Two polls of NH Pubs put up on the same day, and Jeb has twice as high a percentage in the other poll as this one. Assuming both polls are accurate, can that be within the margin of error? Muon2? For an individual candidate, with a poll have a 4.8% margin of error, can an individual candidate polling at 4%, really have a 5% chance of really being as high as 8.8%? Inquiring minds want to know. 4.8% of 4% of course, would be next to nothing.
That's not how margins of error work. The figure of a MoE is not in %, it's in percentage points, and it's given for scores of 50%. When the score decreases or increases, the margin of error decreases.
In this instance, the poll has a 4.8 pt MoE for scores of 50%.
For scores of 20% or 80%, the MoE would be around 3.8 pt.
And for scores of 5% or 95%, the MoE would be around 2.1 pt.
So for a score of 4%, the MoE should be about 2 pt.
So in this poll, a candidate polling at 4% would have a 95% chance of really being inside the 2-6% bracket.
If he's polling 8% in the other poll, with 600 LV polled and a 4.1 pt MoE for a 50% score, the MoE for his 8% would be around 2.2 pt, so he would have a 95% chance of really being inside the 5.8-10.2 bracket.
Safe to assume he's around 6 in reality.
Go to page 5 of
this pdf file, everything's there.