Bush up 7 in Florida, says SurveyUSA

[Fmr. Gov. NickG]

723 Florida LVs surveyed 6/12-6/14

Bush 50
Kerry 43

http://www.surveyusa.com/currentelectionpolls.html

[Lunar]

Ouch.

I'm waiting for Mason Dixon to do a new on there though.

[JNB]

   As with all polls, I would like to see a demographic breakout. My theory here is Bush is doing better among the Jewish vote than in 2000, possibly getting 30% rather than the 18% he got in FL last time.

[millwx]

723 Florida LVs surveyed 6/12-6/14

Bush 50
Kerry 43

One potential serious problem with this poll.  It's party affiliation was 41% Republican, 34% Democrat, 24% Independent/other.  It's registered voter sample was only slightly shifted (by about 2%) more to the Democrat side, still giving Republicans a four percentage edge.

This RV sample is important, even though the poll's results used only the LV pool, because it should match the state's registration closely.  It does NOT.  According to the Florida Department of State the official party registration numbers for Florida are: Democrat 42%, Republican 38%, None+Minor 20%.

So, the SurveyUSA RV sample favoring Reps by 4% is off by about 8%.  That's practically LA Times material (in the opposite direction, and I would highly doubt the SurveyUSA bias is intentional).  I'd wait for a poll with a more accurate sample than this before making any conclusions from this.

[millwx]

You are looking at the RAW data, it got weighted back in the reported results.

I'm certain you know how they do business better than I, but I can say that their description of their weighting methodology within the PDF file with the results does not mention weighting by party affiliation.  There are other weighting factors indicated there, which may or may not catch the party affiliation "issue" in their raw data.

[millwx]

Look at page 4 of the .pdf, that is where they apply their party affiliation indexing.

Yeah, I saw that indexing, but it wasn't clear to me what that "meant".  It didn't look like it was "weighting".  Using their raw likely voter party registration (41/34/24) and applying it to the percentage in that "index" breakdown gives you their final results (e.g. 50% for Bush).  So, I'm still missing where they wieght back to the actual party affiliation breakdown.  Maybe I'm just stupid :-)  ...but, seriously, I'm missing it.  I don't see it, and unless the other weighting compensates (which it might), I'm hesitant to trust this poll.

[The Vorlon]

Look at page 4 of the .pdf, that is where they apply their party affiliation indexing.

Yeah, I saw that indexing, but it wasn't clear to me what that "meant".  It didn't look like it was "weighting".  Using their raw likely voter party registration (41/34/24) and applying it to the percentage in that "index" breakdown gives you their final results (e.g. 50% for Bush).  So, I'm still missing where they wieght back to the actual party affiliation breakdown.  Maybe I'm just stupid :-)  ...but, seriously, I'm missing it.  I don't see it, and unless the other weighting compensates (which it might), I'm hesitant to trust this poll.

I never trust ANY single poll

Like I said - it's one poll, one more dot on the graph.

I will go over this SUSA in some detail later, and post in Dave's polling section.

They typically do a pretty decent job of weighting things back however.  The raw sample in their last PA poll was a mess, but they brought it back pretty well actually.

The other thing to consider too is that it was taken right after Reagan died.

These are SELF IDENTIFIED party affiliations, I suspect a few "independants" and maybe even a Democrat or two "self identified" as a Republican in this survey.

[millwx]

They typically do a pretty decent job of weighting things back however.  The raw sample in their last PA poll was a mess, but they brought it bacl pretty well actually.

The other thing to consider too is that it was taken right after Reagan died.

These are SELF IDENTIFIED party affiliations, I suspect a few "independants" and maybe even a Democrat or two "self identified" as a Republican in this survey.

Actually, I tend to like SUSA as well.  The do seem to do a decent job.  This is the only one I've bothered to look at so closely, since it was a wee bit out of whack with other recent FL polls.  That's when this stood out to me.

And your "self identified" point is an excellent one.  I would self-identify as either independent or Republican... yet, I'm registered Democrat!!  How/why?  The Rep party has changed much since the late 60s.  I'm more of an old-fashioned Rep (socially moderate, fiscally responsible), but don't find the current Rep party to embody this.  Neither does the Dem party, but I live in a state where Indys can't vote in the primary, and I choose to have an impact.  So, I switched my official affiliation to Dem.  So, I, personally, am a clear indication of what you're talking about!  :-)

[bullmoose88]

They typically do a pretty decent job of weighting things back however.  The raw sample in their last PA poll was a mess, but they brought it bacl pretty well actually.

