« From the Streets of New York City to Main Street, USA | Main | General Election Cattle Call, August 30 »

Monday, August 30, 2004

Pennsylvania Shifts

Posted by Seamus

As a follow-up to my prior post, I thought I'd start a thread dedicated to the other news coming out of Pennsylvania. The new Gallup Poll (already being discussed in the Open Thread) shows a tie in Pennsylvania:

(Aug 23-26)
Kerry/Edwards 47
Bush/Cheney 47

These "likely voters" numbers are a bit of a swing where Kerry was soundly ahead of Bush not so long ago. Interestingly, if you add Nader and consider "Registered Voters" the numbers look like this.

(Aug 23-26)
Kerry/Edwards 49
Bush/Cheney 43
Nader 3

So much of these numbers are obviously in the "Likely Voter" calculations. Lets hope those can be upset. But these latter numbers are interesting given the newest information that Nader won't be on the ballot in Pennsylvania. I am not one who thinks Nader will affect Pennsylvania whether he is on the ballot or not. This makes that debate irrelevant.

Posted at 05:54 PM in Pennsylvania | Technorati

Comments

Why is big media basing it's pollin of LV on a turnout of 100-105 million when it's more likely to be 115 million turning out?

Posted by: pc at August 30, 2004 07:01 PM | Permalink | Edit Comment | Delete Comment

I believe they base those likely voter numbers on past elections. I'm not sure they have any foundation for what will happen in 2004. It seems to me that past elections haven't had things like multi-million dollar movies and the internet influencing turnout quite like they will in 2004.

Posted by: seamus at August 30, 2004 07:28 PM | Permalink | Edit Comment | Delete Comment

I think that there will be a much larger turnout than 2000. That is why I like Zogby because his poll's are based on higher turnout than Gallop OR WSJ. Many are making fun of his inter active polls but I think he will emerge right on top as the closest projection of the final results.

Posted by: bobm at August 30, 2004 09:10 PM | Permalink | Edit Comment | Delete Comment

Regardless of whether the likely voter models of Gallup and others are sincere attempts to predict turnout (and I believe that they are), and regardless of whether it's too early to be using those models (and I believe it is), simply looking at the numbers they're coming up with should give the pollsters pause. When one candidate is ahead with registered voters, and the lead turns over when the likely voter transform is applied, it has *got* to raise some internal suspicions that their model is skewing the results is a way that's destructive.

That Gallup (and other pollsters) haven't taken the hint from their own results and re-thought their likely voter models (and stopped reporting those results until they do) is unconscionable, but in the meantime, I'm taking the registered voter results as being more representative, and ignoring the likely voter results for the time being. Perhaps when the election gets closer I'll pay more attention to them, but only if the pollsters perform some due diligence and take into account the greatly increased activism among Demeocrats compared to 2000.

Posted by: Ed Fitzgerald (unfutz) at August 30, 2004 09:52 PM | Permalink | Edit Comment | Delete Comment

Ed, your point is kind of what I was trying to hint at. How can the likely voter model work when you add Nader and shift it "registed voters" there is such as huge shift? That would suggest to me that either those likely to vote are just huge Bush fans (get real) compared to others or that the model is broken. And I agree with your point on this, the model is obviously broken.

Posted by: seamus at August 30, 2004 10:16 PM | Permalink | Edit Comment | Delete Comment

Well, one thing I've noticed is that most people are willing to believe which ever polls confirm their own political views, regardless of how legitmate the poll is. I think this is one of the dangers of polls, in that they can influence people whether or not they are accurate. That said, it does appear that Bush is gaining momentum in PA, a state where he seemed all but dead just a short while ago. No one has been able to explain to me what resuscitated him there, which is what I'm must curious about.

Posted by: Pepe at August 31, 2004 06:46 AM | Permalink | Edit Comment | Delete Comment

Pepe. I don't think there has been a definitive shift yet in PA. And even then we can't tell how big it is. As for what could have caused a shift? Take your pick. Could be SWBVT, perceptions of iraq, the olympics (nationalism), rnc, etc..

Posted by: seamus at August 31, 2004 01:00 PM | Permalink | Edit Comment | Delete Comment

To be fair to Kerry, these polls should include Badnarik, the Libertarian presidential candidate. Even at 1-2%, Badnarik can pull votes from Bush, but not if the media keeps shutting Badnarik out at Kerry's detriment.

Posted by: Wilson97 at September 1, 2004 09:28 PM | Permalink | Edit Comment | Delete Comment

Well, one thing I've noticed is that most people are willing to believe which ever polls confirm their own political views, regardless of how legitmate the poll is.

I think that's a rather harsh indictment, and I don't think it applies to a fair number of people, liberals and conservatives alike, who run the electoral college tracking sites I've been surveying.

In my own case, I don't make status decisions (assigning a state to either candidate or as a "toss-up") on the basis of partisan polls, whether they come from Democratic shops like Lake Snell Perry and Democracy Corps or Republican ones like Strategic Vision and Public Opinion Strategies. I also don't make major adjustments relying solely on Zogby's battleground polls (although I do use them to assign toss-up states to one side or the other), because of the widely voiced concerns about their methodology -- but that applies whether they show a lead for Bush or for Kerry.

And when three non-partisan polls tell me that George Bush leads in a state, I switch that state to George Bush, no matter that I'd obviously prefer to see it assigned to Kerry -- and that state stays assigned to Bush until the circumstances change, and by the same criteria by which a Kerry state would change to Bush.

My methodology is not as strictly objective as Chris' is, or that of some other of the e.c. trackers, but it is as fairly and even-handedly applied as I can make it be. I may not like what a poll is saying, I may discount the size of a lead because a likely voter model is used instead of a registered voter model, or because Nader is included in the poll when Nader is not on the ballot, but those are objections and adjustments made on a rational and factual basis, and it doesn't mean that I willingly disbelieve those polls just because they don't tell me what I want to hear.

Posted by: Ed Fitzgerald (unfutz) at September 2, 2004 12:35 AM | Permalink | Edit Comment | Delete Comment