Another new state poll (Nevada)...

[millwx]

Survey USA poll...

Bush: 45%
Kerry: 49%

...haven't looked at the details yet.

...incidentally, on the poll averages I posted on the "New state polls" thread, the SUSA poll would now make Nevada dead even at 45.7% Bush to 45.7% Kerry.  In a day or so an old poll will drop from Nevada.  Barring a new one showing Bush ahead, Kerry will move out to a slight lead in NV.

[millwx]

Hmm. . .  They oversampled Las Vegas by 10% and undersampled both parties by about 5% in the registered voters, thus oversampling independents by 10% (rounding making the discrepency.)

Maybe, maybe not.

1) Since they also undersampled Dems... undersampling both parties... there is zero indication of the "summer Republican fade".

2) Moreover, w.r.t. party ID, this is often a self-identifier, which can sway with the match-up numbers.  Not a reliable piece of data.  I only balk at it when it is severely out of whack... like in the old LA Times poll.  And it's why some people don't like to weight by party ID (even our resident expert, The Vorlon, doesn't like party ID weighting).

3) It's impossible to say if and by how much they oversampled the Las Vegas area, because we don't know what the "Las Vegas area" means.  For example...

-If it's Las Vegas proper (not likely, since they specifically say "area"), that's less than 25% of the state; therefore, SUSA way oversampled the area.

-If it means all of Clark County (which is highly plausible), that's about 60% of Nevada, meaning they slightly oversampled the Las Vegas area.

-If it means all of Clark, Nye and Lincoln Counties (an awful large region, but the largest towns/cities in Nye and Lincoln are rather close to Clark and are far outer suburbs to Las Vegas... so, this is also feasible), they total very near 2/3rds the population of Nevada; in that case the "Las Vegas area" was still oversampled, but barely... and MAYBE not at all, since...

4) We don't see their "adults" breakdown (which is what's important, because we need to make sure their RESPONDENTS are spread properly... not their registered or likely voters; registered and even more so likely voters can be impacted by the specifics of the election).  I mention this because the "registered" breakdown is less Las Vegas oversampled than the "likely" breakdown.  That makes me strongly suspect that the "adults" breakdown - if "Las Vegas area" includes these three counties - is not oversampled at all.

[millwx]

1- The Republican fade is largely from the independents area of a poll.  Oversampling them will exaggerate the fade effect.

2- S-USA uses either a self identifier (do you consider yourself a . . . ) or asks what party they are registered (Are you regestired in . .  .) depending on client requests.  We're not sure which method wa sused, but S-USA prefers self identifiers.  I know the local NC polls done by them use the second method since the local stations want that.

3- Clark county has 62% of Nevada's registered voters.  If we include the bordering counties it goes up to 64%.  The poll has 72% from Clark County.  

4- Registered voters are what needs to be representative, since that is what we have data on.  We can check with the state to see how many people are registered in each party and get various breakdowns of where they are.

People who are not registered to vote are irrelevant to a poll as they cannot vote in the only poll that matters.  To survey them is a waste of time and money.

1) Makes no sense at all, since, by your own judgement, there is also a Democratic fade in the poll.

2) You're right, we don't know.  But, as you point out, if SUSA had the option, it's a self-identifier.  So, it's relatively meaningless anyway.  This also makes the entire discussion in point #1 moot.

3) Based on 2003 Census Bureau estimates, and extrapolating to 2004, Clark Co is 72% of the state.  Clark plus Lincoln and Nye is over 75%.  Now, I presume you're not looking at population, but at the RV table.  In which case, you're right, Las Vegas is oversampled.  I would conjecture, however, that many folks answered "registered" who weren't, rather than a polling bias/error/etc by SUSA.  Why?  Given the massive population shift in NV (Clark is now 10% more of the state population than they were in 2000), many Las Vegas residents may not have even been eligible to vote until very recently.  So, there may be a massive number of these people who PLAN on registering but are not presently, so they responded that they are registered.  Other than a bizarre 21,000 drop in RVs in April (cleaning the voter registration rolls??), Clark has increased by an average of 10,000 RVs/mo this year.  That should only accelerate.  So, accumulating another 60,000-80,000 by the end of October is highly feasible, if not outright likely.  At 80,000 Clark Co would close in on their 72%.  Of course, other counties will also see registration increases, so even with 80,000 Clark might come in under 72% of the state's RVs, but they'll be close; and with Lincoln and Nye included, it could hit that.  I will certainly concede that you're right that there may be some oversampling in the Las Vegas area.  For one thing, these folks AREN'T registered now.  For another, did that many lie simply because they intend to register?  I doubt it.  And, lastly, even if a whopping 80,000 Clark voters register (possible, but probably on the high end of what's possible) Clark will still fall short of its 72% representation in this poll.

