Could "Sex" Be Why Polls Are All Over the Place???

[ItsMyParty]

I was polled by Gallup within the last year. They asked 3 questions up front (and explained to me the one that seemed 'odd').

They asked how many people in my household were over 18 and registered voters (or some such question). I said 2. They asked if I was one of the 2. I said yes. Then, they asked me if I was the one of the 2 who has had the 'latest' birthday. I said that I was. They explained to me that then they could continue with the polling otherwise they would have had to end it because they had to talk to the person with the latest birthday in order to make their sample random. It was way down the list after they determined by those 3 questions that I could be interviewed, that they asked my 'sex'.

Since 'sex' wasn't one of the determning factors for being polled, it could very easily be that they could interview 1,000 people for one of their polls and end up with things like 75% of them being of one sex and 25% the other. In other words, if they happen to interview a lot of men using the above criteria, I'm sure you're going to get an 11 point spread for Chimpy. If you get a mix, I'm sure you're going to get a much closer race. And if you get a lot of women, I'm sure you are going to have a Kerry lead. In fact since race and income were way down on the list, too, think of what would happen if they got into a real run of white, fairly well off, males. Could this be why we have poll numbers all the hell all over the place????

[cheezus]

1. in a large enough random sample

there WILL be 50/50 males females

I mean, if the poll can't even get that right about the general population, how can we expect that the data on political preference is even close to valid?

[TomNickell]

5. Unless it's NOT random--Sampling Bias Vs Sampling error...

If the sample is not random--if there are more women answering the phone that night, for example--making more calls won't help.

It gets the Sampling Error down, but the Bias is still there.

They can call phone numbers randomly, but the actual sample -has- to be influenced by the TV schedule, holidays, time-of-day etc. During the RNC, they are going to get more Repubs home watching Dick and George. During the Holiday Weekend, there may be fewer, or more, people home.

[seabeyond]

2. they really should use gender and not sex

:)

[tedoll78]

3. This is why I like weighted polls better.

I'm noticing that the weighted polls are producing one group of similar results, while the non-weighted polls are producing another group of similar results. And it's of great comfort that the weighted polls (esp Zogby) were more accurate in the last presidential race.

[tinanator]

4. sample size matters

the ludicrous poll results these fascists produce are skewed by their sampling methods, and betrayed by their paucity. They are setting up another theft, and you best get ready to fight back with exit polling. Unless someone has a better idea?

[Warpy]

6. A poller called me last night

and said he needed to speak to the youngest male in the household. When I said that would be my cat, he lost interest in asking any questions, thank you and good night.

You bet they're manipulating the sample.

[Medium Baby Jesus]

7. No Sex Threads!

Sample size matters indeed.