Republicans talking about skewed polls were the most egregious example, but nonpartisan media was generally calling the election “razor tight”, “a tossup” and similar things, too. In their case, the reason seems to be an ignorance of how statistics works. E.g. seeing polls with Obama up by 2% and a margin of error of 3, they would label it “a statistical tie”, even though a) even with a single such poll, it implies a much higher chance of Obama winning, and b) with many polls giving numbers in that range, the chances of Romney being actually ahead drop to near-zero, barring systematic error.
True, also the media will tend to exaggerate the tightness of any race to make their news more exciting. Who will say up until the wee hours of the morning watching commercials and news, if the outcome is certain?
The normal margin of error on a political opinion poll would be 1.96 sigma—a 95% confidence interval (that’s how you’d get a margin of error of just over 3 percentage points on a poll of 1000 people.
Republicans talking about skewed polls were the most egregious example, but nonpartisan media was generally calling the election “razor tight”, “a tossup” and similar things, too. In their case, the reason seems to be an ignorance of how statistics works. E.g. seeing polls with Obama up by 2% and a margin of error of 3, they would label it “a statistical tie”, even though a) even with a single such poll, it implies a much higher chance of Obama winning, and b) with many polls giving numbers in that range, the chances of Romney being actually ahead drop to near-zero, barring systematic error.
True, also the media will tend to exaggerate the tightness of any race to make their news more exciting. Who will say up until the wee hours of the morning watching commercials and news, if the outcome is certain?
Assuming the “margin of error” is one sigma, that’s a 75% probability of Obama winning, which hardly qualifies as “much higher” IMO.
EDIT: Retracted. If, as James_K says, the margin of error is 1.96 sigma, that’s a 90% probability for Obama.
The normal margin of error on a political opinion poll would be 1.96 sigma—a 95% confidence interval (that’s how you’d get a margin of error of just over 3 percentage points on a poll of 1000 people.