Thursday, April 5, 2012

One thing I do agree with Keen on

Is the criticism of representative agent models.

Represenative agents aren't the end of the world, of course, but I think they're very misrepresented. All representative agents give you is a macroeconomics grounded in optimizing behavior. That is the sense in which they are "microfounded". A better word might even be "optimization-founded". That's an important technique to use, but it's not entirely satisfactory. Agent based modeling (or something else other than representative agent) is what really gets you microfounded models.

Do agent based models get you substantially different insights than well done representative agent models? That I don't know.

2 comments:

  1. Daniel Kuehn: From what I've read of ABM, the agents are designed to behave randomly according to the computer program. That way, you see power laws seen in the data generated. Agent-base models aside, would Daniel Ellsberg's decision theory provide better results than the Subjective Expected Utility-based representative agent?

    ReplyDelete
  2. I don't see representative agents 'microfounded' so much as 'microfied' i.e. they are simply a micro model extrapolated up to the entire economy, with no thought given to coordination problems or fallacies of composition.

    ReplyDelete

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