Welcome, Guest. Please login or register.
Did you miss your activation email?
July 30, 2014, 10:20:57 pm
HomePredMockPollEVCalcAFEWIKIHelpLogin Register
News: Atlas Hardware Upgrade complete October 13, 2013.

+  Atlas Forum
|-+  Election Archive
| |-+  2012 Elections
| | |-+  2012 U.S. Presidential General Election Polls (Moderators: Tender Branson, Sheriff Buford TX Justice)
| | | |-+  .
« previous next »
Pages: [1] Print
Author Topic: .  (Read 816 times)
.
bawlexus91
Rookie
*
Posts: 21
Political Matrix
E: 4.00, S: -3.30

View Profile
« on: May 24, 2012, 09:18:26 am »
Ignore

.
« Last Edit: June 07, 2012, 01:51:58 am by . »Logged
ajb
YaBB God
*****
Posts: 872
United States


View Profile
« Reply #1 on: May 24, 2012, 10:43:39 am »
Ignore

Actual data from poll here:

http://msnbcmedia.msn.com/i/MSNBC/Sections/A_Politics/_Today_Stories_Teases/NBC%20News-Marist_Poll_Florida.pdf

Which makes clear what is implied in the blog post, which is that this is a poll of registered voters.
Actual voter registration numbers in Florida as of April 2012:

D 41
R 36
Minor 3
Other 21

http://election.dos.state.fl.us/nvra/affiliation.asp

Logged
ajb
YaBB God
*****
Posts: 872
United States


View Profile
« Reply #2 on: May 24, 2012, 10:53:09 am »
Ignore

Actual data from poll here:

http://msnbcmedia.msn.com/i/MSNBC/Sections/A_Politics/_Today_Stories_Teases/NBC%20News-Marist_Poll_Florida.pdf

Which makes clear what is implied in the blog post, which is that this is a poll of registered voters.
Actual voter registration numbers in Florida as of April 2012:

D 41
R 36
Minor 3
Other 21

http://election.dos.state.fl.us/nvra/affiliation.asp



Interesting, so based on actual 2008 turnout, AND state party registration, the sample is still a little over DEM.

Actual 2008 turnout is, of course, irrelevant to a poll of registered voters.
Logged
M
YaBB God
*****
Posts: 2496


View Profile
« Reply #3 on: May 24, 2012, 11:00:05 am »
Ignore

Actual data from poll here:

http://msnbcmedia.msn.com/i/MSNBC/Sections/A_Politics/_Today_Stories_Teases/NBC%20News-Marist_Poll_Florida.pdf

Which makes clear what is implied in the blog post, which is that this is a poll of registered voters.
Actual voter registration numbers in Florida as of April 2012:

D 41
R 36
Minor 3
Other 21

http://election.dos.state.fl.us/nvra/affiliation.asp



Interesting, so based on actual 2008 turnout, AND state party registration, the sample is still a little over DEM.

Actual 2008 turnout is, of course, irrelevant to a poll of registered voters.

A poll of registered voters is, of course, irrelevant to life on Earth.
Logged

Recently moved to Jackson, Mississippi.
ajb
YaBB God
*****
Posts: 872
United States


View Profile
« Reply #4 on: May 24, 2012, 11:07:30 am »
Ignore

Actual data from poll here:

http://msnbcmedia.msn.com/i/MSNBC/Sections/A_Politics/_Today_Stories_Teases/NBC%20News-Marist_Poll_Florida.pdf

Which makes clear what is implied in the blog post, which is that this is a poll of registered voters.
Actual voter registration numbers in Florida as of April 2012:

D 41
R 36
Minor 3
Other 21

http://election.dos.state.fl.us/nvra/affiliation.asp



Interesting, so based on actual 2008 turnout, AND state party registration, the sample is still a little over DEM.

Actual 2008 turnout is, of course, irrelevant to a poll of registered voters.

A poll of registered voters is, of course, irrelevant to life on Earth.

Less so than a poll of likely voters six months before an election. Voters themselves don't know if they're going to vote in November (at least those on the margins of voting/not voting).
Logged
ajb
YaBB God
*****
Posts: 872
United States


View Profile
« Reply #5 on: May 24, 2012, 11:09:26 am »
Ignore

Actual data from poll here:

http://msnbcmedia.msn.com/i/MSNBC/Sections/A_Politics/_Today_Stories_Teases/NBC%20News-Marist_Poll_Florida.pdf

Which makes clear what is implied in the blog post, which is that this is a poll of registered voters.
Actual voter registration numbers in Florida as of April 2012:

D 41
R 36
Minor 3
Other 21

http://election.dos.state.fl.us/nvra/affiliation.asp



Interesting, so based on actual 2008 turnout, AND state party registration, the sample is still a little over DEM.

