I agree. Surveys may do a decent work when they emulate simple behavior i.e. ask the Rs to do in a survey what they would be doing in real life (e.g. vote, or pick an item) - but they are useless to gauge opinions. Their main problem is that the make "heroic" assumption about cognitive processes, chief among them being that views remain relatively constant and context independent, and that opinions form a linear continuum (i.e. the distance between strongly agree and moderately agree is more or less the same as that between moderately agree and moderately disagree), or perhaps curvilinear is you are a utility function worshipper, rather than discrete "chunks" (e.g. various forms of 'agree' forming a 'chunk that is qualitatively different from other types of opinion such as disapproval or using an altogether different evaluative scheme). I think both assumption are demonstrably false in respect to cognitive processes.
The reason why surveys are so popular is not because they give us a reasonably accurate assessment - which they do not - but because they are a relatively inexpensive magic ritual that gives us an illusion of knowledge and control in situations that are ambiguous, unpredictable and difficult to measure. They are modern version of the magic rites performed by "primitive" societies to "control" volcanic eruptions, make rain, or ascertain safe return of the fishing boats from the high seas.
While it is possible to assess people's opinions and understand mechanisms that govern their changes in different directions, the methodology to gather that information is quite expensive and difficult to implement (c.f. focus groups or what Touraine called 'sociological intervention which can be thought of as a hybrid between focus group and the Delphi method). Surveys, otoh, are relatively cheap and easy to implement - they turn out results quickly and in the form that is easy to understand by "busy executives" (translation: dolts in authority positions who would like to think they know what they are doing but are too stupid and lazy to read anything longer than a paragraph). Hence the ubiquity of surveys.
As to attitudes toward the Bush administration - a great deal of them are not rational and thus do not follow any form of "utility function" i.e. preferences changing in response to perceived utility derived from policies pursued by the administration. Most people tend to "support" Bush as a part of their emotional reaction to social events - e.g. as a part of vicarious "ass kicking" of unpopular minorities. This is why certain people love Rush even though they know he is a fat slob and a dope head (i.e. a member of otherwise "despicable" to them category of people) and they love Bush even though they know that he is lying to them and screwing them up big time. But these rational utility considerations are trumped by emotional gratification they derive from the virtual lynching of the minorities they despise, which Rush or Bush do for them. So these people will be supporting right wing trolls regardless of the damage they inflict on the country and the economy.
Wojtek