<<this US Election is eating into my sleeping time!!!>> More response later, but for now it's giving me absolute nightmares. A lot of that going around. Eternal question -- enough for the electoral college? Julie Krueger ========Original Message======== Subj: [lit-ideas] Re: polling Date: 10/10/04 8:00:35 PM Central Daylight Time From: _judithevans001@xxxxxxxxxxxxxxx (mailto:judithevans001@xxxxxxxxxxxxxx) To: _lit-ideas@xxxxxxxxxxxxxx (mailto:lit-ideas@xxxxxxxxxxxxx) Sent on: Monday, October 11, 2004, 12:07:50 AM, JimKandJulieB@xxxxxxx wrote: Jac> Oh dear -- I'm probably going to be banished from the list for asking such a Jac> stupid question but what is the technical diff between a survey & a poll? Jac> To me the words always generally meant asking a bunch of people their opinion Jac> and making a tally. There isn't a difference usually -- you're right -- CNN/Gallup call this a survey but perhaps they shouldn't. They've based their Electoral College figures on polls plus interviews with others they mention. I can't see anything wrong with doing that, so long as they make it clear their claims are based on it. Je> Zogby International's polls are the ones whose method differs Je> from others, in just one, possibly fairly crucial, respect. I Je> can't decide whether they're more likely to be right than the Je> others -- I'f say if the apparently very small % of undecided Je> voters we've been cited from the beginning masks reasonably Je> large shifts (and I'd now say it does) they could be. Jac> Can you explain what that "possibly fairly crucial respect" is? I thought I could but now suspect (having just read a piece by John Zogby) that the account I read -- of the difference in methods -- was wrong. Here's what seems to be the real difference and the possible reason why Zogby got the result right in the last 2 elections. First, here's the reason why he chose a different method! Partisans tend to be overrepresented in "likely voter" polls, and further (for example, but an important one) a particular group of partisans can be overrepresented in a given "likely voter" poll. So -- it seems -- Republicans were overrepresented in such a Gallup poll following the Republican Convention (one that gave Bush an 11 point lead; FYI the Gallup poll of registered voters apparently showed a far smaller lead). It's also true, I now learn, that their regular registered voter sample has a Republican lead; not in terms of how people intend to vote this time, but in terms of "Party Identification", which is -- supposedly -- a measure of more enduring support. That would tend to an overestimation of support for the Republicans. They aren't the only pollsters to overrepresent "Republican Identifiers" (not, I add, that they do it deliberately; various reasons have been suggested for it). What Zogby has done (as have some others) to try to avoid this, is (I now read) to weight his samples so they reflect the expected proportions of Democrats, Republicans and Independents in the elctorate (given the known "Party Identification" data, just as others weight their samples to reflect the numbers of men and women (as does he). The account I read got this the wrong way round, and I see that I agree more with Gallup etc. than with Zogby re weighting by PI. But if you want polls whose method has the best track record of late, go with Zogby, while bearing in mind that the US is now more polarized than before -- but figures for "Party Identification" don't take this into account -- and that turnout is expected to rise this time. Aaaaarrrrgggghhhh, who would be a pollster? Or, indeed, a political scientist?! I can get you some more stuff on all of this but not right now -- this US Election is eating into my sleeping time!!! -- Judy Evans, Cardiff, UK mailto:judithevans001@xxxxxxxxxxxxxx ------------------------------------------------------------------ To change your Lit-Ideas settings (subscribe/unsub, vacation on/off, digest on/off), visit www.andreas.com/faq-lit-ideas.html ------------------------------------------------------------------ To change your Lit-Ideas settings (subscribe/unsub, vacation on/off, digest on/off), visit www.andreas.com/faq-lit-ideas.html