Cathy O’Neil on EconTalk  
 
I really enjoyed last week’s EconTalk with Cathy O’Neil (who blogs at mathbabe.org/). The discussion was about math, hedgefunds, the financial crisis and Occupy Wall Street. While the host, Russ Roberts, is pretty hard-core-Chicago-School (his PhD advisor was Gary Becker!) and O’Neil is now working for the Occupy movement, they agreed on a surprising amount about what caused the 2008 crash and what good solutions might look like.
 
O’Neil made a few points that really resonated with me. First, she took a little shot at Nate Silver (who has argued that the financial crisis was the result of the challenges of modeling complex financial phenomena and the solution is to simply make better models). O’Neil pointed out that in finance (unlike baseball, poker, or other finite games) you don’t often get to see the realized outcome to know where you were right or not. In finite game statistics, we all find out (fairly unambiguously) who wins and therefore the incentives of the modeler are to be as accurate as possible to get a statistical edge. Finance, on the other hand, is a non-finite game embedded in a complex political system with lots of (potentially contradictory) pressures (also your model may directly modify the system inputs). In such a system, the incentives of modelers are not necessarily to be as accurate as possible. Modelers might have incentives to get a big bonus, to not get fired, to come up with the rating that a company has paid you to give them, etc...
 
Second, O’Neil argued that there are few incentives to acknowledge the (often huge) error bars on the estimates we get from financial models because modelers are not paid to provide conditional, equivocal answers. She elaborates this point on her blog here where she argues that there should be a macho way to say “I don’t know.” It strikes me that the use of the word macho hints at a possible relationship between gender and this problem... but that’s a thought for another day.
 
 
Thursday, February 14, 2013