Polls - take with a grain of salt - or skepticism

Posted on January 26, 2008

One of the first thing my favorite prof in college taught us in Statistics 101 was don’t take seriously a survey with a small sample size.

There is a poll of South Carolina ‘voters’ which on its face looks like 53% of Hispanics are in favor of Obama and 32% for Hillary.

Now most polls use a sample of about 1000 minimum. This one interviewed 2100. Ok. But then it determined only about 1900 were actually registered to vote.

Then it determined only some 685 were “likely” to vote in the primary today. Getting low. I would doubt any results from this survey just on that basis.

Go further - only 1% identified themselves as Hispanic. Now 1% with no decimal should be something between 0.51% and 1.49%. So it is possible that as few as 4 Hispanics were actually included in the results. Working backwards from the stated results it seems that 23 Hispanics were included in the final results. (Either they mis stated the % of Hispanics in the final survey or the 1% was on the original sample of 2100).

So of the 23 people - 12 said Obama, 4 Hillary, 5 said someone else and 2 said Edwards. Hardly a convincing statement on what the thousands of Hispanics will do in SC today. Nor is it any indication what Hispanics elsewhere will do.

I have trouble with trusting any political poll but certainly one reporting results based on minuscule sample sizes is really of no value.

» Filed Under dirty tricks, Democrats, general nonsense, Hillary, National Politics