« Convention Bounce Predictions, Round 2 | Main | Kerry Ad Buy Hints at Battleground Locations »

Wednesday, September 01, 2004

Another Quiet Period for State Polls

Posted by DavidNYC

Around the time of the Dem convention, we didn't see too many state polls. It looks like pollsters have similarly gone quiet. SUSA's most recent polls are from 8/25, and the only new material available on Polling Report (subscription required) are surveys from Strategic Vision, a Republican outfit. Even the prolific Rasmussen is on hiatus.

So I guess us poll junkies will have to wait a few more days before we have our next batch of polling heroin. In the meantime, you should check out Ed Fitzgerald's latest meta-projection. I can't imagine how much work he puts into it, consider there are now 46 sites in his survey. (Everyone has an opinion, eh?) I especially appreciated this useful graph, which shows that most watchers think the race is getting tighter:


I'll be really curious to see who winds up with the most accurate projection in the end.

Posted at 02:40 PM in General | Technorati


Here are three new state polls

Posted by: PF at September 1, 2004 06:51 PM | Permalink | Edit Comment | Delete Comment

I have to say that I remain extremely suspicious about the Strategic Vision polling. I do not put it beyond the scruples [sic] of the GOP to insert deliberately skewed polling into the mix in other to create the impression of momentum in their direction, and I'm struck by the fact that this virtually unknown Republican-connected polling firm all of a sudden ramps itself up to doing almost a dozen 801-person polls each and every week, week after week. That's far more volume (as far as I can figure) than Rasmussen, Zogby, SUSA or ARG do. Someone's footing the bill, I think, and it's not impossible that someone's also pre-determining what certain results will be.

But that's mere speculation and suspicion, perhaps even paranoia on my part. There's another, simpler, less devious (perhaps) explanation for the fact that SV's polling seems to skew to Bush in any circumstance in which a Bush lead would appear to be (if only barely) credible, and that is that all of their polls are of likely voters. Depending on what model they're using to get their likely voter results, that could skew their results in the same way that Gallup's results are skewed, in favor of Bush.

(Of course, this doesn't completely put to rest the possibility that someone's paying to get the SV results out there knowing that the likely voter model is going to, in many circumstances, help Bush.)

People are starting to take the idea that SV is a Republican firm seriously and label their results as such, but a lot of people still take their results at face value, which, I think, is a mistake.

Posted by: Ed Fitzgerald (unfutz) at September 1, 2004 07:35 PM | Permalink | Edit Comment | Delete Comment

Dear colleague (I also do Electoral College projection analysis)

I had the same reservations you have about SV, but have found that even though results maybe a little skewed they usually are not.

Don't take my word just go an analysis of same period polls. Use the state of Florida polls available at RealClearPolitics.com. You will find out that most of the time SV is right on the money. You will find out that even respectable polls like Quinnipiac College seem to be wrong some times.

Why don't you do like I do. Don't use obviously flawed polls as main source for EV assignment, use them only as corroborating material in the absence of the best polling sources (Mason-Dixon, Gallup)

I am surprised that you claim that Gallup is biased toward Bush. If anything they are biased toward Kerry since no matter how bad a poll looks on Kerry they always find a favorable headline to use to make Kerry look good.

Anyone that does serious electoral college analysis needs to leave his political preference outside.

I laugh loudly when I see analyses pointing to either a Kerry or Bush landslide at this juncture.

By the way if you care, this is my current projection:

Bush-Strong/Lean = 217

Kerry-Strong/Lean = 233

Toss Ups - 88

Assigning Toss Ups: Bush 284 Kerry 254

If you include my projection in your web page refer to me WalterAlonso. I will provide updates to your e-mail if you provide it to me.

E-mail at matchmail@excite.com

Posted by: Walter A. Avellaneda at September 2, 2004 11:20 AM | Permalink | Edit Comment | Delete Comment


I believe this is a copy of an e-mail you sent to my blog's address, so I'll respond specifically to you there.

