"When these assumptions [of equilibrium, competition, and the completeness of markets] fail, the very concept of rationality becomes threatened, because perceptions of others and, in particular, of their rationality become part of one’s own rationality. Even if there is a consistent meaning, it will involve computational and informational demands totally at variance with the traditional economic theorist’s view of the decentralized economy." -- Kenneth J. Arrow (1986). "Economic Theory and the Hypothesis of Rationality" Journal of Business, V. 59, N4, Pt. 2 (reprinted in The New Palgrave)A lot of this attempt to conceptually clarify the meaning of "rationality" uses game theory. In parallel with this theoretical work, economists and others have performed a body of experiments. These experiments found many cases in which the predictions of neoclassical theory were falsified. Daniel Hausman, in a chapter in his The Inexact and Separate Science of Economics, treats the resulting work as a case study. He tries to understand why mainstream economists retain a theory with such a poor track record in controlled experimental settings.
I have blogged on experimental economics before. (A Mark Thoma post post inspired me to put this post up.)
Posts
8 comments:
It seems that the existence of a single irrational individual would lead to the breakdown of the entire system, thus nullifying all conclusions made under the assumption of rationality. If there exists an irrational individual and I base my (rational) decision off of a choice made by the irrational individual, my decision could have less than ideal results due to the irrationality of the other individual. This could lead to a domino effect of choices and decisions made with imperfect information.
Robert,
Keep nibbling at the heels of the "mainstream economists." We need more of your kind of "nibbling."
As to rationality I keep close James March's A Primer on Decision Making; How Decisions Happen. In his study, March first lays out the "loosness" and "inconsistencies" in common uses of rationality, and notes, after defining rationality, that "Pure rationality strains credulity as a description of how decisions actually happen."
Then March proceeds with various chapters dealing with individual and group decision making including political decision making, all building from the frame he and Herbert Simon called "bounded-rationality."
The book is worth a look. Keep nibbling...
He tries to understand why mainstream economists retain a theory with such a poor track record in controlled experimental settings.
As opposed to that theory with a good track record in controlled experimental settings? Which theory would that be? Prospect theory, which is for all intents and purposes mainstream?
And of course some controlled experiments - Vernon Smith's for example - confirm that a good deal of mainstream theory works pretty well even when many of the assumptions of the theory fail.
Also, in game theory, the requirement of rationality, well, rationalizability, at least, is a very weak one. Essentially it only requires that players avoid strictly dominated strategies, like failing to block in a game of tic-tac-toe when your opponent has two X's or two O's in a row. Other than that a lot can happen, and that, if anything is the weakness of the theory.
If one in a hundred people behaves irrationally, the system breaks down? Sean's conclusion seems like it must be too extreme. But I cannot see anything wrong with his understanding of the logic of game theory.
I think there is another philosophical problem in game theory. In calculating the Nash equilibrium, for example, one considers a player's response to other players' moves, including "irrational" ones. What kind of counterfactual is this?
Thanks for the encouragement Sean and Dave. I haven't read James March. I do find interesting the little of Herbert Simon I've read. (The Hawkins-Simon condition is important in the Input-Output theory I often go on about, but that is a very small part of Simon's work.)
Whether or not prospect theory is now mainstream , philosophers of science can argue about the optimal rate at which it should have been adopted by economists. But does prospect theory really get that kind of space in intro teaching? If a body of experiments falsify a theory, the theory is falsified. The existence of confirmatory experiments helps suggest conditions for a new theory, but cannot make it unfalsified.
But does prospect theory really get that kind of space in intro teaching?
Does Expected Utility Theory? I don't know, I don't teach micro. I mention both in intro Macro when we talk about financial markets - the area where PST and EUT have the most relevance. For other areas, as Kahnneman himself said, standard theory is fine.
Vast majority of econ departments have one or two behavioral people and maybe an experimental person. In smaller departments it's either one or the other or the person plays a double role. Most offer an intermediate class in behavioral theory or experimental economics, or behavioral finance. The stuff is a bit hard to teach at the intro level unless you expect undergrads to be able to generalize Savage's axioms and show that a prospect utility function exists. Best you can do is just mention loss aversion or the fact that things get weird in the small-probability/huge-payoff cases. At the intermediate level you can set up the weighted utility function.
A theory is composed of many different parts. If it is falsified in an experiment, it's worth looking into which part of the theory is screwing things up rather than just throwing the whole thing out the window. It's worth remembering here that Expected Utility Theory can be seen as a special case of Prospect Theory.
And of course we can have an experiment falsify a theory, simply because the experiment is set up incorrectly.
"Cartwright establishes that such arguments are in principle acceptable; that, as Friedman argued many years ago, there is a strong case for using models based on unrealistic assumptions. However, for economists, the sting in the tail is that she argues that many of the assumptions used by economists reflect the poverty of the economic theory; because there are so few principles to rely on, large numbers of assumptions are necessary in order to provide models with enough structure to obtain results. Her conclusion is that most economic models are too 'model-dependent' (dependent on assumptions that are made solely to give models enough structure to generate results) for there to be much hope that they will have external validity."
Roger Backhouse, Review of "The Experiment in the History of Economics" Cambridge Journal of Economics, June 2006.
H. E., thanks for the reading suggestion. I believe you are referring to a review in the Economic Journal, not the CJE. My favorite university library doesn't have the book under review. My ability to read the review is also somewhat limited, since my favorite home computer is under repair.
Robert, You're right about the source of the quote. I haven't read the book myself; I liked the quote. Another in the spirit of Kirman, regarding the arbitrariness of auxiliary assumptions that economists make to produce models with stark predictions.
Post a Comment