Wednesday, February 23, 2011

Cargo Cults

Someone once accused me of practicing "cargo cult economics." At the time this just sounded goofy to me, but a commenter clued me into the fact that this comes from a commencement address that Richard Feynman made at Caltech in 1974. In general, it's quite entertaining, though maybe of limited use for economists as he was thinking mostly in terms of experimental science. However, there are some very nice parts. For example, here he is discussing honesty in science:
It's a kind of scientific integrity, a principle of scientific thought that corresponds to a kind of utter honesty--a kind of leaning over backwards. For example, if you're doing an experiment, you should report everything that you think might make it invalid--not only what you think is right about it: other causes that could possibly explain your results; and things you thought of that you've eliminated by some other experiment, and how they worked--to make sure the other fellow can tell they have been eliminated.
The following sounds like Ed Prescott, don't you think?
There is also a more subtle problem. When you have put a lot of ideas together to make an elaborate theory, you want to make sure, when explaining what it fits, that those things it fits are not just the things that gave you the idea for the theory; but that the finished theory makes something else come out right, in addition.
More on honesty:
The first principle is that you must not fool yourself--and you are the easiest person to fool. So you have to be very careful about that. After you've not fooled yourself, it's easy not to fool other scientists. You just have to be honest in a conventional way after that.
And here's one for Paul Krugman:
I would like to add something that's not essential to the science, but something I kind of believe, which is that you should not fool the layman when you're talking as a scientist.
Happy reading.


  1. To me "stimulating" the economy is roughly equivalent to the "cargo cult". Essentially, you observe something - control tower, airstrip, landing lights (aggregate demand, GDP) - and then you make some causality assumptions which, however, depend on peoples decisions.
    The cargo cult thought that if you build a proper airstrip, the airplanes start flying. The usual explanation "aggregate demand failed, the government should step in and replace it" is roughly equivalent to the cargo cult problem. The aggregate demand was higher before the crisis, therefor if we force it back, economy will start working again....when the lights were on, the airplanes were flying - let's switch the lights...


  2. > … though maybe of limited use for economists …

    I'm not so sure whether this is true. On the other hand monetarists believing in Friedman's 'The methodology of positive economics' must come to this conclusion. I've read it, had a good lough and thought his methodology is only ridiculous.

  3. "The following sounds like Ed Prescott, don't you think?"

    Thank you for pointing this out! I think that the fundamental difficulty in business cycle macro, really, is the lack of clear criteria for theory evaluation. There isn't enough data to identify many empirical patterns in a robust way, and it's virtually impossible to test "out of sample" unless you're willing to wait a few decades. (You can claim that you're testing "out of sample" by estimating a model on data from an earlier time period, and testing it on a later one. But since you're almost surely aware of what happened during the later time period, and at some level are using it to inspire your theory-building, this exercise isn't really honest.)

    So we get a lot of vague handwaving. (Oh, look, it matches these moments; therefore it must be at least partly right!) I've never understood, for instance, how the fact that the basic RBC model roughly matched the relative movements of investment, consumption, and GDP was at all a success; since you're using GDP to construct TFP in the first place, there aren't many degrees of freedom here, and the only remotely novel prediction of the model is that investment should fall by more than consumption. (Which is something you can get in pretty much any model with a neoclassical backbone, TFP shocks or not.) The subtler predictions of the model, like the strong procyclicality of the real interest rate, aren't right at all -- yet they're fundamental to the RBC account of how recessions work. Why was this ever supposed to be a successful theory?

    Ultimately I think macroeconomic models must be judged by the strength of their microfoundations, and by introspection about whether those microfoundations really match up with the world. But this is a tough, ambiguous way to evaluate models, and it's no surprise that economists will sometimes screw it up.

  4. If there's one thing I really can't hear anymore then it is this "microfoundations of macroeconomics". This is an oxymoron! Micro ecomomists invading another domain like Dschinghis Khan with their banana/apple curves.

  5. @Stephan: You are saying that positive economics is ridiculous and that "microfoundations of macroeconomics" is an oxymoron! We are scientists, so it would be nice if you can contribute to the discussion by making an intellectual effort and provide arguments to support your claims!

  6. @German Cubas
    "We are scientists, …" Good joke ;-) Before you go through the roof. This was also a joke.

    You want an argument? How about this paper I agree with Friedman, Positive Economics, and the Chicago Boys Of course you must be willing to put to a rest (only for some minutes) the chronic autistic behavior of "scientific" economists.

    If you want a further argument why positive economics is dubious I think a picture tells more than a lot of words: Economists and Statistics

  7. Stephan said "you must be willing to put to a rest (only for some minutes) the chronic autistic behavior of "scientific" economists."

