Saturday, July 28, 2012

Jonathan on Friedman on Positivism

These are all good thoughts from Jonathan Catalan:

"Unfortunately,  don't own the actual book and I can't access the complete article online, so I'm stuck reading only the first and last thirds of the article. But, this is enough to get the gist of Friedman's argument, which I actually find somewhat persuasive in some regards.  He argues that all theories necessarily abstract from some aspect of reality, and I think he's mostly correct.  However, like I write in my reaction to Caldwell's article on Post Keynesian methodology before, I don't think the preference for predictive over explanatory hypotheses is well grounded; rather, I think economics should seek to explain phenomena.  Sometimes these explanations, which establish laws of causality, can be useful for making general, relatively short-term predictions, but this isn't the main purpose of economic science.  Finally, it also seems to me that Friedman underrates the benefits of studying the same phenomenon from different angles — i.e. different models which abstract from different realities; this is readily apparent in the classical monetarist explanation for business cycles: a scarcity of money (as opposed to this being a secondary consequence or even unimportant)."


  1. Is Milton Friedman a Friedmanite in methodology? Apparently, Dr. Michael Emmett Brady disagrees with that notion, given how Milton Friedman changed his methodology. See the following article in the Journal of Economic Issues, which accepted Dr. Brady's article out of "pure luck".

  2. If Friedman had just been defending some abstraction, then he would have been ok, since in practice abstract simplifying assumptions are very often essential to good science.

    Even predictivism (the belief that the only non-instrumentally valuable virtue of a scientific theory is its predictive strength) would be fairly respectable. Philosophers of science like Bas Van Fraassen have gone down that route and put up some good arguments for it. However, Friedman's predictivism is unusual, because he confines the value of prediction to the output part of a theory. The assumptions of economic theories are as much part of their predictive content as hypotheses derived from the asssumptions + initial conditions. Put another way, assumptions are predictions, so the assumptions of a theory are part of its predictive content.

    Furthermore, "explaining phenomena" is very ambigious. Does it mean a) providing a deductive-nomological explanation i.e. a deduction of the phenomena from universal laws + premises about initial conditions? Or does it mean (b) a Max Weber-style kind of empathetic understanding with agents in a social process? Or does it mean (c) providing a causal story about the (interesting) causes of the phenomena? Or does it just mean (d) just retrodictions i.e. predictions of observed phenomena from a theory? I'd say that, given the policy-orientated nature of modern economics and the shortcomings of the deductive-nomological model, (c) is the most important.

    Unlike you and (just about all) Post-Keynesians, I'd say that Friedman managed to be a great economist in spite of his lousy philosophy of economics. However, I think it hampered him at key points e.g. he would have been far more successful in arguing for monetarism and had more of a shot at achieving a lasting move away from the income-expenditure framework had he provided a strong causal theory about money supply determination. Some monetarists, like Tim Congdon or Gordon Pepper, got round to this after the monetarist wave had peaked (and ended up going back in some ways to Fisher 1933) but by this time even the word 'monetarism' had generally become taboo.

    Discussing Friedman's methodology is one of those things that gets me into a headspin, because I often find myself arguing the against Friedman against people who are thoroughly anti-Friedmanite and on the left. I've actually seen people go from being extremely keen on Friedman's methodology to condemning it as stupid and vice versa, as soon as they find out that it's the same Friedman who wrote "Free to Choose". It's as if the methodology of economics was as tied-up with political ideology as the rest of economics and the social sciences...

    1. Oh I am definitely of the view that Friedman was a great economist. There's no second-guessing that on my part.

      I think your c. version is the most important as well. Prediction is valuable, of course, I just have doubts about the ability to do that in a study of a complex system, particularly given data limitations. The real contribution economics can make, at least now, is to explain observed phenomena. Of course we can use those explanations to generate some fairly powerful prescriptions, but those are not based on any kind of foreknowledge derived from prediction.

