The graph above comes from Rudy Espino. It displays the level of reported turnout (as a deviation from the overall level of turnout) for General Social Survey respondents grouped by astrological sign. The interviews were conducted in midterm years; the graph depicts respondents’ self-reported turnout in the previous presidential election.
This graph’s origins are explained in this tale from Rudy’s statistics course:
bq. In class, I walked through an example [of logit] from a single GSS survey to estimate the likelihood of turnout. The usual suspects were thrown on the right hand side – age, education, church participation, union, et cetera.
bq. Two of my students noticed a variable called “ZODIAC” and wanted to see it included as well. “Now, now,” I told them. “You do remember what I told you about theory and model specification? So why would you possibly use somebody’s astrological sign to predict their likelihood of voting?”
bq. These two students came up with a well-thought out statement. They expected that Virgos would, on average, have a higher likelihood of voting. It has nothing to do with some magical alignment of the stars. Rather, it has to do with the parents of most Virgos, they explained. The birth date of Virgos occurs right before the age cut-off in schools. These students went on to explain that some parents will plan to have their children around this time period because it gives them an opportunity to place their children into school at an earlier age. It is those type of parents who, having invested so much time already in planning their child’s birthdate, will continue to invest in that child in other ways – Baby Einstein, music lessons, MENSA puzzles, et cetera. So that the parents of Virgos will, on average, raise more civic-minded individuals. Not all parents of Virgos are like this; but, on average, more of them are like this.
bq. Their explanation sounded reasonable enough for me to create a dummy variable for Virgos and include it. I smiled before executing the command knowing that I would be able to prove to these students that something like turnout, which has been examined in such detail over the years, could not be significantly explained by something as trivial as an astrological sign; and I, of course, would lecture them yet again on the importance of existing theory.
This cliffhanger continues after the jump.
bq. To my surprise, the dummy variable for Virgo came out positive and significant. I re-examined the basic summary statistics over and over and was horrified to see that Virgos had such a high propensity for voting. Did my parents not love me as much because I was born a Gemini? Was I just an accident they had to put up with? Would I have won a MacArthur genius grant had I been planned?
bq. I ran back to my office and did some searching. I discovered some interesting literature that assured me that I must have encountered an aberration: Mock & Weisberg (1992) and Sigelman (1982). Of course, Sigelman would have already examined astrological symbols! [No link is available to Lee’s pathbreaking analysis of presidents’ astrological signs and their personality attributes (Presidential Studies Quarterly 12: 434-39). Suffice it to say that the article is so thorough and brilliant that no one has touched the topic since. — Ed.].
bq. I then downloaded the full time-series of GSS data and was reassured to find that there is nothing special about Virgos. In any particular cross-section you might find individuals under a particular sign with above average levels of voting but nothing meaningful across time. The chart shows this.
This is comforting news to me, as a “sentimental” Cancer possessing a “narrow patriotism” that naturally makes me a regular voter. Had it been true that Virgos were the more devoted citizens, my “inferiority complex” would have made me “sulky” for days.