Austrian Game Theory

Not precisely, but Steve Verdon posts that Austrians are 'winning' an aspect of the theoretical/methodological war with the Neoclassical mainstream in that general equilibrium models are being junked in favor of game theoretic models.

As I noted in this post the Sonnenschein-Debreu-Mantel theorem basically did in general equilibrium models. Now the new hot area is game theory, and game theory has taken a couple of interesting turns of late. The first is the theory of learning in games. Much of the early literature in game theory focuses on the equilibrium--i.e., how the game is played. There was little research into, how did you get to that equilibrium. Now this research looks at precisely that quesiton, how do you get to an equilibrium.
[...]
Also, in standard game theory one of the main tools for finding a Nash equilibrium is the best response function. The best response function for any given player is a function of the opponents' strategies taking the player's strategy as given. This is brings in another objection of the Austrians: perfect knowledge. In constructing a best response function I have to know what the other guy is going to do (or might do) and respond accordingly. The replicator dynamic is not a best response function.

Finally, in looking at some of the Austrian literature on the market process it is notable that the Austrian's refer to the market process as an evolutionary process (link, link, link). But if you look for an Austrian response to game theory you wont find much (at least I haven't found much so far). By ignoring game theory and the recent advances that look at the process of arriving at an equilibrium, the Austrians risk having one of their very complaints voided (or reduced to arguments over methodology).

Lynne Kiesling adds in response:

One sticker for me is going to be that in Austrian economic theory, one of the important features is that through the process of competition you learn both what your preferences are and what the opportunity set is. Game theoretic models usually fix those dimensions to make the models more tractable. And how do you formally handle the fact that not everything can have a probability distribution fit to it -- in Frank Knight's terms, we face not just risk but also uncertainty. Formal models don't do a good job of capturing the joint dynamics of the discovery process and the omnipresence of uncertainty, not just risk, on some dimensions.

Interesting!

Share this