Friday, September 21, 2012

Good Polls Versus Bad Polls

As we get closer and closer to the election we are being inundated by polls and interpretation of what they mean.  But the big problem with these polls is that some of them are simply unreliable.  I was a political science major as an undergrad and one of the things I remember from a class on polling is that any major poll,as a general rule, with a margin of error greater than 3% is not a reliable poll.

Now there are some exceptions to this as a general rule.  If you were polling a small local election it might be hard to get a large enough sample to get a really good margin of error, and there are other times when it's not really hugely important.  If you were polling on whether people liked a new soft drink, for example, and 20% said they didn't but 65% said they did and the margin of error was 5%, that would probably be acceptable.

But in larger political polling it's not acceptable.  In a poll released yesterday on Obama/Romney in Florida, the margin of error was 5%.  5%?  That is a completely useless poll, and to be honest there is absolutely no reason why the press should have ever even reported on it, because they might as well as have just guessed.

Polling is a science and those who do it well are really good and should be trusted, but always pay attention to the margin of error.  If, as a general rule, is greater than 3% then the poll is not worth anything and you should keep looking.

No comments:

Post a Comment