[USFRIENDINVICTORIA]

"* Romney's support drops to within the margin of error of not existing (that's 3% support in a poll where the margin of error is 3%). "

This statement is a bit right -- but the margin of error on it is so large, that it's actually wrong. </smarmy sarcasm>

Let's assume that 1,000 people were polled (the exercpt doesn't make that clear) -- Romney got the support of 3% of those polled -- i.e. 30 people. Even if, by some incredible fluke, those were the only Romney supporters in existence -- they still exist. Therefore, it's not correct to say the support has dropped: "to within the margin of error of not existing".

If Romney had polled 4%, it would be fair to say that his support could be as low as 1% (provided you also added the confidence level. I.e. how many times out of 20 would a smaller result be expected?)

It's important to be picky on poll results -- they have become a political weapon.

[supercat]

Let's assume that 1,000 people were polled (the exercpt doesn't make that clear) -- Romney got the support of 3% of those polled -- i.e. 30 people. Even if, by some incredible fluke, those were the only Romney supporters in existence -- they still exist. Therefore, it's not correct to say the support has dropped: "to within the margin of error of not existing".

I expect that what's meant is not "total support", but rather "support above the nearest competitor". And that could easily be absorbed completely by the MOE.

[RobbyS]

Polls have become the equivalent of politic ads, or, to put them down even more, the equivalent of blubs on the back of a book.