AP Electoral College Analysis

[CARLHAYDEN]

Suggest everyone check out the AP Electoral college analysis.

http://apnews.myway.com/article/20040724/D841CACO0.html

[© tweed]

I saw that on AOL.

We know more than they do.

[millwx]

Suggest everyone check out the AP Electoral college analysis.

http://apnews.myway.com/article/20040724/D841CACO0.html

I just wish some folks would stop calling states "toss-ups" that aren't.  Examples?  AP calls the following states toss-ups: Florida, Ohio, Iowa, Nevada, New Hampshire, New Mexico, Wisconsin, Michigan, West Virginia, Pennsylvania and Oregon.

What is AP basing this analysis on.  I'd presume polling data.  But...  A poll average any one of us here can create is just as accurate as the data they're dealing with.  And even the most conservative of analysts would deem a lead 5% or greater to be NOT a toss-up.  Now, let's look at their toss-up states and the current month-long poll averages...

Florida: Kerry +1.4%
Ohio: Bush +1.5%
Iowa: Kerry +2.5%
Nevada: Kerry +1.0%
New Hampshire: Kerry +5.0%
New Mexico: Kerry +7.0%
Wisconsin: Kerry +4.0%
Michigan: Kerry +4.2%
West Virginia: Bush +7.0%
Pennsylvania: Kerry +6.2%
Oregon: Kerry +8.7%

FIVE of their eleven states deemed "toss-ups" simply are not.  I'm not saying they're "heavy leans"... like NH, which just barely hits the 5% range... but they are at least "slight leans", not toss-ups.

Besides that, on the flip side, there are ALSO toss-ups that they're classifying as "leans".  Using the 5% "rule" Arkansas, Maine, Minnesota, Missouri and Virginia all meet that criteria.  Let me be clear, I do not view some of these as toss-ups myself, as I don't use 5% as my criteria... nor do I even necessarily believe the poll average (I'm merely trying to be as objective as possible).  For example, as it stand now there's no way in hell I'd call Virginia a toss-up.  But that's beside the point.  The point is, this AP analysis appears to be based on some very subjective, and thus, potentially horribly wrong, analyses.  Impact-wise, what does that mean?  Well, using 5% as the toss-up rule, here's where the poll average stands... Kerry 217EVs, Bush 192EVs.  I'd trust that long before the AP analysis, derived from what?, we don't even know.

[TomC]

They like to keep us watching. This race will be a nail biter if for nothing else, we'll watch and read.

[CARLHAYDEN]

Suggest everyone check out the AP Electoral college analysis.

http://apnews.myway.com/article/20040724/D841CACO0.html

I just wish some folks would stop calling states "toss-ups" that aren't.  Examples?  AP calls the following states toss-ups: Florida, Ohio, Iowa, Nevada, New Hampshire, New Mexico, Wisconsin, Michigan, West Virginia, Pennsylvania and Oregon.

What is AP basing this analysis on.  I'd presume polling data.  But...  A poll average any one of us here can create is just as accurate as the data they're dealing with.  And even the most conservative of analysts would deem a lead 5% or greater to be NOT a toss-up.  Now, let's look at their toss-up states and the current month-long poll averages...

Florida: Kerry +1.4%
Ohio: Bush +1.5%
Iowa: Kerry +2.5%
Nevada: Kerry +1.0%
New Hampshire: Kerry +5.0%
New Mexico: Kerry +7.0%
Wisconsin: Kerry +4.0%
Michigan: Kerry +4.2%
West Virginia: Bush +7.0%
Pennsylvania: Kerry +6.2%
Oregon: Kerry +8.7%

FIVE of their eleven states deemed "toss-ups" simply are not.  I'm not saying they're "heavy leans"... like NH, which just barely hits the 5% range... but they are at least "slight leans", not toss-ups.

Besides that, on the flip side, there are ALSO toss-ups that they're classifying as "leans".  Using the 5% "rule" Arkansas, Maine, Minnesota, Missouri and Virginia all meet that criteria.  Let me be clear, I do not view some of these as toss-ups myself, as I don't use 5% as my criteria... nor do I even necessarily believe the poll average (I'm merely trying to be as objective as possible).  For example, as it stand now there's no way in hell I'd call Virginia a toss-up.  But that's beside the point.  The point is, this AP analysis appears to be based on some very subjective, and thus, potentially horribly wrong, analyses.  Impact-wise, what does that mean?  Well, using 5% as the toss-up rule, here's where the poll average stands... Kerry 217EVs, Bush 192EVs.  I'd trust that long before the AP analysis, derived from what?, we don't even know.

