with one foot in scalding hot water and the other in freezing ice water. Someone asks him how he's doing, he says on average, I'm comfortable...

"Likely voters" generally means randomly selected people from a list of people who've voted in 2 of the last 4 elections.

"Likely voters" generally returns results that are slightly biased towards Republicans over Democrats. For instance, in 2008, "likely voter" polls were skewed towards McCain, due to the large number of new voters in 2008.

It's not a simple thing, bias in polling.

No, it's not simple. The polls leading into the election though did still have McCain losing. It was nothing like the Kerry exit polls. Though those apparently were rigged.

Exit polling is particularly inaccurate, due to the fact that you can't talk to every single person leaving the polling place, so the pollsters have to choose people to talk to - which itself tends towards bias.

 

[kaz]

Exit polling is particularly inaccurate, due to the fact that you can't talk to every single person leaving the polling place, so the pollsters have to choose people to talk to - which itself tends towards bias.

Republican voters are more likely also to blow off the pollster. Democrats love telling everyone they are sheep and proud of it.

I live in North Carolina and I get called by pollsters all the time. I have yet to answer the phone. I'm voting Romney because Obama's a Marxist. I'd voted third party/independent for President every election since 1988.

 

[g5000]

For the sake of exercise, let's stipulate all Democrats hate donuts, and all Republicans love donuts. Half of Indendents hate donuts while half love them.

And let's say 50 percent of Americans are registered Democrats, 30 percent are registered Republicans, and 20 percent are Independents.

If you call 100 Americans and ask them how they feel about donuts, at least 60 of them (50 Democrats and 10 Independents) are going to tell you they hate donuts, and 40 (30 Republicans and 10 Independents) will tell you they love donuts.


If the polling agency reports "60 percent of Americans" hate donuts, that would be a legitimate result.

If a Republican complained that for every 3 Republicans that were polled there were 5 Democrats polled, that would not be a legitimate complaint. The poll is reflecting the actual demographics of Americans.


As I said above, there are far more registered Democrats than there are registered Republicans. So of course if you call 1000 Americans, a lot more will "self-identify" as Democrats than Republicans.


Mystery solved!



.

 

[kaz]

For the sake of exercise, let's stipulate all Democrats hate donuts, and all Republicans love donuts. Half of Indendents hate donuts while half love them.

...

If the polling agency reports "60 percent of Americans" hate donuts, that would be a legitimate result

Which doesn't explain why polls strongly skew left. A good argument at least covers empirical data...

 

[blackhawk]

One reason I have heard for the oversampling of Democrats in these polls is and it goes back to the phone sampling is they tend to call more cell numbers than home phones which tends to lead to younger people who usually vote Democrat. If this true I don't know is just one reason I heard about.