Wednesday, September 08, 2010

ed morrissey never heard of "margin of error"

the obvious answer is not a left-leaning media trying to find a poll to claim GOP momentum is reversed, but rather that last week's generic ballot poll showing a 10-point GOP advantage and this week's generic ballot poll showing a tie are probably both statistical outliers. the "real" generic ballot result is probably a GOP advantage of 4-6%.

it really is remarkable how many people who regularly watch the polls completely disregard that bit at the bottom that says "For results based on the total sample of registered voters, one can say with 95% confidence that the maximum margin of sampling error is ±4 percentage points." it's not just right-wingers who make this mistake.

and that's on top of the fact that the "generic ballot" question itself is pretty meaningless. generic ballot questions are like the iowa straw poll. they are one of those things that political reporters pay attention to, which fools political junkies into thinking generic ballot polls mean something about an upcoming election, when there is little evidence that they predict much at all.