Patience is a virtue, I keep telling myself.

So, to repeat what I (an Vorlon) have posted (in brief).

You cannot trust many of the public opinion polls, as they are extremely biases and very defective (CBS, LATimes, and Princeton Associates come to mind).

Other polls can suffer from a number of defects which are unintentional, but nonetheless, significant.

Historically, when a Republican is President and seeking reelection, the Democrat performs better in the summer polls than on election day.

Polling is also more difficult today than a few years ago due to changes in the telecommunications industry.

There are other problems, but I have other things to do, so just look up some of the comments Vorlon and I have made over the past couple of months.

In conclusion, don't place too much confidence in the polls you see, as they generally are not very reliable.

[millwx]

In conclusion, don't place too much confidence in the polls you see, as they generally are not very reliable.

I understand what you're saying, but unless we just make crap up, they're all we have to go on.  And when many polls in a row agree, I think we can have a fair amount of confidence.  Also, poll averages can smooth things out if the poll errors are random.  An example of consistency in a state that's AP is calling a toss-up (to their credit they do say it may fall to Kerry soon)... Oregon.  There has not been a poll since May there showing Bush ahead or even tied.  There have been seven polls there since.  The closest Bush came was 4% in a Public Opinion Strategies poll.  Unless your analysis is biased, flawed, or just plain bad, Oregon is in no way shape or form a toss-up.  That's just one example... Pennsylvania is almost just as bad, but an Opinion Dynamic poll at Bush +5% there does muddy the waters a bit; still that OD poll was over a month ago and no poll since shows Bush closer than 5% and every other June poll prior to OD had Kerry ahead.  Other states have fewer polls, so some concern over the polls or a polling average may be more valid.  Still, those polls are fairly consistent in thos non-battlegrounds (NM - 6 polls since May & Kerry tied or ahead in all, NH - 9 polls since May & Kerry tied or ahead in 8, WV - this is the only questionable one... 7 polls since May & Bush ahead in 5, but of the non-Zogbys it's two polls favoring Kerry, two favoring Bush).

The point is, you can whine and gripe about polling problems all you want.  But they don't all suck.  So, when they all repeatedly and consistently show something, it's probably true.  As such, there's no basis for this AP analysis.

[TomatoSoup]

I saw that on AOL.

We know more than they do.

That would be true.

[CARLHAYDEN]

In conclusion, don't place too much confidence in the polls you see, as they generally are not very reliable.

You said:

I understand what you're saying,

I reply:

based on the rest of your post, no you don't

You said:

but unless we just make crap up, they're all we have to go on.  

I reply:

Wrong again.  I have posted a rather lengthy and detailed explanation in the 2004 Predidicts - discussion thread, page 157.

You said:

And when many polls in a row agree, I think we can have a fair amount of confidence.  

I reply:

This simply is NOT the case.  Apparently you are not familiar with what is known as the 'lemming. problem.  Eliminating the completely absurd polls (CBS, LA Times, Zogby and Princeton Associates), and looking only at the others, all the publicly reported polls suffer from some defects which to some extend skew they results.  

Historically (as I have previously noted), Democrat challengers perform better in summer polls, so even very good public polls (I am very impressed with both ISPOS and Mason-Dixon) are skewed somewhat for this reason.  

Other pollsters have additional problems in that their sample composition is significantly overweighted with Democrats.  I look specifically at the internals of polls and have frequently commented with specifics about such problems.  Interestingly, PEW has noted this problem in a roundabout way.

You said:

Also, poll averages can smooth things out if the poll errors are random.  

I reply:

As I noted above, and have noted before, the errors are NOT random.

You said

The point is, you can whine and gripe about polling problems all you want.  

I reply:

Again, you have not been reading my posts very carefully.  I have repeatedly, and in detail noted the defects in many of the polls.  Check it out yourself.  See what Vorlon has posted in the past.  (I'm looking forward to his return in about two weeks, maybe you will believe him).

You said:

But they don't all suck.  So, when they all repeatedly and consistently show something, it's probably true.  As such, there's no basis for this AP analysis.

I reply:

Again, you are relying on reported polls.  AP uses ISPOS (a high quality firm), to advise them.