Actual 2008 turnout is, of course, irrelevant to a poll of registered voters.

Says who? But even if I agreed with you, then this poll of registered voters has still produced a sample that is D+8, which is still much more Democratic than actual 2008 turnout, and hugely more Democratic than '04. Also more Democratic than actual party registration, as you pointed out.
Because a poll of registered voters polls registered voters. If you weighted the poll according to 2004 turnout, or even 2008 turnout, then called it a poll of registered voters, you'd be guilty of a gross distortion.
Logged
ajb
YaBB God
*****
Posts: 872
United States


View Profile
« Reply #6 on: May 24, 2012, 11:17:43 am »
Ignore

Actual data from poll here:

http://msnbcmedia.msn.com/i/MSNBC/Sections/A_Politics/_Today_Stories_Teases/NBC%20News-Marist_Poll_Florida.pdf

Which makes clear what is implied in the blog post, which is that this is a poll of registered voters.
Actual voter registration numbers in Florida as of April 2012:

D 41
R 36
Minor 3
Other 21

http://election.dos.state.fl.us/nvra/affiliation.asp



Interesting, so based on actual 2008 turnout, AND state party registration, the sample is still a little over DEM.

Actual 2008 turnout is, of course, irrelevant to a poll of registered voters.

Says who? But even if I agreed with you, then this poll of registered voters has still produced a sample that is D+8, which is still much more Democratic than actual 2008 turnout, and hugely more Democratic than '04. Also more Democratic than actual party registration, as you pointed out.
Because a poll of registered voters polls registered voters. If you weighted the poll according to 2004 turnout, or even 2008 turnout, then called it a poll of registered voters, you'd be guilty of a gross distortion.

Whether it's of registered voters or not, the number of Democrats polled exceeds the number that turned out in 2008. That fact isn't dependent on whether the poll is of registered voters or anything else. So yes, the poll is of  registered voters. And yes, the registered voter sample is 5 points more Democratic than '08 turnout was

So the question is, what do you think about polls of registered voters? That's the only question at issue here. I think there's value to such polls six months out. Do you?
Logged
Niemeyerite
JulioMadrid
YaBB God
*****
Posts: 6472
Spain


View Profile
« Reply #7 on: May 24, 2012, 12:21:53 pm »
Ignore

Bawlexus, not everything is wrong if the party sample doesn't match the 2008 one. Stop reading on party ID, please, it's becoming annoying. Are you umengus or what??

 
Logged

My evolution (by The Political Matrix):
E: -6.06 -> -6.97 -> -6.97 -> -8.13 -> -7.29 -> -8.26 -> -8.65 -> -7.03
S: -6.78 -> -6.09 -> -7.30 -> -7.13 -> -8.09 -> -8.35 -> -9.04 -> -8.61
HagridOfTheDeep
YaBB God
*****
Posts: 4768
Canada


Political Matrix
E: 3.35, S: 0.52

View Profile
« Reply #8 on: May 24, 2012, 12:44:55 pm »
Ignore

Well, I appreciate your analyses.
Logged

M
YaBB God
*****
Posts: 2496


View Profile
« Reply #9 on: May 24, 2012, 01:14:50 pm »
Ignore

Bawlexus, not everything is wrong if the party sample doesn't match the 2008 one. Stop reading on party ID, please, it's becoming annoying. Are you umengus or what??

 

Not everything is wrong, but the poll is considerably less accurate.

In fact, most likely the actual electorate will be less +D than 2008 in most states, and the most accurate pollsters will correctly predict this. However, more +D than 2008, a blowout Democratic year, is a highly unlikely scenario. Without accounting for this unaccountable methodology, the polls employing it become actively misleading.
Logged

Recently moved to Jackson, Mississippi.
ajb
YaBB God
*****
Posts: 872
United States


View Profile
« Reply #10 on: May 24, 2012, 02:21:50 pm »
Ignore

Bawlexus, not everything is wrong if the party sample doesn't match the 2008 one. Stop reading on party ID, please, it's becoming annoying. Are you umengus or what??

 

Not everything is wrong, but the poll is considerably less accurate.

