A recent Yahoo News article asks the creative question of whether your name can predict your politics . It turns out, it can. On some level this isn’t terribly surprising – names vary based on gender, ethnicity, and region of the country, all of which predict electoral choices. On a different level, it is sort of cool that you can predict who a person is going to vote for based on the name. Regardless, I’d like to offer a word of praise, and a word of complaint about this article.
First, the praise – the author makes the source code for the analysis freely available. This is a wonderful move, as it allows others to delve into the methodology and determine whether there are errors, false assumptions etc. I think this should be standard for data mining analysis. So many numbers and stats are thrown about and we have no idea how they were arrived at. It’s nice to see somebody be so upfront about exactly what was done.
Now the complaint. “if you want to raise a Democratic son, name him Willie. Democrats expecting a daughter should go with Gwendolyn, the most pro-Obama girl’s name on the list.” This is a classic correlation vs. causation error. Willie’s are more likely to be Democrats but that doesn’t mean that being a Willie causes you to vote Democrat. It may be that people who lean republican prefer to go by Bill or Billy or Mac or Buddy. So, it’s not that being a Willie makes you a democrat, but rather that being a democrat makes you a Willie. Or maybe there’s a third variable that causes people to both be democrats and be Willies. Maybe it has to do with regional/demographic differences in names and political attitudes, for example.
The point is, that you can’t draw causal inferences from correlational data. Being a Willie doesn’t chain you to the democratic ticket, even though many people who share your name vote democrat, you are still a free Willie.
The data from this article are useful in helping candidates identify likely voters or campaign contributors, but it’s not necessarily useful for interventions by parents to try and influence the politics of their kids.
Funny enough, I was just looking at this guy’s data when you posted this. Seems to me that he has another problem: he’s data-mining without theory, and so it is impossible to even know for sure if his outliers are interesting, or if they are just caused by randomness. After all, if you look at 1000 names, you are likely going to find a couple names that are outliers in our direction or another–just by random happenstance. Now, if he had a theory to explain the outliers (after all, names are correlated with geography, ethnicity, race, and income, just to name a few things, none of which he controls for or addresses), that would be interesting. It would be even more interesting if he did control for those factors, and name was still an important component of voting. But as it is, I’m not even convinced he’s found an interesting correlation–and that’s before we get to the massive causality problem that you’ve so astutely indicated.
Unlike ordinary sun glasses, those from Eagle Eyes do not use dark lenses to indiscriminately block different colors. Eagle Eyes sunglasses try to mimic the lenses in the eyes of hawks and eagles to bring you the best possible clarity when you are wearing them. This doesn鎶?mean that it will look as if you have the eyes of an eagle, but it will certainly feel like it when you wear Eagle Eyes sun glasses. fake oakleys