[lbo-talk] Excellent analysis of Time Poll

ira glazer ira at yanua.com
Sat Sep 4 07:48:12 PDT 2004


http://www.dailykos.com/story/2004/9/4/03450/25839

My thoughts on the Time Poll

by mattb25 http://www.dailykos.com/user/mattb25

This is going to be a somewhat stream of consciousness analysis; not so much of the data in the poll, but of its circumstances. The bottom line for me is that while it's always worrisome to see any poll have Kerry 11 points down (amongst either LV or RV), I'm not gonna get truly alarmed until I see this validated in a minimum of two more polls (I say two more because it truly wouldn't shock me if CNN/Gallup has Bush up 15 in its LV poll)

1. I'm curious as to the exact timing of this poll, specifically as

to whether or not it included interviews taken after Bush's

speech. If all of its data is pre-speech, then its CLEARLY an

outlier, as four other polls tell a wholly different story

(Rasmussen, Zogby, ARG, Economist).

2. If the poll does include data from interviews taken after the

speech then, quite frankly, I find it less alarming. Consider who

might have been likely to answer the phone last night to talk to a

political pollster. The same is really true of the whole week.

It's important to keep in mind that Time DID NOT poll over the

four days of the Kerry convention. It waited until August 3-5,

while Kerry accepted on July 29th. Newsweek and Zogby were the

only polls to be in the field during the actual four nights of the

Dem Convention (July 26-29). Of those, Newseek was the only poll

that actually sampled on July 29th, the night of Kerry's

acceptance speech, and on that particular night, it had Kerry up

54-41 amongst Registered Voters. So basically, the only pollster

thus far who has released polls during the EXACT same time frame

for both conventions is Zogby. He released polls conducted over

the first three days of each convention and found Kerry to be up 5

during the DNC, and Bush up 2 during the RNC. My point here is

really this: we have no idea what a poll conducted over the four

nights of the Dem convention would have showed, since nobody

conduced such a poll. Hence, we have no idea if Kerry would have

had a similar type of lead, that then quickly modified itself

3. I'm not going to suggest that Time is part of a conspiracy, as I

don't really believe too too much in that kind of thing. But I

will say objectively that their polling typically shows the

largest margins between the candidates post-convention. In the

8/3-8/5 Time poll, Kerry led by 7 amongst LV, and 8 amongst RV;

those were, respectively, his largest margins in both categories

in any post DNC poll (a later Zogby Poll also showed Kerry up +7).

So, for one reason or another, Time seems to really swing in both

directions, which brings me to my next point:

4. I'm not sure whether Time is weighing its sample. They do not

indicate that they do in their methodology, but that doesn't

necessarily mean that they don't. Some folks on here knock

weighing without knowing what it really means, but in my view,

it's often essential to an accurate poll, mostly so during a

period like a convention, when one side is more likely to be

paying alot of attention (and note--weighing is wholly different

from determining who is and who is not a "likely voter.") But the

real point here is that if Time does not weigh its sample (and

again, I don't know for sure), then the results are explained

right there: there is absolutely no doubt that during the RNC,

repubs are more likely to be home watching the convention, to

answer the phone, and to identify themselves as likely to vote.

Likely voter models skew Repub in general (LV models are also much

more inaccurate at this stage, particularly this election cycle),

but ESPECIALLY so during a major Repub event. The prime example:

right after Reagan's death, the Harris Poll showed Bush leading

51-41 amongst Likely Voters. That poll was the only poll around

the period with an LV model; other polls, including Pew, that

measured RV did record a Bush bump, but a much smaller one. The

Harris poll turned out to be a complete outlier. Why? Because

Repubs were the folks watching the whole freaking week, and they

were the folks inclined to answer the poll.

Zogby, for instance (and most reputable pollsters), weighs his sample. What that basically means is that he has a model Dem/Rep breakdown that he uses to adjust the raw count if it's off by a significant margin from the ideal. I am not sure as to his current breakdown, but in the past his model was based on a typical random sample consisting of 34.5% Dems and 34.0% Repubs, the rest independent/third party (weights are also applied for certain demographics). So, for example, if Zogby's raw count showed 45% Repub and 30% Dem, the Dem votes would count more than once, in order to bring the sample into line with the model. I don't wholly agree with the weighing method, but it is is fairly invaluable during an event like a convention, when most of your responses are likely to come from one party. I haven't seen the Time breakdown anywhere, but I'm definitely willing to bet that the party ID breakdown has a major skew, particularly amongst Likely Voters.

Amongst RVs in a two way race in the Time poll, it's Bush 50, Kerry 42. I find that slightly more plausible, but I definitely have a gut feeling that Time may just be reporting its raw count, which probably has more self-identifying Repubs than you would find during ANY other polling period.

5. I don't think that Time has an inherent Bush bias in its

horserace, but, without a doubt, its approval ratings for the

Chimp are consistently above the average. The Time Poll and the LA

Times Poll are the ONLY TWO polls that have never had Bush job

approval under 50. Never. I mean, even Rasmussen and Fox have had

Bush slipping into the 40s.

6. The RV numbers aren't good, but they're not earth-shattering, the

race is over type of thing. Bush's re-elect number amongst RV's is

still under 50 (49) and in three way RV race, Bush is still below

50 (49-40-5). Oh, that's another thing--Time consitently has Nader

running higher than anybody else. Nader isn't pulling 5 anywhere

now. This may just sound like spin, but if Bush, during the

ABSOLUTE FOUR BEST DAYS HE WILL HAVE, can't get over 50 amongst

RV's, then I still think he's in trouble.

My point in this post isn't to try to deny that Bush has made some gains. He certainly has, but I still absolutely rate this race a toss-up. Like I said above, we've seen Bush up by about this much in individual polls before (Harris Poll, 51-41), and it didn't last. I'm definitely not gonna say that this is the position I wanted to be in, but I'd bank alot of my credibility on saying that Kerry's not down 11. 7, 8 I could buy, though I really think within a week to 10 days, Bush will up 2-3 tops. Tommorrow's polls may make me look like an utter fool, but I'm willing to take the risk. Bottom line: I think that the poll is an outlier. Not because I don't like it, or because I think there's a conspiracy--but because of its circumstances and possibly its methodolgy.

> and a comment of his on Gallup

I honestly put nothing past Gallup at this point

They were done with me when they showed a "Bush bounce' during Kerry's convention. I don't really think it will be 15, I just meant that if it was, I'd just shake my head, laugh, and drink a beer. What's weird with them is that they were much more in line with other polls right up until the Dem convention, and then they began a total outlier. Weird. I'm really not a conspiracy person, but sometimes I wonder, given how blatant the CNN bias is. Basically, after this election, I'm either gonna think that Gallup is indeed a genius organization or I'm never gonna look at another poll they put out.



More information about the lbo-talk mailing list