A "Means" to an End

Last night I heard a caller to a radio talk show (forgot the name of it) say to a political pollster that his research is flawed because he only talked to 500 people. How can he project the results to the entire U.S. population, the caller asked. The pollster assured him that this is statistically valid.

But I beg to disagree. Even a sample of 1,000 carries a 3% error margin for a U.S. population of 250 million!