    2. I'm not even all that hostile to his politics, although I wouldn't take a lot of his points as far as he does.

  3. Daniel Kuehn,

    "Prediction is valuable, of course, I just have doubts about the ability to do that in a study of a complex system, particularly given data limitations. The real contribution economics can make, at least now, is to explain observed phenomena. Of course we can use those explanations to generate some fairly powerful prescriptions, but those are not based on any kind of foreknowledge derived from prediction."

    My thoughts exactly.

    As for his politics, I often find myself agreeing with his conclusions but not for his reasons e.g. I prefer his pragmatic policies like a negative income tax, school vouchers and university bonds to the kind of radical libertarianism that one finds on and so on, but Friedman's argument for this pragmatism was that (a) we know where we want to go, i.e. to a minimal J. S. Mill-type state but (b) we have to go there gradually, for political reasons and because people have to adapt.

    The British conservative (small c) in me disagrees with (a) - I don't think we can work out a priori where the boundaries of the state should be, so we should try various tweaks to institutions and see if they work/fail and why. For example, if I ever believed the Keynesian arguments for the possibility of a liquidity trap, I'd go for some minimally innovative solution e.g. underfunding existing expenditure to boost the money supply via fiscal means rather than a big "stimulus".

    Friedman always said he wasn't a conservative, and he was right. He had a 19th century liberal's confidence in freedom. Instead, I think that D. C. Stove's "Columbus Argument" ( ) applies to right-wing policy changes as well as left-wing policy changes. Maybe my view is affected by the fact that there are quite a few state interventionist policies in the UK (my home country) which I can see working well e.g. the monopsony power of the National Health Service seems to get the balance about right between market and state, so I'm willing to apply Hayekian/traditional conservative reasoning in FAVOUR of the state as well as against it.

    What do you think? Which parts of Friedman's politics do you like/dislike?

  4. I believe the Austrian complaint to Milton's positivism is that theories shouldn't abstract /any/ material and relevant details.

    If my theory is that depletion of stocks raises subsequent savings, then I can omit nothing relevant (e.g. does a given household have insurance, or are stocks depleted because of cancer diagnosis). I might abstract irrelevant details, but nothing whatsoever that's both material and relevant.

    I'm thinking of Kruman's paper on country size and first-mover advantage, where (to my mind) he abstracts to absurdity. Specifically, he ignores other factors so noisy that country size becomes an error term in a very large regression. Because he's abstracted those factors, he gets a clean result, but in abstracting almost the entire regression, he's left saying that backing out of a parking space means you'll never get to your destination.

    1. "I believe the Austrian complaint to Milton's positivism is that theories shouldn't abstract /any/ material and relevant details."

      The criticism is more along the lines that assumptions have to be realistic, not that theory shouldn't abstract from certain details. If you didn't abstract from certain details then you'd never be able to establish causality between events. And, well, what's relevant to a particular event will depend on how the economist interprets it.

    2. Yes, good point. I like keeping all the details, but then I do case analysis. Indeed a general theory would have to drop detail.

  5. That should read Krugman. As in Paul.

  6. As W. Peden points out, there is a good bit of ambiguity in the question of prediction vs. explanation. For me, prediction includes postdiction. The point is taking data that are given and estimating data that are not given (even if they have already happened.).

    To say that explanation is the object makes me nervous, because humans are extremely good at coming up with explanations. E. g., Thor is angry. OTOH, Velikovsky predicted that Venus would be extremely hot, when most astronomers thought that it would be cold and damp. But his explanation for the extreme heat of Venus is crazy. ;) So this is not an either/or question.

    I find the retrospective explanations for the recent and ongoing economic crisis interesting but mostly unconvincing. It is easy to make up post hoc explanations. At the same time, I do not fault the economics profession for failing to predict it. At least, for failing to make any kind of precise prediction.

    Here is a challenge. Interpolation is easier than extrapolation. Given the state and history of the economy up to January, 2001, and the state and history of the economy since the end of December, 2008, explain and predict the intervening states of the economy. If that is too hard, start from the end of December, 2005.

    If that is too hard, why do we place any credence in economic projections 75 years into the future? ;)


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