Eran Shmaya, in a beautiful recent blog post, adds some sociological angles to the economists’ never-ending introspection of whether their field is a science, where the starting point of such discussion is Popper‘s falsifiability. The discussion about whether economic theory has any significant prediction power has certainly picked up lately due to the financial crisis that convinced many people that the answer is “No”. (With a variant taken by many micro-economists claiming that micro is fine and only macro-economics is a superstition.)
Whenever I hear these types of discussions about the prediction power economic theory, I always think about the irony that, as an algorithmic game-theorist, I am mostly interested in using economic theory as a branch of mathematics… Arrow’s theorem is simply true, not falsifiable. So is Meyrson’s expression for optimal auctions, as well as many other mathematical tidbits that we, computer scientists, pick up from economic theory. We don’t really care how well these theorems relate to human situations, since we use them in different circumstances, those related to computer systems. I often wish that theoretical economists would allow themselves to stray even farther away from the realities of their human subjects.
This is not to say that we, in algorithmic game theory, need not worry about how well our theories correspond with reality. We should. We have not really started doing so. And, it’s getting high time we do. But our concerns may be more like those of engineers rather than those of scientists. From economics, I’m happy to just take the theory…
Turing's Invisible Hand 
Great post Noam.
Creating a theory of the real (social) world is hard though. We may start with easy situations.
Just be careful about zombie Karl Popper
BTW, in a corporation, I have been witnessing the inaccuracies of predictions of micro-economics regularly for the last ten years. These inaccurate predictions affect one situation at a time, so it does not become a glaring example.
I believe in the micro-economic theory and also in macro-economic, but I also would not recommend solely dependent on them in a complex situation, since we overlook so many real variables and so many futuristic things always remain uncertain or unknown.
Arrow’s theorem is not true, nor falsifiable.
Arrow’s theorem follows from a set of axioms, or assumptions made formal.
The utility of those axioms lies not in being true, mostly true or something like that. They are plausible formal interpretations of intuitions that we have.
Axioms are translations of pre-theoretical ideas which are truth preserving.
Formalization allows us to see underlying sameness in phenomena that we would not have seen but for the formalization.
But, we are not getting closer to Platonic truth with formal models. The necessity in the model may or may not have any relationship to truth outside the model, whether the outside is computers or people in a market.
There are two possible motivations for studying a subject, its utility (useful predictions it makes) or its inherent beauty.
No one would study about cars if they did not take us places, while number theory would be studied purely for interest.
Economic systems are fairly complex and ugly beasts, and not sure they would be studied if not for their utility.
Rationality is beautiful and deserves to be studied for its own sake. That’s why we have game theory and microeconomics.
@ annon who writes: “There are two possible motivations for studying a subject, its utility (useful predictions it makes) or its inherent beauty.”
No, the utility of studying a subject is also the control or manipulation of the subject. Useful predictions is just a small part of utility.
@ anon who writes: “Rationality is beautiful and deserves to be studied for its own sake. That’s why we have game theory and microeconomics.”
You are studying rationality in game theory and microeconomics, you are studying some optimization or maximization devices. Without some analogue of Church’s thesis for rational behavior, there is no conceptual link between the two.
Human processors are freedom-order-fairness (i.e.symmetry)-efficiency aware agents (all these concepts can be preciselly defined).
Avalaible economic theory/game theoretic models of rational agents takes into account only efficiency awareness (i.e. profit maximization) wich is a very restricted model of the “real” rational agent.
Therefore the two facts you comment in your post (failure of economic theories at predictions, in my experience including microeconomy, and good aplicability to AGT)depends on the same cause and should not surprise anyone.
Great post. So which bird (if any) would computer scientists be?
Amphibians?
Of course, there is a tension between the original aspirations – to explain social behavior – and the more ‘mundane’ engineering uses.
If the major theorems were just viewed as mathematical constructs and if there were none of the potential connection to policy or empirical behavior, do you think the best economics students would still be attracted to the field?
From just one quick look at the blogs by big-name economists, one senses that they may not bother if there weren’t the potential for a grand unified theory (however distant that may be from where the theory is now)…