Friday, July 18, 2008

Polls don't matter

A new poll, that has been reported on by all the major news organizations, says that over 60% of white Americans have a negative view of Obama. We all know to get an accurate poll, the poll must have a large enough sampling of the population to properly represent the populous. So how many people were polled: a million, 100 thousand, 10 thousand? The actually polled less than 2,000 people. There is no way to get a legitimate poll with this small of a sampling. Depending on econmic levels, education levels, and geographic area of pollees can make a huge difference. However polling just a few of each group is simply ridiculous. People wonder why polls keep being wrong. This is why.
Posted by at
Labels: , , ,

1 comment:

John Snowden said...

As far as satisticians and political pollsters care...the statistical need for accurate sampling for polling is a random sampling of 1,500 constituents which yields a +/- of 3%.

But I think most polls which report a +/- 3% variance would be more appropriately labeled as +/- 30%. But that's just me.

August 28, 2008 at 12:22 AM

Post a Comment

Subscribe to: Post Comments (Atom)