PDA

View Full Version : Polling Is "Not Simply A Sampling Of People's Preferences"


RINGLEADER
10-18-2004, 11:09 PM
Sounds to me like John Zogby is admitting that he pretty much takes raw data and then uses his own guesses to determine the results of his polls. Weighting a poll for partisan response is one thing...weighting it based on culture and sociology seems like he's stretching:

"Polling can be uncertain only if you rely on statistics alone," is Mr Zogby's riposte. "That's why an effective pollster has to rely on culture, history and sociology. I repeat, polling is the study of human behaviour, not simply a sampling of people's preferences."

|Zach|
10-18-2004, 11:24 PM
Polling in and of itself is ugly...

Do you get a random amount of people for your poll? If the poll is random then is that fair? The odds of matching up the demographics that represent your population from random sampling is small...

How do you do it? What questions do you ask? Seems like there is a lot of loaded\leading questions out there...

Does the act of polling itself but a slant on before any polling has begun? Seems to me there are ceratain people have personalities that make them never respond to polls. I wonder if these people that have a tendancy not to respond have a tendancy to go a certain way in whatever poll is given?

|Zach|
10-18-2004, 11:28 PM
Zogby seems to have been as accurate of poll as any...

2bikemike
10-18-2004, 11:32 PM
Polling in and of itself is ugly...



Rainman is not going to like hearing this!

|Zach|
10-18-2004, 11:33 PM
Rainman is not going to like hearing this!
I believe I would get my ass kicked by Kevin for such words. ROFL

Mr. Kotter
10-19-2004, 06:44 AM
Wow.

I understand his point, but that is a very murky journey he's embarked on then... :rolleyes:

Mr. Kotter
10-19-2004, 06:48 AM
Zogby seems to have been as accurate of poll as any...

He was accurate in 2000. His "record," otherwise, isn't any worse or better than the other major polling organizations.

So the question is, did he get lucky? Or is he onto something?

I'd suggest the criteria should be SCIENTIFICALLY conducted polling; 9 out of 10 times, if done SCIENTIFICALLY the polls will be damn close.

Entering "historical" and "cultural" factors to try and adjust for "human behavior" would really be nothing more than a guessing game really.

Cochise
10-19-2004, 06:55 AM
We already get enough opinion presented as fact in the news media. Polls are supposed to be raw data IMO.

Mr. Kotter
10-19-2004, 07:35 AM
We already get enough opinion presented as fact in the news media. Polls are supposed to be raw data IMO.

Exactly. And much less verbage than my rant. :thumb:

|Zach|
10-19-2004, 09:14 AM
I think you folks are very naive when it comes to polls.

Mr. Kotter
10-19-2004, 09:39 AM
I think you folks are very naive when it comes to polls.

What does that mean? Please explain?

|Zach|
10-19-2004, 09:50 AM
What does that mean? Please explain?
The idea that polls are just spouting raw data. Polls have to be done by people and put together by people and given to the public by people. It is done not in a vacuum.

For polls to be accurate humans have to sort through data. If you don't like the idea of this then I would plug your ears and not listen to a single poll.

But like I said Zogby must be on to something. If you want to call them "lucky" for their work in 2000 then I dont know what to tell you...

Stinger
10-19-2004, 10:00 AM
I think you folks are very naive when it comes to polls.

WtF do you mean? Hang on let me check that opinion poll to see wether I agree or disagree with this post.

Mr. Kotter
10-19-2004, 10:13 AM
The idea that polls are just spouting raw data. Polls have to be done by people and put together by people and given to the public by people. It is done not in a vacuum.

For polls to be accurate humans have to sort through data. If you don't like the idea of this then I would plug your ears and not listen to a single poll.

But like I said Zogby must be on to something. If you want to call them "lucky" for their work in 2000 then I dont know what to tell you...

I understand your point, but would offer the following.

