The question, does game theory work, is two fold.
First, does game theory predict how people will actually behave? The answer is a resounding no.
And what was their boo-boo? The very same one as permeates all econometrics, as opposed to Austrian Economics.
Everyone realizes that economic reality is very complex. It’s about complicated people in complicated situations. So something has to be “ignored” when analyzing a situation.
But what should we ignore? Here’s where AE and the mainstream part ways. AE says we should keep the human, since economics is about humans acting, and simplify the situation. Thus the famous ceteris parebis assumption that permeates AE.
But the mainstream decided that the very first thing to go has to be the “human” part of human beings. Real people are just too messy and complicated to understand. So let’s replace them with zombies and call that economics.
And that’s exactly where game theory has failed and continues to fail. It predicts how zombies will behave in a given situation, but somehow real people stubbornly insist on acting like real people, not like zombies.
From as far back as the 1950s, when game theory
was in its infancy, researchers wanted to investi-
gate the extent to which the exciting new theorems
and models they were developing could predict
actual behavior. Merrill Flood conducted one of the most famous early
experiments to investigate this question.2
This work introduced the prisoner’s dilemma,
now regarded as one
of the most important formal games
in the game theory canon.
The most important game in the whole canon. Surely the most important game in the whole canon will, you know, actually tell us something?
Surely our simplification of replacing people, who are complicated, with homo gamicus, who is simple, has not thrown out the baby with the bath? We, the brilliant game theorists, can’t be that stupid, can we?
The test was made. Real people replaced the “models”.
So, what did they do?
Well, they didn’t choose to follow the
prescriptions of game theory, which
in this case points to mutual defec-
tion in every round of the game. In
fact, mutual cooperation occurred
nearly two-thirds of the time…
In other words, out of 100 predictions, the most important game in the whole canon was dead wrong 67 times.
Guys, you are taking tests now in school. What would your teachers say if you got 67 out of 100 questions dead wrong?
“Why these poor grades in psychology, son?”
“I dunno, Professor Mises. I studied the zombies on The Walking Dead, and based all my theories on how they did things.”
“No wonder you failed, you dope.”
This happened in 1950. Being the brilliant folks that they were, they realized, eventually, that something was amiss. Research was begun.
A pair of game theorists won a Nobel Prize for stating the obvious:
Their starting point
was the observation that expected
utility theory doesn’t predict how
people make decisions in practice.
And when did they win this prize? 1951? A year after the great fiasco described above? Nopers.
It took fifty two full years for things to sink in. That Nobel Prize was awarded in 2002.
And they still haven’t realized what the problem really is.
Much work remains on reconciling the theoretical models and solutions of game theory with observed
Imagine if the student, who got a grade of 33 on his test, told the prof something like that.
“Much work remains on reconciling the acts of the zombies on the Walking Dead with observed human behavior.”
So much for the first part of the question, does game theory work. Does it predict anything at all? The answer is a resounding horse laugh, or in academic doubletalk, much work remains on reconciling things.
Now for the second aspect of the question does game theory work. Forget about those dummies in the real world. Who cares what they do? Is game theory a reliable guide for what we should do if we were placed, say, in the prisoner’s dilemma? Maybe everyone else, who doesn’t follow game theory, is losing out, but we smart ones who know game theory will do much better than them.
Again, they are making the exact same mistake, conflating their stupid models with, you know, reality.
The first point to make here is that
often, game-theoretic models are ap-
plied in entirely inappropriate circum-
stances. Such models are predicated
on a host of assumptions—some that
are easily justifiable, others that are
perhaps harder to justify.
But they did have two spectacular successes, they think.
The UK govt auctioned off rights to use certain electromagnetic frequencies to cell phone companies.
This happened …in the first half of 2000…the peak of the
dot-com boom, when wild speculative investment was made in IT and
The peak of the dot-com boom. Wild speculative investments. I wonder what happened? Did these companies overpay, perhaps?
They sure did, to the tune of 34 billion dollars. The govt had hired game theorists to set up the rules of the auction, and they are taking all the credit.
Was the fact that this was the peak of the dot-com boom relevant, when wild speculative investments were made in cell phone technology? Did it have anything to do with these results? Nope, say the game theorists. That was a “coincidence”.
The second spectacular result happened in Los Angeles World Airports (LAX). Game theorists came up with a model of when to search people who might want to blow up the airport, and guess what? It works. Nobody has blown up anything at LAX.
“You mean the prisons are full of terrorists now, that the game theorists have nabbed?”
“No, I mean nothing happened at all. It doesn’t get any better than that, right? Plus one for game theory. Total proof that we outsmarted the hoards of terrorists out there.”
But hey, let’s be fair. We normal people may think a score of 33 out of 100 is a total flop, but it only looks that way.
Game theory can work under both its descriptive
and normative interpretations, although it might often appear that it
“Professor, I could have gotten a perfect score on that test, although it often may appear that I didn’t.”