Posted on 10/12/2004 5:45:03 AM PDT by GEC
Zogby now has the race back to a tie. It is very difficult, within a multi-day tracking poll, to come up with individual day samples that result in a spike one way and a return spike the very next day.
Of course he had.
Special sauce, lettuce, cheese...
bump
Unless...you play games with the political affiliations of those being polled.
Do the words WISHFUL THINKING ring da bell? ;)
>>realpolitics.com<<
Bad URL. Will take you to RAT site.
Whenever Zogby is part of a story we need a BARF ALERT
Mine doesn't.
Up until 2000, Zogby's polls were accurate and were much quoted here. But what really put the seal on it was 9/11. Zogby is not only a Democrat, his brother is a bigwig in one of those Arab-American rights groups. He cooked the books a little for the Dems, but now he is cooking them a lot for the Arabs.
One trick you can count on with these pollsters is that they will slant the numbers before election day and then adjust them at the last minute, when people aren't looking, so they can preserve their reputation for accuracy. It's too early to be doing that now, but he must have some reason for making this adjustment.
Monday 50
Tuesday 50
Wednesday 37
Thursday 37
Friday 37
Saturday 50
On Wednesday, Freepers would freak out en mass simultaneously declaring the race over and the poll bogus. But here is what the raw data looks like to generate those results...
Monday 50
Tuesday 50
Wednesday 10
Thursday 50
Friday 50
Saturday 50
Just a single day of bad data shifts the average of what is actually a pretty steady race. Obviously that is an oversimplified example, but it shows how just a single day of bad polling data can effect a rolling average poll. That is why it is important to watch subsequent days of a major shift in a rolling poll to determine whether there is a trend developing, or if just one day is off.
I should add that nobody knows WHAT Zogby is doing with his numbers. He has changed his methods from 2000 but hasn't revealed his new techniques.
But here what we had was a spike in Kerry's direction followed by a spike in the other direction the very next day. Note that in your example, the spiked movement lasted three days.
He's weighting to D39% R35%. He said so recently in an interview (sorry, link not handy).
True. In the sample I provided, the day after the 10 point data sample would have to be a 90 point data sample to reverse the trend. That is why I believe Zogby is greatly manipulating his figures beyond generating a 39/35 R/D split. From what I've read he is creating his own demographic breakdown based on results from his own polling to determine shifts in voter sentiment. That requires him to assume his demographic polling is accurate in the first place. If it isn't, his horserace poll is completely bogus (and I believe it is). While he may use a consistent 39/35 split, his likely voter model must shift like Kerry's opinion. Something certainly does. Based on that, his poll isn't even useful to determine trends.
I have not put much stock in Zogby's polls for quite some time.
(Which means- I agree with you!)
Well Bush went up 1 and kerry down 2 in a ROLLING AVERAGE POLL. Just what kind of numbers spiked Kerry up 2 day before and down 2 today? Day 3 is a draw or Kerry 1 so 400 or so respondents on DAy 2 and 3 to get Kerry +3 assuming a 1200 person 3 day tracking poll:
Day 1:
Kerry 4x46=184 respondents for Kerry
Bush 4x45=180 respondents for Bush
Day 2:
Same Margin so same numbers just doubled
Anoterh 184 for total 368 for Kerry
Another 180 for total 360 for Bush
Day 1 to get extra 2 points for Kerry:
196 for total of 564 for Kerry (47%)
168 for total of 528 for Bush (44%)
So you see what has to happen in one day to get that 3 point lead. Now drop off Day 1 and imagine what today had to look like to in terms of one day difference between the two to drop Kerry 2 and raise Bush 1. THe difference in Day 3 was 28 where the other 2 days was 4. Now to swing this thing back, we have to use Day 2 and Day 3 as starting points.
184 +196 for Kerry=380
180+168 for Bush=348
32 respondent difference.
Now for the kicker
New Day 3:
160 for Kerry = 540 (45%)
192 for Bush = 540 (45%)
A 24 respondent uptick for Bush and a 36 respondent down tick for Kerry. Really, really, really wierd. This race is all over the map or Bush still has 2 point lead. CBS was as accurate as Zogby in 2000 I believe and they have Bush up 3. Perhaps, despite Rassmusen, there is a weekend trend and Gallup's weekend only poll ran into that as did Zogby. Whatever, that swing in last two days of Zogby to give kerry 3 pts back to tie in single day is asounding.
Basiclly, the New Day 3 poll had a 60 respondent turn around to affect a 1200 person tracking poll the way it did. That is just weird. So to expand on what I said earlier, polling on the weekend gave us Zogby's uptick and Gallup's numbers. Add in a weekday and you get a swing back in Zogby's numbers, you get a CBS poll with Bush up 2. Imagine a weekday only polling sample with this trend? If I were pollster, I'd only do Sunday, Monday, Tuesday polling. Thursday, Friday, Saturday is too much going on.
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.