Are the Cook Polls biased by design?

[SoCalDemocrat]

Lets talk about this poll and the conclusion below which they somehow derive from the poll:

"For example, an August AP-Ipsos poll had 50 percent self-described Democrats or those who leaned Democratic and 44 percent Republican or those leaning that way. In the September AP-Ipsos poll, 50 percent were Republican or leaned toward the GOP, and 43 percent were inclined to call themselves Democrats."

First, how do we guage whether the electorate is shifting one way or another by use of this poll? It doesn't poll the same number of Republicans and Democrats each month. By their methodology, all they do is make sure they poll approximately 788 registered voters among the total 1000 persons polled. That means the number they get from each party is totally random?

So, if they polled 33% Republicans in August, and 40% Republicans in September, their polling would give the appearance of a shift towards Republicans. The fact that they don't measure the political party of the persons polled renders use of these results in this manner unreliable.

The only way we could expect to draw such a conclusion would be if each and every month they consistently poll the same percentage of Republicans, Democrats, and Independents/Other. Then we could measure the leanings as this reporter and the poll insinuate.

Credible polls, such as employed by Gallup, state the exact mix of participants and apply those metrics consistently.

If the poll does not disclose the percentage of registered voters sampled, we cannot use it to draw any meaningful conclusions. This month they got lucky and hit 500 Republicans in their polling, bu tnext month they'll hit 512 Democrats and it will appear to skew badly the other direction.

Another aspect of this poll which I disagree with is the "pushing". Their methodology implies pushing undecided persons who are polled to make a stand on their position one way or another. That makes it entirely subjective and subject to influence based upon who is doing the pushing and how they're doing it. A poll should be completely objective and neutral in its language and application. This one would appear to fail on that count as well.


http://www.cookpolitical.com/poll/default.php

Twice monthly, Ipsos Public Affairs in conjunction with the Cook Political Report conducts a national survey of 1,000 adults (of whom approximately 788 are registered voters) on a variety of political and economic questions. Question 1 (right direction/wrong track) reports the results of the full sample of adults while the political questions (presidential job approval, presidential re-elect and generic ballot test) are just of registered voters, unless otherwise noted. For adults samples, individual samples of 1,000 adults have a margin of error of +/-3.1 percent, all with a confidence level of 95 percent, but also reported are data for two consecutive surveys, totaling 2,000 adults with a margin of error of +/-2.2 percent and quarterly data of 6,000 interviews with a margin of error of +/- 1.3 percent.

For political questions, the margin of error for individual surveys of approximately 788 registered voters is +/-3.6 percent, all with a confidence level of 95 percent. For two combined surveys with approximately 1,576 interviews with registered voters, the margin of error is +/-2.5 percent while for quarterly compilation of data with approximately 4,727 registered voters, the margin of error is +/- 1.5 percent.

Some questions report a "mean" result as well. In the political questions, the mean is the average response for all those interviewed with a value of 7 given to those who strongly approve of President Bush’s performance, a six is one who somewhat approves, a five for those who initially are not sure or have mixed feelings but when pushed say that they lean towards approval, a four to those who are not sure or have mixed feelings and do not lean either way, a three for those who are initially not sure or mixed but lean toward disapproval, a two for those who somewhat disapprove and a one for those who strongly disapprove.

[tritsofme]

1. Gallup does not weight its polls for party idenification

They use a random sample, and they do not reweight partisian numbers.

Pollsters like Rasmussen and Zogby do.

[Teaser]

2. They do, however, "quota fill"

Or so I have been lead to believe.

[tritsofme]

4. They do weight for factors they can check against census figures IIRC,

but I believe that is the extent.

[SoCalDemocrat]

3. bogus polling methods

Based on random sample, how in the world do they justify drawing and printing such conclusions?

1. They have no idea who they polled each month
2. Their poll includes use of push tactics to elicit a response in favor of Bush vs. Kerry

I don't find any credible statistical information to be drawn from such a poll. Can anyone help out here? What exactly is this poll telling us?

The only assertions you can make from such a poll might be:

X percent of registered voters think that Bush is doing a bad job.

Drawing comparisons between parties based on such polling is inaccurate at best, and at worst misleading and fraudulent.

[tritsofme]

5. Their rationale is that party ID can be fluid and is not set in stone

and that using predetermined partisian weightings could miss shifts from one party to another as some people choose to identify with the people who are currently "ahead."

I don't how you consider this to be "push polling"

Now, suppose that the presidential election were being held today, and it included John Kerry and John Edwards as the Democratic candidates and George W. Bush and Dick Cheney as the Republican candidates. Would you vote for – ?

[SoCalDemocrat]

6. fluid?

A persons party registration is not fluid. You're registered to a party or you're Independent. That fact is used by reputable polling agencies to make more accurate predictions.

It's pushing when you have the pollster pressing the person on the other line who is sitting on the fence to make a decision one way or another. How do we know that the individual pollster doesn't have an agenda? How do we know they're not trying to illicit a particular response. It biases the poll to have the pollster attempting to influence someones opinion. The Cook poll has that built into their survey process!

"Are your really sure you don't like Bush? He's done a lot for our country!"