Friday, September 12, 2008

Correlation Not Quite the Same as Causation

I've talked before about the silliness of comparing conditions under Republicans versus Democrats and pronouncing one to be better.  After all, you have a sample size of around six, and without a causal pathway between the President and tangible policies, you're looking at noise.  

In interest of balance, I should point out that a University of Chicago professor, of all people, goes down this route in a WSJ article arguing that Republicans reduce the wage gap.  Granted, he's also written a paper which I haven't read, which may have more substance.  But the editorial is absolutely horrible.  There's no discussion of any policies that could drive this trend or even a thought given to the possibility that we have a Legislature which passes laws and partisan identity might matter there as well.  

This is one example of why data isn't everything.  Granted, small sample size is wildly skewing these results, and there might be something behind a more robust finding across, say, conservative parties worldwide.  Or it could be noise, or driven by some other factor (maybe people prefer a party in times of plenty; maybe they have an anti-incumbency bias in bad times).  Theory isn't dead, and that's something you might expect a Chicago professor to know.  

No comments: