Seeking a Robust Model of Disequilibrium
The Santa Fe Institute has just published a new volume called Complexity Economics, which collects together the talks and panel discussions from a symposium held in November 2019. The book is edited by Brian Arthur, Eric Beinhocker, and Allison Stanger, and is available for just ten dollars in paperback and just three as an e-book.
I was at the meeting and can confirm that there were some fantastic talks and conversations involving prominent scholars such Matt Jackson and Scott Page and practitioners such as Katherine Collins and Michael Mauboussin. Reproduced below is an extract from my own contribution to the proceedings, with minor edits and links added.
Modern economies are characterized by path dependence: seemingly insignificant early decisions or minor random shocks can have cumulative effects that become highly consequential over time and can change the course of history.
This is also true of the study of economies: fortuitous exposure to an idea early in one’s academic life can alter the course of a career. I can vouch for this through personal experience.
One of my favorite activities as an impecunious graduate student was browsing through the shelves in the basement of the iconic Strand bookstore in New York City, where review copies of recently published books are still sold at deep discounts. It was there that I happened to chance upon a slim volume called The Economy as an Evolving Complex System, edited by Philip Anderson, Kenneth Arrow and David Pines, based on a workshop held at the Santa Fe Institute.
The chapters in this book were all concerned in one way or another with the economy, but the contrast of their vision with that found in standard textbooks was stark. It was as if these chapters were describing an entirely different and vastly more interesting object. Where the textbooks featured mutually consistent plans unfolding along equilibrium paths, this volume opened my eyes to genetic algorithms, rugged landscapes, disequilibrium dynamics, and more generally processes of adjustment in complex environments.
There are many famous and familiar names among the contributors to this book, but the chapter that had the most enduring impact on me was written by someone who is rarely mentioned today. This was the Brazilian economist Mario Henrique Simonsen, and the title of his chapter was “Rational Expectations, Game Theory, and Inflationary Inertia.”
Simonsen was a former finance minister of Brazil, and very familiar with the practical difficulties faced by policy makers. His chapter involved mathematical reasoning applied to a very real policy problem: how to deal with hyperinflation. He started off with an abstract question: when might we expect Nash equilibrium to be a predictively useful hypothesis in game theory? He classified games into those in which it could reasonably be assumed that players would coordinate on equilibrium strategies, and those in which such coordination would be unlikely; he called them A and B games. This was where I first encountered what today would be called a beauty contest or guessing game; in this chapter it was called half-the-average.
Simonsen argued that Nash equilibrium would not be predictive in this game, and he was completely correct. About seven years later, Rosemary Nagel published a classic experimental paper in the American Economic Review that established this decisively, and set in motion a large and still vibrant literature on alternative solution concepts in the theory of games.
But Simonsen’s concern wasn’t just with abstract theorizing, he was concerned with more practical issues. There was a debate raging at the time between those who favored gradualism and those who argued for shock therapy in addressing high inflation. Shock therapy involved a drastic cut in the rate of growth of the money supply coupled with a public announcement to shift expectations. The argument was that inflationary expectations in the economy would jump down discontinuously and actual inflation would drop without the need for a sharp rise in interest rates or mass unemployment.
This argument depends on complete and uncritical acceptance of the so-called rational expectations hypothesis, which states that the subjective probability distributions held by agents in the economy are self-fulfilling, in the sense that they match the objective distribution to which they give rise. As Herbert Simon pointed out as far back as 1978, this hypothesis does not “correspond to any classical criterion of rationality,” and the label accordingly provides it with “rather unwarranted legitimation.” It is an equilibrium hypothesis, asserting the mutual consistency of individual plans, rather than a behavioral assumption akin to rational choice in consumption or production decisions.
Simonsen’s point was that there are situations in which the consistency of beliefs assumed by equilibrium theory can lead one astray in dealing with important practical problems, and that this necessitated an explicit consideration of disequilibrium dynamics.
The point remains valid but largely neglected to this day. Economic theory continues to lean heavily on a notion of equilibrium that requires massive coordination and consistent expectations involving millions of economic agents, without specifying a plausible process that could conceivably allow such coordination to arise. In fact, as far as the graduate school curriculum and the content of the leading journals are concerned, economics has moved towards even greater reliance on equilibrium methods and even greater resistance to non-equilibrium reasoning.
Why, despite some obvious problems with the equilibrium approach, have explicit models of disequilibrium dynamics fail to take hold? A cynical reason is that the profession is stuck at a local peak on a rugged landscape, with no incentives for anyone to explore distant terrain. But there is a more valid reason that is worth contemplating. Disequilibrium models in general—and agent-based computational models in particular—are too easy to construct and too hard to evaluate. There is really no accepted methodology that allows an editor or referee to clearly and convincingly distinguish a robust model from a fragile one.
Missing is a canonical model of disequilibrium dynamics that can serve as an exemplar for graduate students and young researchers, in much the same way a classic paper by George Akerlof opened the doors to a generation of work in the economics of information.
It is worth recalling that Akelof’s paper was rejected by three journals before eventual publication. Somewhere out there, perhaps yet to be written, is a paper that will do for the economics of complexity what Akerlof’s paper did for the economics of information. When that sees the light of day, the floodgates will open, and the promise of The Economy as an Evolving Complex System will finally start to be fulfilled.