I've been voting since the 1984 election. We know more about Zogby's reliability after this election. I don't think ANYONE here is suggesting history and culture don't matter; but how the heck do you QUANTIFY that? Most reliable polling organizations rely on very rigorously SCIENTIFIC methods of polling; to introduce subjectivity in the polling data, it becomes more of a qualitative process and less of a quantitative process.

Speaking strictly from a analytical perspective, that fact should be troubling to anyone viewing data collected the way that Zogby is suggesting he is....

Polling is enough of a crap shoot as it is; this makes it even more so.

Mr. Kotter
10-19-2004, 11:00 AM
I understand your point, but would offer the following.

I've been voting since the 1984 election. We know more about Zogby's reliability after this election. I don't think ANYONE here is suggesting history and culture don't matter; but how the heck do you QUANTIFY that? Most reliable polling organizations rely on very rigorously SCIENTIFIC methods of polling; to introduce subjectivity in the polling data, it becomes more of a qualitative process and less of a quantitative process.

Speaking strictly from a analytical perspective, that fact should be troubling to anyone viewing data collected the way that Zogby is suggesting he is....

Polling is enough of a crap shoot as it is; this makes it even more so.

I guess YOU told them, SD. :thumb:

BigOlChiefsfan
10-19-2004, 11:00 AM
I've mentioned Sabato before, not exactly a poll, but not to be sneezed at, either...here's this weeks update.

http://www.centerforpolitics.org/crystalball/2004/president/electoratemap.php

Mr. Kotter
10-19-2004, 01:51 PM
I've mentioned Sabato before, not exactly a poll, but not to be sneezed at, either...here's this weeks update.

http://www.centerforpolitics.org/crystalball/2004/president/electoratemap.php


Sabato is good; but left of center....

Mr. Kotter
10-19-2004, 05:00 PM
I've mentioned Sabato before, not exactly a poll, but not to be sneezed at, either...here's this weeks update.

http://www.centerforpolitics.org/crystalball/2004/president/electoratemap.php

I also would check electoral-vote.com and federalreview.com for other perspectives on the electoral vote tally.

This site now has a state by state breakdown...interesting. :hmmm:

http://www.federalreview.com/

Rain Man
10-19-2004, 09:52 PM
Polling in and of itself is ugly...

Do you get a random amount of people for your poll? If the poll is random then is that fair? The odds of matching up the demographics that represent your population from random sampling is small...

How do you do it? What questions do you ask? Seems like there is a lot of loaded\leading questions out there...

Does the act of polling itself but a slant on before any polling has begun? Seems to me there are ceratain people have personalities that make them never respond to polls. I wonder if these people that have a tendancy not to respond have a tendancy to go a certain way in whatever poll is given?

When I do polls (in real life, not on Chiefsplanet), we develop weighting factors to correct for the fact that we don't get a representative sample of people in our random draw. It's not too hard to develop correction factors for that, at least within reason.

How the questions are asked is a big deal. We can really drive the results of a survey with subtle changes. The legit market research firms always work toward asking neutral questions, even if it's more of an art than a science to make that determination.

And yeah, there's probably a difference between people who agree to participate in a poll and those who don't. There are ways to calculate that error, but in real life almost no one goes to that expense, since it's a real headache.

That said, some types of polls ask about current opinions and others ask about planned future behavior. On current opinion polls, it's generally okay to just demographically weight each survey response to correct for the cooperation bias factor discussed above.

For predictive polls, you have to go a step further. If someone gives you their opinion that they would vote for Candidate X, that doesn't mean a thing if they've never voted and don't really intend to this time, either. If you want to predict outcomes, you then have to model two steps: what is the person's opinion, and how likely are they to follow up that opinion with action (e.g., casting an actual vote in the case of political polls)? I think it's perfectly legitimate to weight each survey response in a political poll based on that person's estimated likelihood of voting. It only makes sense, and that approach could certainly be applied to other factors as well. In these cases, I think the polling part of the process is just one input into developing a predictive model.