Marianne Bertrand and Sendhil Mullianathan contemplate (pdf) the limits of surveys:
Simply because the surveyor is asking the question, respondents believe that they should have an opinion about it. For example, researchers have shown that large minorities would respond to questions about obscure or even fictitious issues, such as providing opinions on countries that don’t exist.
Robin Hanson reflects on this finding:
I’m not clear on just how far this effect goes, but one lesson is: you have fewer real opinions than you think. If you talk a lot, you probably end up expressing many opinions on many topics. But much, perhaps most, of that you just make up on the fly. You won’t give the same opinion later if the subject comes up again, and your opinion probably won’t effect your non-talk decisions.