Just because you poll a lot of people doesn't mean the poll is any good.
Remember, the quoted margin of error is just the statistical error, not the systematic error.
I would love to know where this idea came from the larger = more accurate.
In 2006, if was at 1,000 or so, everyone always said it was an indication that it was of less quality then polls with 300-500. The same way people claim the use of decimal points indicates it is of less quality. They are playing a game and trying to make you think they are better then they actually are by giving a large sample or in case of decimals by giving impossible levels of precision (hundredths, thousandts, etc). In doing so, they ironically communicate their lack of reliability.