Of things that are relevant here, my writing was sloppy, I didn't mean to imply that Gallup was biased toward Bush per se, but merely to say that their emphasis on reporting their likely voter results skews things in Bushs's favor, and since SV reports *only* likely voter results, their findings are likely to be skewed in that way as well. (See Ruy Teixiera's analysis of the polling in Pennsylvania, which shows that when you compare apples to apples -- registered voter results to registered voter results, a lot of "tightening up' of the race there simply go away. The same is true of the national results.)

I do indeed do as you suggest -- I don't use or de-emphasize partisan polls (including Strategic Vision's), likely voter results, results that include Nader in non-Nader states, and the Zogby battleground polls. That doesn't mean I ignore them entirely, I simply weigh them depending on what I perceive to be their flaws and their virtues.

Posted by: Ed Fitzgerald (unfutz) at September 2, 2004 12:39 PM | Permalink | Edit Comment | Delete Comment

Oh, right now my projection is a little higher for Kerry than the average tracking site is, but in the past it's also been a little lower. That's because I don't allow states to move quite as easily as others do -- many sites have states changing status on the basis of whatever the latest poll is. For me, it takes several (at least) polls with a consistent result to change a state's status, and even then states have to move from a candidate to "toss-up" before they can move to the other candidate. This makes my projection inherently less volatile, but also means that it tends to lag behind the projections on other sites.

(The exception for me is toss-up states, which I *do* allow to move freely, assigning them to the winner of the latest poll, on the theory that a "toss-up" is a state without a consistent result allowing me to assign them to one candidate or the other, and therefore should float until it settles down.)

Specifically, Bush's leads in Wisconsin were sufficient to move it from DEM to TOSS-UP, and it was therefore assigned to Bush, the winner of the most recent poll. When a new poll came in with Kerry in the lead, it went to Kerry. If there are several more Kerry leads, the state will go back to being DEM, but if there are several Bush leads it will go to GOP and not change from that status until several (or more) Kerry leads in a row provoke me to move it back to TOSS-UP. The same situation is true in Ohio, Tennessee and other toss-up states.

My methodology is my methodology, and I try to stick to it and not adjust it because it "seems" too high for one candidate or the other.

Posted by: Ed Fitzgerald (unfutz) at September 2, 2004 01:01 PM | Permalink | Edit Comment | Delete Comment

You do good work Ed. Don't worry about defending it (even though Walter wasn't attacking you).

Posted by: Michael at September 2, 2004 02:16 PM | Permalink | Edit Comment | Delete Comment

OK, prompted by Walter (and a comment on my own blog) I did take a closer look at Strategic Vision's results -- the findings are here.

Summary: In six of seven states I looked at, based on the last month of polling, Bush did better in SV polls than he did in polls by other organizations. Across the seven states, he averaged an advantage of over 2.5 points.

Posted by: Ed Fitzgerald (unfutz) at September 2, 2004 02:21 PM | Permalink | Edit Comment | Delete Comment

Good job Ed !!!

Lesson learned: if someone uses a Strategic Vision poll be aware that it maybe about 3% from the consensus of other polls.

You can say the same about Fox.

In the same way you can assume most of the time that polls done by NPR are biased to Kerry

Posted by: Walter at September 2, 2004 05:08 PM | Permalink | Edit Comment | Delete Comment

The difference is that NPR isn't releasing several polls a day in crucial battleground states. They've put out two national polls in two months, Strategic Vision has put out three or four dozen polls in the same time.

Posted by: Ed Fitzgerald (unfutz) at September 2, 2004 09:25 PM | Permalink | Edit Comment | Delete Comment


One could say then that they have a strategic vision on how to get their guy elected. (Sorry couldn't resist.)

Posted by: Michael at September 3, 2004 01:36 AM | Permalink | Edit Comment | Delete Comment