Warning

 

Close

Confirm Action

Are you sure you wish to do this?

Confirm Cancel
Member Login
Posted: 10/26/2004 2:42:34 AM EST
I don't pay much attention to the polls, but something caught my attention and raised a question.

In this lastest ABC? poll that shows Kerry ahead, it sounded like they said the actual polling numbers were "adjusted" for:

1. People's geneal opinion of the direction of the country

2. The fact that terrorism is NOT the major concern anymore

3. something else - I can't recall.

Here's my question -

In these polling numbers, do they take the ACTUAL poll results, and then INTENTIONALLY adjust (read: skew) them with other factors like the ones mentioned above??



Link Posted: 10/26/2004 2:49:29 AM EST
Why...that would be dishonest!


Sgatr15
Link Posted: 10/26/2004 2:53:02 AM EST
Polls are never accurate, they reflect the consensus of a group that is normaly found at that site/news station.
A good example would be to post a poll of ARFCOMers that support the second amendment.
Post the same poll at both Fox News, and CNN and see how the numbers change.
Link Posted: 10/26/2004 3:08:01 AM EST

Originally Posted By Gunslinger808:
Polls are never accurate, they reflect the consensus of a group that is normaly found at that site/news station.
A good example would be to post a poll of ARFCOMers that support the second amendment.
Post the same poll at both Fox News, and CNN and see how the numbers change.



My understanding is these are random phone polls - i.e. somewhat more representative than asking a HCI convention if guns are bad.



Link Posted: 10/26/2004 4:01:43 AM EST
Good polls will release all their data. I don't know if this ABC poll did that. Most or all of the polls "weight" their results by likely voter and party affiliation, but both of those weights can introduce errors. I suspect this could be part of the Wash Post's story line that "Kerry is a good closer". It's actually "Kerry is a hoser".

GunLvr
Link Posted: 10/26/2004 4:03:25 AM EST

Originally Posted By GunLvrPHD:
Good polls will release all their data. I don't know if this ABC poll did that. Most or all of the polls "weight" their results by likely voter and party affiliation, but both of those weights can introduce errors. I suspect this could be part of the Wash Post's story line that "Kerry is a good closer". It's actually "Kerry is a hoser".

GunLvr



"Eh, hoser....."

I think that's what I'm getting at.

Introducing even a few statistical adjustments pretty much turns polls into crap shoots at best, and ideologically based shilling at worst.

Link Posted: 10/26/2004 4:08:44 AM EST
[Last Edit: 10/26/2004 4:09:15 AM EST by PAEBR332]

Originally Posted By garandman:

Originally Posted By GunLvrPHD:
Good polls will release all their data. I don't know if this ABC poll did that. Most or all of the polls "weight" their results by likely voter and party affiliation, but both of those weights can introduce errors. I suspect this could be part of the Wash Post's story line that "Kerry is a good closer". It's actually "Kerry is a hoser".

GunLvr



"Eh, hoser....."

I think that's what I'm getting at.

Introducing even a few statistical adjustments pretty much turns polls into crap shoots at best, and ideologically based shilling at worst.




Introducing statistical adjustments can actually make the poll much more accurate, IF the adjustments are done for valid reasons. These adjustments should only be used to remove know sample bias, i.e. getting a sample that is very unrepresentative of the population (like one containing 80% Republicans).

That is why you need to know what the raw numbers were, as well as what adjustments were made and why. If the poll will not divlulge this information, be VERY wary of the results.
Link Posted: 10/26/2004 5:07:39 AM EST
Link Posted: 10/26/2004 5:12:03 AM EST

Originally Posted By sgtar15:
Why...that would be dishonest!


Sgatr15



Are you a dyslexic typist, Sarge?
Link Posted: 10/26/2004 6:14:46 AM EST

Originally Posted By HiramRanger:
Yup, without the methodology and internals I can't judge whether they skewed the results to a desired end or not. It is possible to adjust numbers both to bias the poll and remove bias... but without access to the core data it is impossible to know.



OK.

Nother ? -

Is that type of "adjustment" standard procedure in polling, or are we usually just receiving raw data ? (i.e. 540 voted for Kerry 570 voted for Bush)
Link Posted: 10/26/2004 6:27:31 AM EST
Link Posted: 10/26/2004 6:28:26 AM EST
My feeling is that without seeing the poller's actual methodolog, I would go with the number of different polls and who they favor. Most of the polls in my opinion don't same a big enough sample size to make an accurate inference on the entire population. Polls such as the LA Times to me has a history of being wrong because of their political bias for the Dems.

Check out HiramRanger's Electoral Vote Map October 26.
Link Posted: 10/26/2004 6:42:23 AM EST
I said this earlier as far as how much I am relying on the polls:

In the past 100 years, the electorate consisted of about 35%-40% of the total people eligible to vote. They are now estimating that it can be upwards of 60% of people eligible to vote who will be casting ballots this year. No pollster or political pundit knows how to even read this. Could it be that the electorate is desperate for change? Can it be because the electorate wants to ensure that we maintain consistent leadership in a time of war? Who knows but something is motivating the populace.

Unless there is some drastic and explainable shift in the poll numbers, I will patiently wait 8 more days

Link Posted: 10/26/2004 6:43:55 AM EST

Originally Posted By garandman:
Is that type of "adjustment" standard procedure in polling, or are we usually just receiving raw data ? (i.e. 540 voted for Kerry 570 voted for Bush)



I don't know whether typical political polls in a non-election year use raw data or not.

Most of the studies the government (or academics) release which involve surveys include weighting of the data. My PhD dissertation weighted the data of the survey I did and I did not choose the weights, it was done by by survey "experts" and out of my hands (although I could dump the weighting factor if I chose).

GunLvr
Link Posted: 10/26/2004 6:44:27 AM EST
Link Posted: 10/26/2004 6:49:43 AM EST
A quick look-see at -Absolut-'s link shows that the majority of the polls has Bush ahead.
Top Top