Letters to the Editor
-
Margins of error, and polling results in the news
I'll spare everyone a detailed statistics lesson (not that I'm entirely qualified to give one) but the 'margin of error' is actually much more complicated than any news outlet ever explains. JeffM, I don't think your explanation was quite correct either, but your point was still valid.
The margin of error (a range of error from the reported statistic) also has an associated level of confidence (90%, 95%, 99%) which is seldom reported. If there's a +/- 2.4% margin of error with 95% confidence, that means we would be 95% sure that the ACTUAL number (in the "real world") is within 2.4 percentage points of the reported statistic from the surveyed sample. Or in other words, there's a 5% chance that the real number is outside the margin of error.
Furthermore, the margin of error our news media reports is usually the 'maximum margin of error' which usually applies to statistics that are in the midrange (around 50%). With a higher or lower statistic (closer to 0% or 100%), the margin of error would be less than is typically reported.
Even more important is that these analyses assume a perfect random sample for a survey or experiment, which I think is a pretty risky assumption these days given the inherent bias in any survey methodology. So it's quite possible that in any given study, the survey bias could produce more error than random variation.
The bottom line is this: the survey statistics we usually hear about in the news aren't explained as well as they should be; the results aren't as simple as they seem, and they're being *dumbed down* for mass consumption.
And my opinion is this: survey results get WAY, WAY TOO MUCH coverage in our political news. I fear people's opinions are being swayed too much by what they hear that other people are thinking via polls.