Economics vs Physics

Apropos our economics vs climatology thread, Jim Manzi has an extremely interesting article comparing economics to other sciences.

In the end, hard sciences produce a body of knowledge that lets us make practical predictions with tolerable reliability. … On the important questions about which we seek predictive guidance from economics … it is possible to find highly credentialed economists who answer one way with great certainty, but it is just as easy to find equally credentialed economists who answer the opposite way with equal certainty.

Also, Stoat and I are having quite an interesting conversation over at his place.

Comments:

  1. The comparison between economics and physics is interesting, but I would be careful with it. There is an unfortunate large misconception carried by many people that science only exists in the laboratory, any any science done outside of the laboratory isn't "real" science. Manzi's argument can potentially be interpreted as being "economics can't be tested in the laboratory, therefore it isn't real science". Of course, much of atmospheric/climate science can't be tested in a laboratory, either, so trying to use such an argument against economics can backfire against us as well. Perhaps we on this site may recognize the distinctions between atmospheric/climate science and economics that would make such a counterargument invalid, but many people may not. Again, I would be careful with this kind of argument.

  2. Manzi's argument leads to his conclusion that, because of the unreliability of economic analysis, laisser-faire is the only sensible way to go. However, given the scale of the climate crisis and our tardiness in tackling it, some dirigisme may actually now be required. Manzi's dissing of economics may not be leading to the answer that some climatehawks are looking for.

    Economics is a fundamentally a more difficult subject than physics. That doesn't mean that we should ignore it, any more than we should prefer playing Checkers to Go.

  3. I'm not saying we should ignore the subject. I'm not so sure about most of its practitioners.

    I am very sympathetic to outsiders coming in and trying to understand climate science and being very dubious about it, because my relationship to economics is much the same. I almost want to start an Econ Audit site.

    Anyway, I think having dubious outsiders is healthy. Hostile ones is another story, alas...

  4. Andy Skuce:

    Manzi's argument leads to his conclusion that, because of the unreliability of economic analysis, laisser-faire is the only sensible way to go.

    Which is really just another manifestation of the "we don't know, therefore we know" canard.

    Economics is fundamentally harder than certain types of scientific research. But that's not the main problem. The problem is that _in addition_ to being fundamentally hard, mainstream economists seem to have an inclination to do it wrong.

    That doesn’t mean that we should ignore it, any more than we should prefer playing Checkers to Go.

    And since you brought up the xkcd comic, AI in general is also fundamentally quite hard, but at least AI technologies are actually getting somewhere (chess, OCR, speaker voice recognition, etc.). And while there'll always be AI cranks (e.g. Ray Kurzweil), at least they don't get in the way.

    -- frank

  5. "AI in general is also fundamentally quite hard, but at least AI technologies are actually getting somewhere (chess, OCR, speaker voice recognition, etc.)"

    Actually, what's happening is mostly that computers have gotten faster and have immensely more storage than in times past.

    In terms of fundamental research (i.e. the "mathiness" bits), AI is very slow-moving.

    Economics is much harder than physics ... it might be easier to master than physics because in some sense its being so difficult means that there's a lot less of it, and what's there is kinda squishy in a way that science based on the physical world isn't.

  6. Economist Paul Krugman pointed out some time ago that without a Keynesian stimulus the Great Recession would drag on and on and on. Indeed it has although not so badly as the similar problem in parts of Europe.

    Despite some distinctly odd individuals and schools (Austrian School) mainstream macroeconomics seems to have some predictive power. So perhaps the underlying concepts are not completely wrong.

  7. Krugman, in my opinion, knows what he is talking about about as well as it is possible to know within the core discipline. Although out of favor in Washington, his approach seems right, given his assumptions.

    But while I think infrastructure investments make a great deal of sense, I don't see him having much interest in resource constraints, or in moving beyond growth.

  8. In terms of fundamental research (i.e. the "mathiness" bits), AI is very slow-moving.

    True enough, but think about how much worse it could be. If AI were done like economics, we'll be seeing the following:

    (1) People in the field will still be talking about "Markovian Techniques in Artificial Intelligence" like it's some new weird thing.

    (2) Almost every write-up on AI will begin with an obligatory mention of the Turing Test.

    (3) Every failure in AI algorithms will be explained away with excuses such as 'well, you know, AI is hard, because humans can always study the AI algorithms and try to fool them', yadda yadda yadda.

    Yet none of this is happening. AI is moving along, guided by the scientific method and crisp performance metrics. (But arguably, that might not have happened without a couple of AI winters...)

     * * *

    Speaking of mathiness in AI, I'm now also wondering about a related question: How much of current advances in climatology is simply due to greater computing power, and how much is due to fundamental advances in meteorological theory or mathematical theory? Perhaps mt knows the answer to this.

    -- frank

  9. Regarding recent progress, I will venture that burgeoning paleoclimate evidence has been at least as important as increasing computational power has been. High end models have become steadily more realistic in output and more complex, and while this pursuit has many uses I wonder whether these are still adding to understanding OR predictive skill (which are different things) commensurate with the effort put in. But the past has been delivering surprise after surprise lately. The last millenium, where McIntyre and his followers spew most of his venom, contains relatively little of interest, but deeper into the quaternary we keep finding unexpected twists and turns. I think a correct picture of the glacial cycle has, arguably, finally emerged recently, with contributions from data and model. I think this was the most important wide-open question in climate a decade ago.

    PS - It's really hard to advance the pure mathematics of the climate system, and I suspect that most of the best work in GFD theory may be behind us.

    PPS - Somebody might conceivably come up with a more practicable discretization of ocean dynamics for very long computer integrations. This would have the flavor of pure math and would be a huge step forward. But it doesn't look promising. The deep ocean just doesn't like to be modeled.


Leave a Reply

Your email address will not be published.