"Only main difference" is an odd phrase in polling. Every nuance of every question liicits a particular bias in the respondent as well as the tone of the voice of the interviewer, time of day, sequence of questions and just about anything else that one may care to cite.
MSNBC Polls will general illicite a more liberal response from the whole universe than will FOX News. Are we to believe that the poll from each actually always taps in to only the group more sympathetic to the famous political leaning of either or both? It is more likely that the biases of the pollers are consistant than the are the people polled.
For this reason all polls are valuable as they are what they are. It is the people who review each of the polls that need to be aware of the biases. Averaging the polls does not average the opinion. It averages the bias. Averaging the bias is worthless. Read the polls individually and know that that the biases exist.
Read the trends exposed by the polls. If Obama's popularity changes by 3 points in all polls regardless of the level reflected by the individual polls, something is happening. Measure the change, not the level. This is true whether the change is negative, like August, or positive, like September.
Reading a poll and seeing only that which supports your own bias is myopic.
Look, there are a lot of polling organizations out there. I am not saying that Rasmussen changes it's numbers, what I am saying is that, as per their own site, which was directly quoted earlier in this thread multiple times, their methodology leads to higher negative numbers, in general.
In addition, that methodology is not used by any other polling agency, and most importantly,
has not been used historically by any other polling agency.
This makes Rasmussen an outlier, by their own admission, from all the other polls.
As I posted earlier, RCP has been proven to provide more reliable numbers, for whatever reason, than Rasmussen.
I'm not claiming Rasmussen, or anyone else for that matter is actually changing numbers.
What I'm saying, is if you
want the numbers to actually mean anything, you must be able to compare them to other historical numbers in similar situations, like Reagan in 1981 for instance.
And if you are goinig to do that with any degree of reliability, you must use the same methodology,
or else any such comparisons become meaningless, making the numbers themselves meaningless.