I’m not generally a political person, but one of my pet peeves is when people, either intentionally or unintentionally, draw conclusions from statistics that the numbers simply won’t support. Lately the seeming obsession in the media with national polls in the U.S. presidential race has been driving me up a wall. This poll says McCain is up by 10 points. That poll says Obama is up by 1 point. Apart from the obvious problem of wildly divergent results, what no one seems to realize is that national polls are absolutely meaningless in a U.S. presidential election.
That’s because the president is elected by the Electoral College, not by national popular vote. Each state is allotted a certain number of electoral votes, and who gets those votes is (usually) determined by who wins the popular vote within each state. A national poll, while it may indicate who is ahead in the popular vote, provides absolutely no information at all as to the number of electoral votes a particular candidate is likely to receive.
The only way to get even remotely useful numbers is to poll state by state, determining who is ahead in each state and then assigning each state’s alloted number of electoral votes accordingly. That’s what this site does. Given how pervasive this stuff is and how easy it is to manipulate, I’m beginning to think statistics should be a required course for everyone in school, or may just for freak-out-prone journalists.