Polls and polling are an important part of politics, and one I don't know much about. To remedy this situation, and so I can write on the topic without flaunting my ignorance, I've been reading the 6th edition of Polling and the Public: What Every Citizen Should Know by Herbert Asher. Take a look at this passage from p. 53:
For example, in surveys in which the wife and husband were interviewed independently, their responses did not agree perfectly about such factual items as the number of children they had. Perhaps errors were made in transcribing their responses. Or perhaps the question was ambiguous. One spouse might have responded in terms of children living at home; the other in terms of the total number. Or one on spouse might have included children from a previous marriage, and the other might not have.
Something to make you think next time you read poll numbers.