You know, I was never the best statistics student. Math's just not my area. But I have enough basic competency so I would never be stupid enough to think a large, non-random internet poll is more accurate than a smaller, random-sample scientific poll. Which, unfortunately, is more than we can say for Rep. Daniel Webster (R-FL).
Webster is pressing to eliminate the American Community Survey (a critical part of the Census), in part because it is "too intrusive", in part because it's "non-scientific". Why is it non-scientific? Because, to quote Webster, it's "random"! The New York Times can barely keep it together as it proceeds to inform its readers "[i]n fact, the randomness of the survey is precisely what makes the survey scientific, statistical experts say."
Random sampling is what makes the survey world go round. It is the only way to cost-effectively gather data about large groups of people given the impracticability of querying all of them. This is Statistics 101.