Accurate surveying is a gold standard in whatever discipline, and that includes politics, unless "the pulse of the public" really doesn't matter, in which case you're probably not as high up in office as you'd thought.
Philosophical question: but how do you determine accuracy if all you have are competing surveyors, isn't that a chicken-egg conundrum? Sometimes you just have to shoot the philosopher (just kidding -- they used to say that about engineers too). Actually there's a better answer: reputable companies tend to reach very similar, if not precisely the same results, i.e. lots of cross-checking and fine tuning goes on (competing surveyors is a good thing).
Social networking software is providing students of public opinion with new tools, little ships in a bottle able to model ships of state, other big enterprises, fairly well, or at least that's what's being advertised, as in: "punch in your data, customize the picture, and the generic power of our model will give you some most interesting predictions, and this ain't just hocus pocus neither."
A side benefit of all these new pools of polling data is it's increasingly hard for fly by night shops to simply make stats out of whole cloth. The bad apples in the bunch become more apparent, as social networking accelerates the reality checking. That's actually how democracy is supposed to work, so don't take this as a report on some crisis. Sit back, relax, and make sure you keep voting (your thinking does matter).