Net Polls

Doug Henwood dhenwood at panix.com
Mon Oct 2 05:15:35 PDT 2000


Gar Lipow wrote:


>Doug -- I'm sending this to you personally, because the panix server is
>refusing to relay this to the e-mail list.

It did.


>Doug Henwood wrote:
>>>Gar Lipow wrote:
>
>>>When I took the quiz it told me two things:
>>>1) My candidate is David M.
>>>) I'm a moderate liberal.
>>>
>>>Uh yeah, right.
>
>>So which is inaccurate?
>
>>Doug
>
>Doug I know I have not posted that often, but do I sound like a moderate
>liberal?

No. I was curious about the David McR part, really.


> >PS: The AAPOR crowd seems mystified by the fact that while net polls
>shouldn't be accurate, they're more accurate than they should be.
>
>If net polls are really coming closer to reality than they should be
>(I'd want more evidence) could it be a matter of sheer size of the
>sample? That is, once you get into gigantic smaple sizes, does the
>ratio of rerpesenative to non-representative possible sample change. It
>has been a long time since my elementary statistic class. I know that
>you don't gain that much when it comes to truly random samples, but I
>don't think we ever dealt with how it would affect biased samples.
>Anyone with the statistical savvy to answer this one?

I'm no stat-head, but a large sample size isn't going to improve things if it's an unrepresentative sample. You'd think that a net audience is richer, whiter, more male, and more libertarian than a random sample of the U.S. pop, wouldn't you?

Doug



More information about the lbo-talk mailing list