Nicely correlating with our presentation of an upcoming appearance by Steven Levitt, of Freakonomics fame, is an interesting post on his blog of the same name. (Note: Registration is full and closed for Tuesday’s talk, so don’t even think about it.) The “Freakonomics” post discusses a recent survey of economists by Scott Adams on how they feel about the upcoming presidential election. (Note #2: What is the proper way to differently style and cite a book and a blog of the same name? I honestly have no idea.) The survey itself is very interesting and worthwhile to peruse, but my intention here is to point out one small but important error in the demographics part.
The survey and the related discussion both point out the easy fact that economists who list themselves as registered Republican or registered Democrat obviously favor the candidate of their party affiliation. But the discussion then goes to note that economists who are registered independent also favor Obama, by a fairly substantial margin. The problem with the party affiliation question is that it does not account for states that do not require, or even allow, voters to register by party. I often hear people in Missouri say while talking politics that they are “registered this-or-that’s.” I always want to say, and sometimes do say, “No, you’re not, we don’t register by party here.” A more accurate way to phrase the question is, “Do you consider yourself a Democrat, Republican, Indepedent, third-party supporter, or non-voter?”
How the hell can this affect the results? Well … if someone, for a poll or survey, asks me whether I am a registered member of a party, it is perfectly legitimate for me to say no. If they instead ask me whether I consider myself a member of a certain party, I would say yes, and any other answer would be lying. We can be sure that some of the responders to this question are from states like Missouri that do not register by party. As such, people from those states with strong political leanings may have very reasonably chosen “registered independent” as the best of the available options, even though they are not truly independents. With a polling universe heavy on people from acadamia, it is likely that many of those not-really-independents favored the obvious candidate.
The results of the poll are not my point. And the overall survey is very interesting. But doing the demographics in a way that assumes all people register by party, or else officially register as an independent, is erroneous.
