Saturday, August 24, 2013

Two interesting observations on this economath discussion

Noath Smith, Paul Krugman, Bryan Caplan, David Henderson (and me, but I'm a bit player) have been talking about math in economics lately. I'll let you look up all the posts.

I've noticed two funny things.
1. The people who are generally opposed seem to treat the argument as economic intuition/relevance vs. mathematical economics. This is a category error. Economic intuition/relevance is something we expect to be in all good economics, whether it is literary or mathematical. And there's definitely examples of both with and without good relevance and good foundational economic principles. That's the substance of a piece of economics. The math/literary question is the medium. The real question is the comparative strength of the medium.

2. If find it interesting that there is a fight about math's role in economics but there seems to be little fight at all about the role of literary economics. In other words you've got people talking as if the highly mathematical papers have somehow gone off the rails but you don't have anyone criticizing the major literary contributions. I find that interesting.
My view is that both are important and we should expect people going through the PhD process to be able to write and do math well and be excellent consumers of both. I am not worried that good people are kept out because of math standards. There is a substantial enough demand for PhD slots that there's no reason why programs can't be choosy and get people good at both. So I don't think we should lower our math expectations - if anything we should raise our literary expectations. This is coming from a guy whose writing is certainly stronger than his math and for whom that certainly made a difference in his PhD application experience, so I'm not giving math expectations a pass because I'm some kind of hot shot.


  1. I think this, from Tim Johnson, is by far the best contribution so far (& Noah Smith agrees):

  2. Agreed, Daniel. I actually can't even seem to figure out what people even really mean by 'economath.' Separating hyperplanes? Statistical theory? Differential equations? Carried over from blog to blog, I don't think that anyone has been wrestling with the same definition.

    Also, as with many things, Krugman wrote about this much, much better back in 1996:

  3. The Krugman article is pretty good, but algebra vs literary is pretty much a slam dunk for algebra. Many making the complaint against economath (or something like it) are using sound "mathematical reasoning" and actually catching most of the errors that are easily disproven. The typical complaints are the Feynman "fooling yourself", over-aggregation, drawing too big a conclusion from empirical data that contradicts sound reasoning, models that are too complex, excess explaining away of data, conflating statistical significance with relevance. Complaints found in Hayek's Pretense of Knowledge, Bowerk's One Good Model.

    One problem is that intuition needs to be deeper than surface level. Deep intuition that challenges your own prejudices is necessary.

    DK- " So I don't think we should lower our math expectations - if anything we should raise our literary expectations."

    I agree with this, as long as it's logical (vs creative) literary skills you are screening for.

  4. Do you remember this discussion we had a while ago, Daniel Kuehn?

  5. I don't think the problem so much is "math" per se as opposed to "useless math that adds nothing of value, obscures whatever actual insight is in the model itself, and only acts as a signaling device". The last one is probably the aspect that people get annoyed with the most.

    I've seen papers in their "Working Paper" version which develop a pretty sensible and straight forward model, where the intuition about the phenomenon in question takes center stage and is transparent. And then I've seen the actually published version of the same Working Paper where it's pretty obvious that either the editor of a particular journal or the referees made the authors complicate it unnecessarily to fit it into the "how modern economics is done" mold. I'm most aware of it in macro but I don't think micro gets off easy either. And no, I'm not speaking of personal experience, just the experience of watching the kind of papers that get published in top journals over the years.

    Simple example - why do an OLG model if there's no integenerational transfers? Why assume weird functional forms for utility and labor supply when it has no effect on the results? There's a lot of stuff like that in the papers once you sit down and work your way through them. Incidentally, this is actually a good bit of the reason why Paul Krugman's "most of it's already in the IS/LM" model is true.


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