The other thing to consider too is that it was taken right after Reagan died.

These are SELF IDENTIFIED party affiliations, I suspect a few "independants" and maybe even a Democrat or two "self identified" as a Republican in this survey.

Actually, I tend to like SUSA as well.  The do seem to do a decent job.  This is the only one I've bothered to look at so closely, since it was a wee bit out of whack with other recent FL polls.  That's when this stood out to me.

And your "self identified" point is an excellent one.  I would self-identify as either independent or Republican... yet, I'm registered Democrat!!  How/why?  The Rep party has changed much since the late 60s.  I'm more of an old-fashioned Rep (socially moderate, fiscally responsible), but don't find the current Rep party to embody this.  Neither does the Dem party, but I live in a state where Indys can't vote in the primary, and I choose to have an impact.  So, I switched my official affiliation to Dem.  So, I, personally, am a clear indication of what you're talking about!  :-)

Please...come back...we need your types


please?

[millwx]

Please...come back...we need your types

Couldn't help but laugh out loud when I read this!  :-)  I'll return when the Reps learn to stop being spend-thirsty theocrats (there was a time when Republicans actually led the civil rights movement... and, no, I'm not talking about having all kinds of entitlement programs... I'm against them, one of few things I still like about the Rep party).  Does anyone notice that the Libertarian Party only gained recognition (albeit slight) when the Reps got "taken over" by the far right?  There's a reason.  Republicans no longer stand for personal freedom.  We/they used to have a saying... "out of your pocketbook, out of your bedroom"... the Reps have abandoned that second premise.  I stuck with the Reps for many, MANY years, hoping they'd go back to their Lincolnian roots.  It's not to be.  And, frankly, I get deeply offended when the far right Republicans invoke the name of Lincoln to try to make themselves sound more acceptable.

Sorry for the tirade, especially on this FL poll thread, but since you brought it up  :-)  Bring the William Cohens, Olympia Snowes, William Welds, Connie Morellas, Rudolph Giulianis, etc to the fore of the party, instead of the Hasterts, Delays, Lotts, Thurmonds, etc and I'll come back.  Let me state that the policies of the "right" have, in fact, personally impacted me in a negative way.  So, I am not spouting some idealistic or partisan rhetoric.  I am deeply and personally affected and offended by these types.  I can no longer support the Rep party until they return to their roots.  And it saddens me, because I certainly don't agree with the Dem party platform either.

...and I should add, based on your "tag line", bullmoose... I'm probably very much in agreement with you on the issues!

[Lunar]

In reality that 37% "for" Kerry is much lower.

It's an election with an incumbant, it's to be expected.

[The Vorlon]

This RV sample is important, even though the poll's results used only the LV pool, because it should match the state's registration closely.  It does NOT.  According to the Florida Department of State the official party registration numbers for Florida are: Democrat 42%, Republican 38%, None+Minor 20%.

Weighting a poll to party ID is something that very smart people can and do disagree on.

My own OPINION is that it is something you should NOT do,

Gallup, Harris, ABC, and TIPP, etc agree you should NOT

John Zogby, Ed Goaes, Scott Rasmussen and Celinda Lake disagree and say you should.

(The quality of the poll has FAR FAR more to do with factors totally unrealated to party ID weighting anyway, so in the end it is a fairly minor issue actually)

I will give you two examples, one historic, and one current why I think weighting by party ID is bad.

Back in 1994 during the congressional elections, a lot of polls started to weight by party ID.  In 1992 Ed Goaes absolutely NAILED the presidential race, he weighted by party ID, so it was all the rage.

Back then when a poll got weighted by party ID it was usually something similar to +5 to +7 more Dems than Reps (give or take a point or so)

Back in 1994 with all the anger against Clinton, the popoularity of the GOP "Contract with America" etc the GOP was hugely energized, and all the polls in the raw data showed that GOP and Democratic turnout would be basically identical at about 40% each.

Polls that weighted back they essentially "threw away" the "extra" GOP 7% when they weighted, and hence there were a lot of surprised pollsters in 1994 when Republicans were indeed 40% of the people who showed up on election day.

To use a more recent example, we have Mr. Reagan's death.