[mddem2004]

Suggest everyone check out the AP Electoral college analysis.

http://apnews.myway.com/article/20040724/D841CACO0.html

CARLHAYDEN,
Just what polls DO YOU TRUST? I see a fair amount of discrediting of polls on your part, but do you trust ANY polling firm?
Forgive me if you have stated so in prior posts if you have, just ain't seen it........

[© tweed]

mddem04,

Carl only trusts polls have have Bush leading.

[CARLHAYDEN]

Suggest everyone check out the AP Electoral college analysis.

http://apnews.myway.com/article/20040724/D841CACO0.html

CARLHAYDEN,
Just what polls DO YOU TRUST? I see a fair amount of discrediting of polls on your part, but do you trust ANY polling firm?
Forgive me if you have stated so in prior posts if you have, just ain't seen it........

Aside from private polls, I think ISPOS is pretty good as is Mason-Dixon.  You do have to allow for seasonal flucuation, but otherwise, they're pretty good.  Also, if you remove the oversampling of black Democrats done by Rasmussen this year (in response to 2000 results), you will get a pretty good result on the national polls there (the state polls are wildly unreliable due to the amalgamation of surveys on different dates).

Oh, and yes, I have stated this in previous posts.

[CARLHAYDEN]

mddem04,

Carl only trusts polls have have Bush leading.

Now Boss, if you search your memory you will note that I was critical of a Harris poll that had an excessive lead for Bush, about five weeks ago.

Also, I did not contest the Humprey Institute polls which show Kerry slightly ahead in the upper midwest.

There are a couple of reason why publicly available polls favorable to Kerry ARE more defective than all publicly available polls.

First, the respondents in those polls do not accurately mirror the actual voters on election day, but instead reflect a more heavily Democrat bias than is the case (this is what, in part, I mean by the 'internals.").  This results in those polls reporting a higher level of support for Democrats than occurs on election day.  

Second, some of the polls are deliberately and heavily skewed to get the results their commissioners want (viz LA Times, CBS news, Princeton Associates et al.)
 

[millwx]

Suggest everyone check out the AP Electoral college analysis.

http://apnews.myway.com/article/20040724/D841CACO0.html

Just what polls DO YOU TRUST?

He trusts his own biases and subjective opinions.  He's made that abundantly clear.  I give up.  He's made negative comments on vitually every thread that includes a poll showing Kerry ahead.  Yet he never complains about one showing Bush ahead.  Meanwhile, he selectively cherrypicks data, as I showed in the Nevada thread.  I give up.  I'm not even going to discuss it any more.  I'll just post the data and he can flame away.  If people want to believe him, fine, their loss.  If they want to believe data, fine.

[Reds4]

Being a Republican, I must say, this AP Electoral college analysis, which is on the front page of my paper today here in Ohio, is certainly skewed to Bush. Michigan is not a tossup. Arkansas and Missouri should be tossups. OR should probably be leaning Kerry, and Iowa should definitely be.

[CARLHAYDEN]

Please note, as I recently replied to Boss, I critized a Harris poll showing Bush with unrealistic lead.

Also if you check on page 157 of the 2004 predictions - discussion thread on this board you will see (but perhaps not understand) the data and analysis I use to make the projections.

So far you show a touching naivite about survey research.

[TeePee4Prez]

Wow, a lot of people think PA is in play.  Forget it George, GO HOME!!

[khirkhib]

Please note, as I recently replied to Boss, I critized a Harris poll showing Bush with unrealistic lead.

Also if you check on page 157 of the 2004 predictions - discussion thread on this board you will see (but perhaps not understand) the data and analysis I use to make the projections.

So far you show a touching naivite about survey research.

You sure do like yourself.

[Gustaf]

I don't know whether to laugh or cry at all those of you attacking Carl for pointing out that your using faulty data...seriously, raise the level of debate a little above "ew, you like Bush, accept our crappy polls or else".

Give us a reason to trust those polls instead, don't whine about subjectivity.

[CARLHAYDEN]

Would appreciate your imput on my posting on page 157 of the predictions thread of this board.

[Gustaf]

Would appreciate your imput on my posting on page 157 of the predictions thread of this board.

Who's? Mine?

[Rococo4]

Seems kind of crazy Kerry would reduce ads in Missouri - maybe both campaigns know something we dont.  Bush has to win MO - you would think Kerry would make him earn it.