In fact, most likely the actual electorate will be less +D than 2008 in most states, and the most accurate pollsters will correctly predict this. However, more +D than 2008, a blowout Democratic year, is a highly unlikely scenario. Without accounting for this unaccountable methodology, the polls employing it become actively misleading.

The trouble with this is that it involves circular reasoning. A pollster has a hunch about which voters will and won't turn out in November, and weights their sample accordingly. The results appear to confirm the original hypothesis about turnout -- but they don't, since the sample was constructed according to that hypothesis.

2008 provides concrete evidence that ignoring this logical fallacy can have real consequences. Pollsters who assumed all year that the turnout in Nov. was going to look like 2004 either got burned, or had to make last-minute adjustments to their methods in order to account for what was obviously a very different year.

I'm not saying that 2012's turnout will look like 2008. Nor am I saying it will look like 2004, or 2010, or that it will be more Democratic than 2008. All of these are possibilities, and it's far too early simply to guess. Much better to poll registered voters for now and then apply a rigorous likely-voter screen come Labor Day or so.
Logged
Niemeyerite
JulioMadrid
YaBB God
*****
Posts: 6472
Spain


View Profile
« Reply #11 on: May 24, 2012, 03:51:36 pm »
Ignore

Bawlexus, not everything is wrong if the party sample doesn't match the 2008 one. Stop reading on party ID, please, it's becoming annoying. Are you umengus or what??

 

JulioMadrid, this is a poll forum where people discuss polling. I do not imply that party samples disqualify any polls, just that they are somehing worth considering. Just pointing out the data. Don't really care whether I annoy you or not, and stop telling me what to do. Please just ignore my posts (as I'll kindly do yours).

OK, you can point out the data, give your analysis or whatever you want to. But saying that a poll is accurate or not based on party ID all the time around (specially with PPP) IS annoying.
Read some umengus posts, then think about them. Don't you think he's a hack, or that his posts are irritating? Well, I'm sorry to write this, but you're more or less the same.

But, of course, you decide what you post. I have nothing against that. But people here aren't dumb. They (we) know when a poll is trash.
Logged

My evolution (by The Political Matrix):
E: -6.06 -> -6.97 -> -6.97 -> -8.13 -> -7.29 -> -8.26 -> -8.65 -> -7.03
S: -6.78 -> -6.09 -> -7.30 -> -7.13 -> -8.09 -> -8.35 -> -9.04 -> -8.61
NVGonzalez
antwnzrr
YaBB God
*****
Posts: 1753
Mexico


View Profile
« Reply #12 on: May 24, 2012, 04:01:10 pm »
Ignore

I do think that the Dems were slightly over sampled. That is offset though by seniors also being slightly over sampled. I suspect Florida will be close enough for another controversial recount However it won't matter much because Obama will have Ohio and Virginia won on election night and should have won reelection by then
Logged




Bernie would probably win Vermont if Obama were deemed to have more than 272 evs in the vag.
M
YaBB God
*****
Posts: 2496


View Profile
« Reply #13 on: May 24, 2012, 09:16:57 pm »
Ignore

Bawlexus, not everything is wrong if the party sample doesn't match the 2008 one. Stop reading on party ID, please, it's becoming annoying. Are you umengus or what??

 

JulioMadrid, this is a poll forum where people discuss polling. I do not imply that party samples disqualify any polls, just that they are somehing worth considering. Just pointing out the data. Don't really care whether I annoy you or not, and stop telling me what to do. Please just ignore my posts (as I'll kindly do yours).

OK, you can point out the data, give your analysis or whatever you want to. But saying that a poll is accurate or not based on party ID all the time around (specially with PPP) IS annoying.
Read some umengus posts, then think about them. Don't you think he's a hack, or that his posts are irritating? Well, I'm sorry to write this, but you're more or less the same.

But, of course, you decide what you post. I have nothing against that. But people here aren't dumb. They (we) know when a poll is trash.

How do you know when a poll is "trash" or "not accurate"? When it doesn't provide the results that you anticipate? Or when the methodology is flawed?

You have to break down the numbers to get at the methodology. That's what the pros do, as you know if you read sites like RCP or Politico. The Vorlon did this for the forum in the 2004 and 2008 cycles and he was one of the most respected members for doing it.

