Monday, December 24, 2012

Hoisted from comments: on the Economic Computation Problem

Jay Ackroyd writes:
These comments are actually inspired by a blog post I read a couple of years ago that I can't possibly find. (Stuart might be able to.)

In it, the policy economist expressed surprise and dismay that a mathematically oriented new classical guy actually took the results of his calculations literally, instead of subjecting them to an Econ 101 evaluation and rejecting them as, well, naive and stupid.

This was telling, doubly, because first, the policy guy was criticizing the competence of the math guy but, second, because he was admitting that these models really don't work.
Or at least, that the models aren't applicable without thought and results have to be checked. Hunh.
For a long time, I've been quoting Robert Solow's Growth Theory, where he describes "stylized facts." I don't think macro can go beyond that--that there are some general principles you can apply, but you can't push them too far. Rigor really isn't possible. (There are good, rigorous reasons in the literature to support this, by the way. The New Classical critique makes some of these arguments. Others are old, like the "adding up" problem. There's no way to add up the preference orderings of a collection of consumers in an economy to arrive at an aggregate demand number. This is at the heart of the NC "where are the people in this model?" critique of Keynsians.)
So in your view, the unpredictability of human behavior invalidates some models, the way it does in some design disciplines. In design, there are problems which can be solved by computation—Will the building stand up? Can this result be computed?—and problems which can only be solved by testing with humans—Will people like this building? Will people be able to navigate this interface? Some designers get good at devising first cuts at such things, but even that group has to test their work with actual humans.

Is it perhaps that the dream of an economic system that is like a machine that computes the best way to satisfy human needs is simply not achievable? This is a criticism which is usually directed at Communism, but in the end the "free" market is another mechanistic solution.

Keynesian economics suggests that an economic system that maintains both good levels employment and production requires governance and therefore must be integrated into a political system. Given that Keynesian economics is valid, what might Keynesian institutional forms be?

[Editorial changes made for clarity on the day of publication.]

No comments: