Brad DeLong QOTD: On the Density of Models

Today I had the pleasure of seeing an excellent job talk in Sociology in the morning, and a fascinating seminar presentation by Brad DeLong of UC Berkeley and blogging fame in the afternoon. I’d never been to an economics seminar here at Michigan before, but I get the sense that this one deviated from the norm a bit. Even so, it was a lively discussion on the topic of the efficacy of fiscal policy during a depression. DeLong and Larry Summers are apparently working on a Brookings paper on the topic, and DeLong was trying out some of his ideas on the crowd here at Michigan.

Modern economists are often criticized for not actually knowing anything about the actual economy. I would occasionally joke with my former roommate – an econ PhD student who at the time was writing an abstract paper using the mathematics of entropy to calculate the value of “an information” – about this. Brad DeLong is perhaps the exception that proves the rule, and his working knowledge of important stylized facts, as well as in depth examples, about every element of the US economy – as well as the crisis in Europe, and so on – was just what you’d expect given his prolific blogging. During the talk, DeLong considered several alternative models to the one he (and Summers) were working on, and at one point mentioned a funny critique of modern economics that apparently comes from Andrei Shleifer:

The set of admissible models is dense in the space of possible conclusions.*

Ok, I admit, you have to be a pretty big econ or math nerd for this to work… but the quote basically claims that for every possible conclusion, there is some model acceptable to modern economics which supports that conclusion. In other words, we have a lot of fancy models that let us justify any claim, but not good enough rules for choosing between the models.

All in all, an excellent day, and a fascinating talk.

* Or possibly “policy choices.” Something like that.



  1. Michael Bishop

     /  December 1, 2011

    “the quote basically claims that for every possible conclusion, there is some model acceptable to modern economics which supports that conclusion. ”

    Allow me to nitpick, I wouldn’t say “every possible conclusion,” but rather, “many possible conclusions.” The former would seem to imply economists aren’t close to consensus about anything.

    • Well, we can argue about the amount of consensus in modern economics (see the excellent table in Marion Fourcade’s book on survey data of opinions held by economists on different issues in different countries). But the quote does, in fact, say that. More specifically, it says that for every conclusion there is a model that justifies a conclusion arbitrarily close (hence “dense” in a mathematical sense).

      More seriously, I agree that there is consensus in economics on many issues. But that consensus may exist *in spite of* rather than *because of* the set of admissible models. In other words, economists may agree that there are models that are acceptable to use for some purposes but that collectively they tend to prefer a smaller subset which generally produce a narrower range of conclusions. So, the existence of consensus as a practical matter, and the existence of admissible models that produce conclusions in contradiction to the consensus, are not contradictory.