Wednesday, October 19, 2011

Assault of Thoughts - 10/19/2011

"Words ought to be a little wild, for they are the assault of thoughts on the unthinking" - JMK

- Ryan Murphy presents a hierarchy of evidence. I think he sets the bar far too high for "science" (only his two highest). I think this sort of high bar (he's not the only one that sets it there) comes from thinking too much about physics and probably too much about most philosophy of science too. If we cut out his #3, 4, 5, 6, and 7 we are eliminating a lot of what we colloquially think of as "science". The list is also clearly weighted towards the empirical potholes that economists hit on a regular basis (I don't think Ryan would deny this point). Also, I think the whole "black swan" thing has limited applications for many scientists, and is less relevant especially when you are not extrapolating into the future. Black swans and fat tails are far less relevant for a variety of elasticities and impacts that economists estimate, for example. That's not to say black swans couldn't show up there - it's just to say that we are consciously taking local effects and we think those are very useful to know, and that even if there are black swans in some circumstances it doesn't make the insights about the local effect any less valuable. This isn't to say he doesn't make good points. I like the way he introduces #3 a lot, for example.

- A great post from Krugman that's been making the rounds. What a great time to be entering the profession.

- Haven't listened to it yet, but this is the new EconTalk. It's on Keynes and Hayek.

2 comments:

  1. Elasticities are exactly the type of thing that's prone to black swans. You fit a curve to the data given small changes. If you extrapolate from that point, you can (and do) get very misleading results because the underlying relationship between the variables fundamentally changes. See this graph (which is roughly the same as the one found in the book) for intuition: http://attaincapital.files.wordpress.com/2010/11/20091130_fig1.png

    ReplyDelete
  2. increasingmu -
    Right, but my point is that it's the extrapolation that introduces the problem - not the estimation.

    That's why I'm making the point that a lot of work we do is estimating local effects. If we remain cognizant of that (as most are, I think) it's quite legitimate.

    ReplyDelete

All anonymous comments will be deleted. Consistent pseudonyms are fine.