Mistaking Beauty for Truth in Science and in Economics

Sean Carroll on Paul Krugman's NYT Magazine piece:

One of the reasons it’s a great essay is that it’s a wonderful example of popularizing science. You can debate all you like about whether economics counts as a science, but there’s little doubt that Krugman does an amazing job at explaining esoteric ideas in non-technical language, and is so smooth about it that you hardly realize difficult ideas are even being discussed. I wish I could write like that.

One part of the essay worth commenting on, or at least musing about, is the punchline. Krugman thinks that a major factor leading to the failures of economics to understand the mess we’re currently in was the temptation to think that beautiful models must be right...

Without knowing much of anything about the relevant issues, I nevertheless suspect that this moral might be a bit too pat. Sure, people can fall in love with beautiful theories, to the extent that they overestimate their relationship to reality. But it seems likely to me that the correct way of understanding all this, once it’s properly understood, will look pretty beautiful as well. General relativity is widely held up as an example of a beautiful theory — and it is, when understood in its own language. But if you put the prediction of GR in the Solar System into the language of pre-existing Newtonian physics (which you could certainly do), it would look ugly and ad hoc. Likewise, Newton’s theory itself is quite elegant, when phrased in the language of potentials on a fixed spacetime background; but if you express the theory in terms of differential geometry (which you could certainly do), it looks like a mess. Sometimes the beauty/ugly distinction between theoretical conceptions is more a matter of how well we understand them, and less about their intrinsic qualities.

Paul Krugman on Sean's point:

[O]n an interesting point raised by Discover (via Mark Thoma): won’t we eventually have a true theory that’s as beautiful as the full neoclassical version? Well, one thing’s for sure: we don’t have that beautiful final theory now, so the current choice is between ideas that are beautiful but wrong and a much messier hodgepodge. But my guess is that even in the long run it won’t be all that neat. Discover suggests general relativityversus Newtonian physics; but a better model may be meteorology, which as I understand it starts from some simple basic principles but is fiendishly complex in practice.

Actually, let me put it this way: the economy is a complex system of interacting individuals — and these individuals themselves are complex systems. Neoclassical economics radically oversimplifies both the individuals and the system — and gets a lot of mileage by doing that; I, for one, am not going to banish maximization-and-equilibrium from my toolbox. But the temptation is always to keep on applying these extreme simplifications, even where the evidence clearly shows that they’re wrong. What economists have to do is learn to resist that temptation. But doing so will, inevitably, lead to a much messier, less pretty view.