Umpires and racial bias (encore presentation)
A few people have written me about this Freakonomics post, from Daniel Hamermesh, the co-author of a just-published study that found racial bias among MLB umpires.
That study is an update of the paper that came out a few years back. The basic data is the same, but the authors have updated the discussion and added a section with PitchFX data.
I haven't gone through the new stuff in detail -- I'll have to do that after the SABR convention, when I have more time. However, the basic data appear to be the same as in the previous version. So the comments I made about the paper back then still stand:
1. The authors of the paper explicitly assume, in their model, that every umpire has the same racial bias in favor of his own race. I think that assumption is unwarranted. I think it's much more likely that umpires have individually different levels of bias, just like everyone in the real world.
2. If you relax the "every umpire is the same" restriction, it's possible that there is much less racial bias than the authors found. For example, if you remove one single umpire from the study, the one whose calls favored his own race the most, the results are no longer statistically significant.
3. Therefore, while I agree that the authors found the results to be statistically significantly different from "no bias at all," I believe the conclusion -- that racial bias is pervasive among umpires -- is not supported by the data.
If you want to see my reasoning in detail, go to my website and look for the first entry under "Research". There's a PowerPoint presentation, an article (.PDF), and a series of nine blog posts. I recommend the article first, and the ninth blog post.