Saturday, September 14, 2013

The economist's utility function

U = [(realism)^a]*[(predictive)^b]*[(explanatory)^c]*[(parsimony)^d]

For most economists there may be some idiosyncratic preferences here and there (Krugman seems to like parsimony, empiricists like explanatory power, people close to forecasting communities like predictive power, and those invested in pedagogical issues like parsimony), but there's no big winner that trumps the rest:

a =?= b =?= c =?= d

Hard core praxeologist types:

a >> b, c, d

Other Austrians:

a > b, c, d

The crude Friedman version/crude Popperians:

b >> a, c, d

Vaguely positivist types:

b, c > a, d

There are very real trade-offs between all four of these arguments in our utility function, whether you want to admit it or not.

15 comments:

  1. a = parsimony = d? Or is there a difference between the two parsimonies?

    ReplyDelete
    Replies
    1. My guess is that parsimony means simplicity, which is not the same thing as realism. I'd argue that "praxeologists" are: a > d > b,c. Most Austrians like to simply things, e.g. the structure of production, the Robinson Crusoe model, et cetera. Non-praxeologist Austrians are probably in between: a,d > b,c and d > a > b,c, because the idea of a synthetic a priori is dropped.

      Delete
    2. I didn't read carefully. Because he didn't give an example for realism ("Krugman seems to like parsimony, empiricists like explanatory power, people close to forecasting communities like predictive power, and those invested in pedagogical issues like parsimony") I somehow forgot about his inital definitions.

      Mea culpa.

      Delete
    3. Jonathan, a Robinson Crusoe economy is a very different kind of simplification from the kind used by mainstream economists. It doesn't distort anything about the phenomena it is supposed to throw light on. See Roderick Long's article on two types of abstraction I recommended to you in one of the discussions on your blog http://mises.org/journals/qjae/pdf/qjae9_3_1.pdf

      Delete
  2. I could get David Romer, Paul Krugman, and the ghost of Milton Friedman to do a YouTube where they say, "Daniel, you are wrong, we don't cherish realism in our model's assumptions," and then Daniel would respond, "Oh damn Bob! I told you so! Do you want some ointment for your burn?"

    ReplyDelete
    Replies
    1. I'm really at my wits end with you on this. I'm not sure where your brain is here or what else to say.

      Delete
  3. I'm confused. Isnt realism (in the Romerian sense) empirical? Isn't explanatory an abstraction?

    ReplyDelete
    Replies
    1. I was thinking of that writing this up actually. Realism is empirical in the sense that it has to do with the correspondence between our assumptions and observation. But when most people think of empiricism they're thinking in terms of the correspondence between conclusions of the theory/model and observation. If I'm not mistaken, though, a lot of Austrians say they are being empirical in precisely this sense - that they use what they know about the world around them to make the modeling assumptions in the first place.

      Delete
  4. What are the units of this function?

    ReplyDelete
    Replies
    1. Or do you mean the units for the inputs?

      Not sure.

      But definitely lollipops and unicorns for utility.

      Delete
  5. Daniel, praxeological Austrians don't just point at unrealism as such. The unrealism we are talking about is not of the kind of a Euclidean triangle approximating close-to-triangular shapers out there (although I wouldn't even call this unrealism but it's another matter). It's an unrealism of a kind that distorts the world it's trying to explain to the point of an unrecognizable caricature.

    Examples of this abound. Treating capital as just an aggregate K, maximization (which we already discussed), one (two)-consumer-good economies with money (hint: money logically can't exist in such economies), treating genuine uncertainty as quantifiable, etc. These things are akin to hypothetical physicists assuming that the Earth is cubic or that molecules themselves decide where they will move. They by definition can't help us understand anything about the real world which is thoroughly different, unless by a happy accident.

    ReplyDelete
  6. You may think me "crude" but "predictive" trumps everything else for me. The rest are just intellectual preferences in how we might best achieve predictive power.

    If you believe in Occam's razor or you are afraid of over-fitting or you genuinely believe that the system is low rank plus random noise then parsimony is an intermediate condition for achieving prediction. If you believe that any reliable predictor of the future must fit the past data then you are into realism.

    ReplyDelete
  7. Is this a sign that you're going to read The Foundations of Statistics by Leonard J. Savage soon, Daniel Kuehn? :-P

    ReplyDelete

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