This got a chuckle out of me, as I wade through data every day and every model, experiment, trade study, AoA, etc. has a raft of assumptions & limitations.
http://gregmankiw.blogspot.com/2007/10/data-free-empiricism.htmlTuesday, October 23, 2007
Data-free Empirics
Recently, I was in downtown Wellesley, in my favorite local bookstore, Starbucks latte in hand, browsing through the new offerings, when I stumbled upon Robert Reich's new book, Supercapitalism. As I stood in the store, I did what every narcissistic Harvard professor (yes, I am being redundant) does: I looked myself up in the index.
To my surprise, I found myself quoted in the book as an authority on the fact that high tax rates are not distortionary. Here is the quotation from me that Reich uses in a footnote:
"CEOs are paid what they are worth to their companies, and their high pay reflects the extraordinary value of their talent. But the supply of talent is inelastic, and the allocation of talent would not be affected if everyone faced high tax rates."
The quotation is taken as support for Reich's broader argument that we should increase taxes on high incomes. He mentions that the quotation is from this blog, but he also cites a David Wessel article from the Wall Street Journal as a source.
Do I really believe that the supply of talented labor is completely inelastic? No, I don't. Let's look at the fuller quotation from my blog entry:
"In his model, high CEO salaries are pure economic rents. CEOs are paid what they are worth to their companies, and their high pay reflects the extraordinary value of their talent, but the supply of talent is inelastic, and the allocation of talent would not be affected if everyone faced high tax rates."
Notice the key phrase "in his model." That is, I was not describing any evidence on the question, nor was I even giving my opinion. Rather, I was describing a particular model presented at a Harvard seminar by Xavier Gabaix on his joint work with Augustin Landier.
In Reich's defense, Wessel's article does not include the key qualifier either. Wessel's article does note, however, that "Messrs. Gabaix and Landier shudder at this suggestion [of higher tax rates]." Reich omitted that part of the Wessel article from his footnote. A reasonable interpretation is that Gabaix and Landier made the assumption of inelasticly supplied talent to keep their model simple, not because they really believe it.
In short, here is what appears to have happened:
1. Gabaix and Landier make a modelling assumption for purposes of analytic convenience.
2. I describe their model and its implications on this blog.
3. Wessel quotes part of that description in the Journal.
4. Reich reads the Journal and cites me as an authority using the partial quotation.
5. As a result, a modelling assumption morphs into an established fact.
You might call that data-free empirics.
Some similar model assumptions that, if taken out of the model & experiment context, lead to absurdity:
1. Flat earth model for small areas of terrain.
2. Infantry move at max rate possible for that terrain type, slope, vegetation, etc.
3. Communications between units is perfect.
4. Laser range finders have no effect on direct fire weapon effectiveness at ranges less than 300m.
5. Weather is constant during game time
6. Unit X will not engage the enemy at ranges greater than 100m
etc....
Each assumption has a valid reason behind it. This is one of those cases where reading the results is not enough. The reader must understand why they were used and the bounds of the trade space after which the model/results are no longer valid.