    Stephan, I am not terribly politically correct but I do draw lines from time to time. One of those lines that people with serious problems -- i.e., autism -- should not be used as cheap throwaway insults. Be glad that you do not have autism.

    I will also note that your comment has no actual content except to express unexplained dissatisfaction with economic research.

    Why don't you put on your thinking cap and make a well reasoned argument as to what you don't like about the state of economic research and how it can be improved.

  8. @Chris
    I wrote a reply. But it did not show up? I don't think it was offensive? Nothing I can do about it.

  9. Strange? I tried it again. The blog does not like my comments ;-) Guess it thinks I'm bad for the karma of other commenter here.

  10. Using a developmental disability as an insult is offensive. Yes.

    These are real people with real problems. They do not need to have you use these issues as a synonym for whatever you don't like about economic research.

  11. Cool down! I've emailed Stephen whether he would post my suppressed comments. Nothing more I can do. But obviously you can only run a lot of regressions but not read a simple sentence.

  12. Nice article, thanks for the information.

  13. Second anonymous,

    What I meant by the "sounds like Prescott" was the idea that you calibrate based on micro evidence and long-run growth observations, then look at what the model predicts about fluctuations. Sounds like Feynman, don't you think?

  14. ^ Ah, I misinterpreted you. I understand your analogy, but frankly I think the analogy works better in the opposite direction, for two reasons. First, "TFP" is a giant black box that is estimated from cyclical production data. It is unsurprising that feeding TFP back into a model should produce something that resembles the cyclical movement in the economy, at least along some moments. (And the core RBC model doesn't do well at matching most moments.) So in some sense, Prescott was violating Feynman's dictum that "those things it fits are not just the things that gave you the idea for the theory" -- not quite literally, but in the sense that the real power of a theory should be its "out-of-sample" explanatory power.

    Second, let me get a little more specific: in choosing a Cobb-Douglas specification for the "composite good" formed by leisure and consumption -- along with a lagged specification for leisure that made intertemporal substitution more likely -- Kydland and Prescott 1982 displayed either technical incompetence or intentional model-fitting. These were extremely questionable modelling decisions that were essentially rigged to make the RBC model display a level of procyclical labor supply that matched the data. Contra your claim here, it's not as if Kydland and Prescott innocently applied a standard neoclassical growth model and found that it explained fluctuations well; instead, they applied a variant of the model that was particularly suited to "explaining" fluctuations. Either they stumbled in this direction (and didn't realize the one-sidedness of the decisions they were making), or they weren't really acting in the way that Feynman suggests.

    And their paper suggests that it was the latter. On page 1351, they note that Grossman and Lucas found that a "non-time-separable utility function is needed to explain the business cycle fluctuations in employment and consumption". Um, I hate to belabor the point here, but if you note that something is "needed" for your model to come out right, and then assume it (thus making your model come out right!), your work is not exactly a triumph of science. They justify this assumption by pointing to vacations (!!!), the fact that some people move in and out of the labor force voluntarily for extended periods (if they were labor economists, they would know that prime working-age men do NOT do this), and the seasonality of employment (which is in large part a consequence of institutional design that lets students out for the summer). This is not intellectually satisfying. Frankly, it's a bit of a joke. An interesting contribution, maybe, but nothing to win a Nobel Prize over. And the contemporary version of Ed Prescott ("unemployment remains high because of... Barack Obama!") is virtually the last economist I'd cite as living up to Feynman's scientific ideals.

  15. Yes, I understand what you're saying. The basic RBC model is pretty simple, and it's easy to pick holes in it. However, the ideas and methods were enormously influential, and it got people thinking in a serious way (as you are here) about the relationship between theory and data. Ed gave an enormous amount of time to students, and taught a lot of excellent economists, who have some very diverse ideas. If you have not talked to Ed, you should do it sometime. He is a very serious scientist, and I'm quite willing to forgive the occasional extreme view. Even in those cases, Ed has thought carefully about the issues and is willing to argue about them in a serious and non-personal way. "Nothing to win a Nobel Prize over." Nonsense.

  16. It is unsurprising that feeding TFP back into a model should produce something that resembles the cyclical movement in the economy

    You don't feed actual data on TFP back into the model. The whole point of RBC methodology is that the model is subjected to random shocks, generated by a computer. These shocks are chosen to match the actual fluctuations in output, because we want to compare the model with the real world business cycle. But there's no reason to expect a correlation between predicted and actual TFP or anything else. If you believe in the traditional Keynesian/monetarist explanation of fluctuations, there should be no correlation between fluctations in TFP and the business cycle (neglecting labour hoarding, which I think was a response to RBCs). So the fact that actual data on GDP is used should not imply a systematic relationship between actual and predicted TFP.

  17. I've been likening most ALL academic economics -- pitched for policy relevance as cargo cult economics for more than 20 years.

    It sure ain't science.