Unlikely Voter

Tweet

Look, I get that we’re way out from the 2012 elections. The dearth of commentary and analysis here at UnlikelyVoter.com is not entirely due to weeks lost to a horrible cold I picked up in our nation’s capital. But hey, Muhlenberg College and Morning Call: If you’re going to do a poll of the 2012 elections, at least try?

Having a poll out for 19 days isn’t the worst thing in the world when nobody’s actually campaigning, but at the same time it still leaves us unclear just what is being measured. It’s some indeterminate mixture of opinions over that nearly three week period, which I’m still guessing is simply the sum of the phone calls made by students on the weekends over those three weeks. That’s problem one with the poll.

Problem two is that the poll is not sampling properly. The poll notes that cell phones are included in the poll. That is flatly the wrong thing to do. Cell phones have strong overlap with residential lines, as many people have both. If you’re not accounting for that and treating cell phones separately, making sure to ensure people with two lines are not oversampled, your poll is now biased. However the given survey does not give any indication of that, nor does the documentation, so I have to conclude the poll oversamples cell phone owners.

Problem three is the plain old sample size. At 395 registered voters our sample size goes up to 5.5%, even if we ignore the above problems. So when I first heard about this poll, my generous conclusion was that Mulhenberg College did the Kessel Run and at least kept the real MoE (taking into account all the issues) under 12 parsecs.