Zogby 48 O 44 M - the poll is crap though

[FreeState]

http://www.zogby.com/news/ReadNews.dbm?ID=1584

In this latest report, McCain gained eight-tenths of a point, while Obama lost one full point.

While the worldwide economic system underwent dramatic turmoil last week, the poll shows the presidential race remarkably unchanged overall at the end of the first full week of daily tracking. Through the week, Obama has always held a small lead, as large as 6.1 points (on Saturday) and as little as 1.9 points (last Wednesday).

The rolling telephone tracking poll included a sample of 1,206 likely voters collected over the previous three 24-hour periods spanning four calendar days – approximately 400 per 24-hour period from Oct. 9-12, 2008.

Obama retains a substantial 17-point lead among independent voters, but that edge receded from 21 points yesterday. In terms of securing their political bases, both Obama and McCain are doing well. Among Democrats, Obama wins 86% support, while McCain wins 88% support among Republicans.

----

Bold added by me to show why its crap - there are more registered Dems than GOP by about 7% - then add the 17% of the 30% of independents. Also the ABC poll was almost 3 times as many people - more accurate - 400 people is to small a sample. (although it does not fair well on the list of accurate polls below as well)


PS I made this an OP - was in a different thread

edit to ad were Zogby is on the accuracy ratings... #17 LOL

[Drunken Irishman]

1. Zogby doesn't understand tracking polls are supposed to be taken in a broader context.

Not just a day's worth of data. :eyes:

[FreeState]

2. I think its a 3 day average - however 400 is way to small a sample for it n/t

[Drunken Irishman]

4. I know...that's not what I'm saying.

I'm talking about changing the tone of the race based on a daily sample.

Over 3 days, Obama has a pretty solid lead.

[FreeState]

5. Oh got you - sorry miss read your comment - must be getting late :) n/t

[Drunken Irishman]

6. It's ok, I'm not very clear.

:P

[endthewar]

7. 400 isn't too small of a sample

Read this to help you understand polling statistics. :hi: http://www.washingtonmonthly.com/archives/individual/2004_08/004536.php

[FreeState]

8. Um - from your link

"Generally speaking, national polls use sample sizes of about 1,100, which translates to an MOE of 3%. State polls often use a sample of 600, which produces an MOE of 4%. Subsets of polls sometimes have MOEs of 5% or higher."

SO the MOE is upwards of 5% and the two candidates are 4% appart in it - so its to small of a sample size to gain any credible data from it. Obama could be up +9 with the MOE in the poll.

Edit:

From Zogby:

The poll was conducted by live telephone operators in Zogby’s in-house call center in Upstate New York and carries a margin of error of +/- 2.9 percentage points.

The poller accuracy listed above ads nearly 2.16% to the poll = 5+ MOE.

[endthewar]

9. deadheat w.r.t. margin of error = misleading

The REAL question from a statistical standpoint is, "What is the percent likelihood that Obama is TRULY leading McCain?" You can use the table in that link to answer that question. The table is just tabulated data from a z-score (boring statistics stuff). The margin of error is determined exclusively by the voter sample, nothing else. n=384 corresponds to a standard error difference of 5%. Using the table in the link, there is a 79% chance that Obama is ahead (4% lead with sigma=5%).

It is statistical malpractice whenever someone says something like, "The lead is 4%, but with a 5% margin of error, Obama might be up by as much as 9% or down by 1%." Anyone saying that has never calculated a z-score before, which is normally done in intro to statistics classes.

Also, every pollster is using different party sampling percentages. People are just guessing how many Republicans vs. Democrats will be voting. The important statistic to look at is independent voters.

[regnaD kciN]

3. Zogby is so embarrassed over his "guaranteeing" a Kerry win...

...on Election Day four years ago, he's got his poll set up to make sure there's no chance of that happening again -- accuracy be damned.

[DCBob]

10. These are minor daily blips in polls. The trends are all good for Obama.

In fact it is probably better to focus on state polls now -- there are plenty of them now. Remember "all politics is local" and Obama is playing the local game much better than McCain.

[regnaD kciN]

11. You might find this chart interesting...

It's from Pollster.com, and plots the margin of each tracking poll versus the overall trendline.

Several points can be made. First of all, Gallup (green) and Rasumssen (pink) seem to hold the closest to the overall trendline. dKos/R2K (red) seems, as I've noted before, to consistently skew in Obama's direction by about 3%-4%. Diageo/Hotline (light blue) has generally stayed pretty close, but with a few days where it was way out of line with everyone else, generally in a favorable-to-McCain direction. GW/Battleground (dark blue) has skewed consistently toward McCain, almost a mirror-image of the way dKos/R2K has favored Obama.

And see those black dots on the right side of the chart? Those are Zogby. Not even close...

[Drunken Irishman]

12. Zogby is bad, but regardless, the trend is clear.

Obama gonna kick some ass on Nov. 4th! ;)

[lessthanjake]

13. The weighting is screwy

If Obama has a 17% lead amongst independents, that would yield a 5.1% lead nationally, assuming independents make up about 30% of the electorate. McCain gets 88% of republicans, whereas Obama is getting 86% of democrats. Assuming his poll finds McCains support amongst democrats as 2% higher than Obamas support amongst republicans as well, this slight difference would actually result in McCain making up 1.4% overall if the remaining 70% of the electorate were split equally between democrats and republicans. This would mean that with equal dem/rep weighting, Obama would be ahead by 3.7%. The poll has him ahead by 4.1%. This means that the weighting is probably about 36% Democrats, 34% Republicans, or maybe Zogby weighted Democrats and Republicans equally but had independents weighted at higher than 30%. I don't know exactly, but you get the gist.

This is really important because Zogby weights by party identification. He is not just going off of what percent of respondents say they are of a certain party. He weights it to meet a certain percent of each party. However, polls such as Gallup that do not weight by party find that there is now a large gap in party identification, with democrats getting about 40% and republicans about 30%. This makes sense because studies have shown that people frequently change their party identification to fit the current political climate or to fit with who they are voting for.

Therefore, it seems faulty to weight by party in the manner Zogby does it.

Just to illustrate this, if we assume that Democrats are 40% of the electorate, and Republicans 30%, and that McCain got 88% of Republicans, Obama got 86% of Democrats, and Obama won independents by 17% we would find a different result. In fact, you'd find Obama winning the election by 11-12 percent (I dont have the exact information on their cross-party support so I can't give an exact number).

The fact is that if Obama is winning independents by 17%, and is pulling a similar percent of Democrats as McCain is pulling Republicans, then Obama will win in a landslide because there are way more people who identify themselves as Democrats now than there are those who identify themselves as Republican.