Posted on 09/22/2004 9:32:46 AM PDT by ironmike4242
There was a bit of grumbling over ARG's sampling techniques. They seem to have oversampled women quite a bit in this poll. So I went through their numbers and re-did a few key states to see how a normal gender distribution might affect their results.
I found that there is a slight difference when you apply a reasonable gender distribution of 51% Women, 49% Men. They were using 55% women, 45% men in several of their states!
Here are results from a few selected states using my gender distribution (51-49) in stead of theirs. Frankly, the differences really aren't that severe.
Mississippi- ARG had Bush ahead 51.0-41.6. My gender weighting shows Bush ahead 51.4-41.1.
Ohio- ARG had Bush ahead 47.6-46.3. My weighting had Bush ahead 47.9-46.0.
Florida-ARG had Kerry ahead 46.3-45.0. I had Kerry ahead 46.0-45.5.
So you can kind of sense a pattern here. A more realistic gender weighting would make a slight difference, but wouldn't change the results in any state except for West Virginia and Wisconsin. ARG had both states tied, but I had Bush up in WV 45.9%-45.0%. In WI ARG rounded it off from Kerry up 46.2-45.7 to 46% each. Fair enough. By my weighting it was "more tied" with Kerry up 45.6-45.5.
This still doesn't validate the ARG poll completely. My numbers didn't take into account any party registration data, which looked skewed in a few states. But fair is fair. I was complaining earlier that ARG's data was "bizarre" and I was wrong. Mea Culpa. It could have been more accurate than it was, but the gender issue alone wouldn't have changed anything. Also keep in mind that without access to their raw numbers it's not possible for me to do a real statistical analysis. I had to rough through it as best I could.
ARG!-charlie brown
-ping to read later
bump
This seems to be a tactic of the more left-leaning polls. Regardless, I don't believe it. I think it's about 2% low for Bush (meaning a 4% spread).
It is likely that Kerry IS GOING to close the gap. Those who put out the "news" will see to that. Rathergate isn't going to stop the media from pushing their side. Quite frankly, I am amazed that any Republican could ever get elected to a national office... So the race is probably going to tighten up folks, IMO. There is just too much left-wing hysteria out there to NOT convince alot of those "swing" voters, who by definition are pretty gullible folks anyway..
I think they're oversampling people who will vote for Kerry.
Its not just the gender gap. Take a look at the voter registration.
Also, I wonder if this is just registered or likely voters.
More thoughts to ponder.
The bottom line, though, my friend is that the ARG poll is to the left of most other polls. Bush is up by 8 points in the Mason-Dixon poll and 11 points in Univ. of Cinci. but just up 2 in this poll. Generally, this year ARG has found Kerry up nationally and in Florida while others showed the race more in the Prez's favor and they've generally given him awful job approval numbers, the lowest of almost any poll - including Zogby. Just looking at their samples isn't good enough. Something's not right with the ARG data and it's just that simple.
Why does Quinnipiac have the race even and SUSA have Bush up by 4 in NJ while ARG shows Kerry up by 8? Why does ARG have Kerry up in the double-digits in Maryland while SUSA shows it even? Why is Bush up by just 9 in MS? Some of the polls look ok, but some look downright wrong. I believe what I have believed all along: ARG is skewed to the Democrats, whether or not it shows up in their state-to-state data.
You are missing the point. In virtually every case the state poll released today by ARG is 5 points better for Bush than same poll a month or two earlier. So these polls confirm what others have, that eveerything else being equal, there has been a 5-10 point shift to Bush in the last month.
Keep in mind that the media has had a Kerry love-fest the last couple days. Kerry is bound to have some good polls the next few days. We have to tighten our shoelaces and get ready because this thing is going to be nasty.
ARG's poll, when you multiply out the state percentages times the popular vote from last election shows a complete dead heat nationwide.
When you use polls from other outfits like Gallup, Mason-Dixon, Strategic Vision, etc. for states surveyed by ARG (and ignoring Zogby, Rasmussen, and SurveyUSA where possible), Bush shows a 4 million vote lead.
The main problem with ARG is that their "likely voter" model appears to almost exactly mirror the actual registered voter counts, whereas it is pretty well known that more Republicans tend to turn out than Democrats.
The other conflict with other polls is they show Kerry getting more of the Democrat vote than others do.
The most suspect polls they are putting out today are Ohio, New Jersey, New York, Delaware, and Mississippi.
And look at the visits..... First Lady in New Jersey (which I think is a deft ploy, I don't think Bush is in a position to win there, but make Kerry spend the $$$$!) The Bush team has somebody in PA almost daily. Minnesota, Wisconsin, Iowa are all getting increased ad buys from Bush, and increased visits.
Forget about the polls. forget about "Bush up 13! or "Kerry up 1." Forget about all of that stuff. Follow the money and it becomes pretty obvious where this race stands right now.
How did you go about reweighting the results? Were you able to obtain the polls sampling weights for each observation or did you reweight aggregate results? I don't think that the latter will give you what you think you want.
This poll is flawed for several reasons, but gender inequality isn't really a contributing factor.
Well, the ARG site gives you a breakdown for each state by percentage. So you can roughly extrapolate out their raw numbers based on gender (600 likely voters, 56% women, 51% support for Kerry=~171 votes for Kerry). So you can rework the numbers using 51% for women instead of 56% which would give you 156 votes for Kerry, and then you re-work the numbers that way.
Your point is well-made and well-taken.
Gender makes a little difference but check out the R/D/I
mix if you can...
Some of the lowball pollsters keep re-weighting their
results to the year 2000 exit polls (I think Rasmussen,
ARG, Zogby and others)
But that ignores the 5 point shift towards the Republican
'party ID' that has happened post-9/11.
And Republicans vote in higher percentages, so make that
a 6 or 7 point shift.
IIRC ARG had Gore winning a handful of Southern states in the final days of the campaign in 2000.
I don't put a lot of credibility toward them.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.