Now, I acknowledge ajb's point that these may be accurate samples of registered voters. That is not, however, a figure that historically correlates in any significant way to the vote on election day. Any weighted likely voter poll, especially this far out, involves some guesswork and extrapolation; some will be off by several points. However, what polls like this are attempting to measure is based on an electorate +D of an election in which Obama won by more than 7 points!

To give a more concrete example of the problem, earlier today on a different thread, user cope1989 asked: "So can someone please explain to me why this election will look like 2004 when all of the polls look like 2008?"

Well, bawlexus91 is attempting to do just that. The reason is that with the major exceptions of Rasmussen and Gallup, these polls are not yet even trying to approximate the actual vote of these states! Largely because the young and lower income voters consistently have lower turnout, the RV number has ALWAYS in modern times been +D of election day. To even get a hint at what these polls mean, you have to weight them yourself.

What bawlexus is applying is political science. He is not the member dragging down the level of discourse on this subforum.
« Last Edit: May 24, 2012, 09:39:57 pm by M »Logged

Recently moved to Jackson, Mississippi.
Smash255
YaBB God
*****
Posts: 14275


View Profile
« Reply #14 on: May 24, 2012, 09:27:55 pm »
Ignore

Absolutely there's value to polls of registered voters. I have no problem at all with them. But that doesn't preclude anyone from comparing a registered voter poll sample to actual turnout in past elections, just because it isn't of likely voters. We compare/analyze what we're given.

I do think the poll oversamples Democrats, but not as much as you might think.

One thing to keep in mind is the difference between asking for Party ID and asking what party you are actually registered with 

The 2004 and 2008 questions were strictly voter ID, this one was party registration  Someone could be a registered Democrat, but identify more with being an Independent or even a Republican, and someone could be technically a registered Republican, but identify as an Independent or Democrat 

You probably do see a fair amount of registered Democrats in the Panhandle of Florida, that might technically be registered as a Democrat, and may show up that way with a registered question, but if it was a party id question might not.
Logged

M
YaBB God
*****
Posts: 2496


View Profile
« Reply #15 on: May 24, 2012, 09:56:32 pm »
Ignore

Bawlexus, not everything is wrong if the party sample doesn't match the 2008 one. Stop reading on party ID, please, it's becoming annoying. Are you umengus or what??

 

JulioMadrid, this is a poll forum where people discuss polling. I do not imply that party samples disqualify any polls, just that they are somehing worth considering. Just pointing out the data. Don't really care whether I annoy you or not, and stop telling me what to do. Please just ignore my posts (as I'll kindly do yours).

OK, you can point out the data, give your analysis or whatever you want to. But saying that a poll is accurate or not based on party ID all the time around (specially with PPP) IS annoying.
Read some umengus posts, then think about them. Don't you think he's a hack, or that his posts are irritating? Well, I'm sorry to write this, but you're more or less the same.

But, of course, you decide what you post. I have nothing against that. But people here aren't dumb. They (we) know when a poll is trash.

How do you know when a poll is "trash" or "not accurate"? When it doesn't provide the results that you anticipate? Or when the methodology is flawed?

You have to break down the numbers to get at the methodology. That's what the pros do, as you know if you read sites like RCP or Politico. The Vorlon did this for the forum in the 2004 and 2008 cycles and he was one of the most respected members for doing it.

Now, I acknowledge ajb's point that these may be accurate samples of registered voters. That is not, however, a figure that historically correlates in any significant way to the vote on election day. Any weighted likely voter poll, especially this far out, involves some guesswork and extrapolation; some will be off by several points. However, what polls like this are attempting to measure is based on an electorate +D of an election in which Obama won by more than 7 points!

To give a more concrete example of the problem, earlier today, user cope1989 asked on a different thread: "So can someone please explain to me why this election will look like 2004 when all of the polls look like 2008?"

Well, bawlexus91 is attempting to do just that. The reason is that with the major exceptions of Rasmussen and Gallup, these polls are not yet even trying to approximate the actual vote of these states! Largely because the young and lower income voters consistently have lower turnout, the RV number has ALWAYS in modern times been +D of election day. To even get a hint at what these polls mean, you have to weight them yourself.

What bawlexus is applying is political science. He is not the member dragging down the level of discourse on this subforum.


Wow. I thank you very seriously for that. Defending me better than I'm able myself.

No problem. Keep up the good work.
Logged

Recently moved to Jackson, Mississippi.
Pages: [1] Print 
« previous next »
Jump to:  


Login with username, password and session length

Logout

Powered by SMF 1.1.19 | SMF © 2013, Simple Machines