Lets say as a hypothetical, that 2% of Democrats, remembering fondly Mr. Reagan, decide that they are actually in their heart of hearts republicans, and will vote for Bush.

A poll which allows party ID to bounce around WOULD pick up this shift, a poll that did a hard weight would miss this shift by "weighting it away"

I do look at party ID in a poll, and especially at the state level where in most cases current registration data is available, weighting to ID is certainly not a "crazy" thing to do by any means.

What I am really looking for is if the sample in the poll is truly a "random" one.

We do know, for example, that on a national basis, the GOP and the Dems are VERY close.

Gallup over their last 40,000 interviews says party ID is within 0.3% of equality.

PEW over their last 26,000 interviews says the Dems have a 1.5% advantage.

EVERYBODY who was even minimal common sense knows its close.  If your a Republican you might say its tied, if your a Democrat you might say +4 to the Dem side, But it's within a very few % points everybody knows and agrees.

When I look at a national poll, I want to see that the partisan breakout is "fairly" close to even.  If the Dems are up 4 or the GOP is up 2, I don't get too excited either way.

It's like the breakout of men and women.  We know the breakout should be pretty close to 50/50. (52/48 at the likely voter level)

If we see the poll has men/women at 57/43 we KNOW it is NOT a random sample, and we toss the whole poll.

If we see a national poll that is 40% GOP and 30% Democrat, or 38% Dem and 25% GOP we KNOW it is not random and we toss the poll just as surely.



The recent Ipsos Poll that has Bush +1 had a 49/42 Dem/Rep party self identification ration - a bit dem heavy, but again not totally crazy.  Three months back the ID was 47/45 to the GOP side - inthe case of IPSOS, it's random, and not systemically biased in any one direction.

That is one of the reasons I semi-like survey USA,  Their methodology is absolutely "straight" - and while I am personally not a fan of the Robo-polling machines, I am quite confident there is no built in bais on way or the other.

Does Survey USA screw up? -

Sure they do - Kerry was never up 10% in Michigan, and Bush was never within 1% in California, but over many polls and many months, I think Survey USA is pretty decent actually.

Bush +7 in Florida? I expect there is a few % of "reagan effect" in there (Florida has a TON of Seniors who remember Reagan) But Survey USA did the correct thing, and clearly flagged the death of Reagan as a possible bais in the poll.

It's one more poll.... I expect we will have 20 others in Florida between now and November.....

[The Vorlon]

723 Florida LVs surveyed 6/12-6/14

Bush 50
Kerry 43

http://www.surveyusa.com/currentelectionpolls.html

Mr. Gov NickG....

Tennessee in pollbooth went from solid to lean for Bush... did I miss a poll somewhere?

[Fmr. Gov. NickG]

723 Florida LVs surveyed 6/12-6/14

Bush 50
Kerry 43

http://www.surveyusa.com/currentelectionpolls.html

Mr. Gov NickG....

Tennessee in pollbooth went from solid to lean for Bush... did I miss a poll somewhere?

Hmm....that's a good call.  I checked my spreadsheet, and apparently last week's update reverted to my old weighting of the 2000 baseline, which I had changed a couple weeks ago.

There are very few states where this new weighting made a difference, but Tennessee was right on the cusp between Lean Bush and Likely Bush...I changed the weighting in the first place so this sort of thing wouldn't happen.  I've fixed my formulas now, but I'll just wait until Friday to fix it on the site.

BTW, Tennessee will by Likely Bush in new update under either scheme, when the results of the new POS surveys are included.

[The Vorlon]

723 Florida LVs surveyed 6/12-6/14

Bush 50
Kerry 43

http://www.surveyusa.com/currentelectionpolls.html

Mr. Gov NickG....

Tennessee in pollbooth went from solid to lean for Bush... did I miss a poll somewhere?

Hmm....that's a good call.  I checked my spreadsheet, and apparently last week's update reverted to my old weighting of the 2000 baseline, which I had changed a couple weeks ago.

There are very few states where this new weighting made a difference, but Tennessee was right on the cusp between Lean Bush and Likely Bush...I changed the weighting in the first place so this sort of thing wouldn't happen.  I've fixed my formulas now, but I'll just wait until Friday to fix it on the site.

BTW, Tennessee will by Likely Bush in new update under either scheme, when the results of the new POS surveys are included.

so it was an old poll expiring, not a new one coming on.. ok... just checking..

Did you find that Mason/Dixon from Washington state BTW...?