4) The RV poll distribution should NOT match the population distribution.  However, if you're comparing to actual RV distribution (which it was not clear that you were... until you posted the link to NV RV data), then, yes, the two distributions should match.

Anyway, in short, I think the truth lies somewhere in between.  I think you're right that Las Vegas was oversampled.  On the other hand, amongst RVs statewide the Las Vegas area is massively underrepresented.  As more voters register (which they will, being a presidential election year), the gap will close.

[millwx]

So, the idea that Clark county has 72% of the state voters seems more than a little far-fetched to me.

Carl, nobody has said this.  The only thing stated (by me) is that Clark County has 72% of the POPULATION.  This is supported by the U.S. Census Bureau estimates for 2003, extrapolated to 2004 (it was 70% in 2003).  The only time I "called out" Tredrick is because I though (it was unclear) he was talking about population distribution (which is nearly 72% in Clark County).  Note that I corrected my statement in complete agreement with him in my followup post.  And note that my original response to Tredrick clear states that I'm referring to population.  Other than that, no one has supported SUSA's 72% claim.  Methinks you also need to get your facts straight... or at least your accusations.

[millwx]

Then you do agree that voters vote and not population.

Please cite a 'fact' which I misstated.

In my math (perhaps a little old fashioned), the extrapolation of trends favors Bush.

No, you posted no incorrect facts either.  What you posted was an incorrect accusation.  I can only assume you were referring to me, since I mentioned the 72% population in Clark County.  My concern with what Tredrick said, and it was entirely a misunderstanding on what he stated (as I pointed out in my followup), was that I believed he, too, was looking at population numbers.  He wasn't.  He was looking at RVs.  That, obviously, is the correct thing to look at... but, at that point, I hadn't, I was merely responding to the population (as I believed the issue to be) distribution question.

As for the trends favoring Bush... maybe.  Why only "maybe"?  Your statistics are entirely correct.  But mine are as well... Clark County has seen an average of 10,000 new registrants/month, excluding the blip in April.  I'm concerned about my exclusion of the April data, in case there's something real there, but the decrease is steep and isolated (only in April).  So, I'm am presuming it was a cleaning of the voter rolls.  Also, the registration trend in Clark County favors the Dems (I continually hear that it's Republicans migrating to Clark County from California).  Excluding April, Rep registration in Clark is up 11,700, Dem registration in Clark is up 19,200.  Also, in January Reps had a 13,000 RV edge; it is now down to 9,000 (and that does not exlude April).  The Dems have closed the gap by 4,000.  Moreover, the Dems were more impacted by the bizarre April.  Excluding April statewide the Dems have gone up 33,000; the Reps have gone up only 16,000... a difference of 17,000.  I just wish I knew what was up in April, so I knew whether or not it's "real"!  So, it is all in what data you decide to put up here.  One can easily argue that the trend favors Kerry.  Clark has the most room for registering voters (largest county, very low percentage of population registered), and it is presently already  impacting the statewide RV numbers.  If the registration rate increases (likely) Dems could easily outnumber Reps by election day.  And this is important because, while you did show the June 2000 numbers, you failed to point out that by the end of October 2000 Republicans had the RV edge (albeit slight... by a mere 1,000).  But this year the RV trend is in the opposite direction

Point is, it's all hand waving.  You cherrypicked a few pieces of data.  Anyone else can do the same thing in the other direction.  I've just done so above.  It's impossible to say who the trends favor.