pudge's Journal: Stupid Polls

Check out this CNN/USA Today/Gallup poll from October 1998, a full two years before the 2000 election, and well before Bush or Gore announced they were running for President.

Note how in the questioning, they only match two candidates from opposing parties head-to-head: Bush and Gore. Note that this is the first question.

And then wonder to yourself, would this influence the later questions, about who you would vote for in the respective parties?

Duh. The numbers for Gore vs. rest of field, and Bush vs. rest of field, are invalid, because the previous question already told the respondents who the pollster considers the most likely candidates to be, and many people are more inclined to vote for who they perceive to be the winner.

This is one of the many reasons you should treat poll results with skepticism, especially if you can't see the actual questions asked, and their order.

rest of world

[nocomment]

Where's the "rest of the world" polls from 1996? oh wait, no one cared back then?

Re:rest of world

[robi2106]

did I miss the part where we are supposed to care about who the rest of hte world would vote for now??? Hummmm I guess I didn't get that memo.

jason

Re:rest of world

[nocomment]

that's my point. We shouldn't care. But it's been all over the place recently. I think there was even a couple different writeups on /. about who the rest of the wrold wants. I quite frankly don't care who they want.

More bad polls

[helix400]

http://www.electoral-vote.com [electoral-vote.com] lately has been railing on polls as well.

Two recent polls I've run into have really irked me. One was an automated telephone poll asking which local state representative I'd vote for. It only gave me two options, and no undecided or third party button to press. This is a major flaw in polling, since they will be unsure what proportion actually are unsure. They could get 500 responses back suggesting a Republican would win, but maybe 2500 undecided hangups who will event

Re:More bad polls

[cascadefx]

I hadn't heard about Richie Robb's [google.com] refusal to cast his Republican electoral vote for Bush until I read this on Electoral Vote'w site [electoral-vote.com] (been out of the news loop lately). Great site and interesting analysis [electoral-vote.com] of a potential tie.

Re:More bad polls

[ces]

"With Salt Lake County Mayor Nancy Workman seeking re-election while facing two felony charges -- for allegedly misusing public monies -- who would you vote for if the elections were held today?"

You know if you got that question in a telephone poll I'd think it was one of those infamous push-polls.

Head of Gallup interview

[cascadefx]

NPR had an interesting segment on polling the other day. In it, they interviewed Frank Newport, the head of Gallup. I think it served more to question poll results, in my mind, and brought up similar issues for me that you have mentioned Pudge.

Check it out. [npr.org]

Re:Head of Gallup interview

[ces]

I heard that too.

I take everything he said with a rather large grain of salt considering that Gallup is one of the organizations accused of using flawed methodology.

They adjust results for party ID and their party ID numbers are way out of whack compared to exit polls from 2000 or the annual Pew survey of party ID. (party ID doesn't tend to change radically year to year)

Indeed if you readjust their numbers according to more commonly accepted party ID numbers you end up with results closer to Zogby, ARG,

Ahh

[buffer-overflowed]

push polling.

Who seriously trusts polls anyway? Sure we all use them to try to bolster bad arguments, but does anyone really trust them?

This is something, in combination with the whole new types of fraud brought in by Electronic voting systems, that kinda lurks outside my consciousness.

It's something I worry about from time to time, as it really is getting easier to subvert the whole process, not harder. And it's getting easier not to get caught.