Do We Need an "Economic Manhattan Project"?

Following a suggestion by mathematician Eric Weinstein of the Natron Group for an Economic Manhattan Project," back in December, Mike Brown, Stuart Kauffman, Zoe-Vonna Palmrose, and Lee Smolin posted "Can Science Help Solve the Economic Crisis?" weighing in favorably on Weinstein's proposal at on the edge. They criticized various assumptions of neoclassical theory and called for complexity science, agent-based models, evolutionary modles, and gauge invariance models to be used for non-equilibrium dynamics modeling, as well as analysis of policy alternatives. Criticism by Nassim Taleb, Michael Shermer, Emanual Derman, Paul Romer, and others can be found under "The Reality Club" at that link. Ronald Bailey of reason weighed in with further critiques, calling the whole thing "inane", with others commenting there as well.

On March 2, Tyler Cowen posted on "Lee Smolin on General Equilibrium Theory". This dealt with some of the related issues and dragged in me making some critical remarks, with replies by both Eric Weinstein and Lee Smolin, who is a quantum gravity theorist at the Perimeter Institute of Theoretical Physics in Waterloo, Ontario. As a result of that, I was invited to participate (and did) in a conference held May 1-4 at the Perimeter Institute on "The Economic Crisis and its Implications for the Science of Economics." The participants included some of the original critics, such as Taleb and Derman, along with a variety of people from physics, math, computer science, evolutionary biology, accounting, finance, and economics, with some of the better known others including Nouriel Roubini, Andrew Lo, Doyne Farmer, Leigh Tesfatsion, Richard Alexander, and Richard Freeman. To the best of my knowledge, no overall summary of what happened there has been posted anywhere, but it was very intense with many ideas and tough arguments going back and forth (the institute has blackboards for walls in much of it, which get a lot of use). So, attempting to go below the fold, I shall try to summarize some of it.

The conference was in two parts, a much larger public conference on May Day (for which Freeman wore a red shirt), and a much smaller workshop the next three days, with some of those speaking on the first day not participating in that ("Dr. Doom" Roubini and "Black Swan" Taleb had to go scare some finance ministers in Singapore, and Lo and Derman also departed). Also, I left after only a bit of the last day, when those still standing were trying to summarize and put forward a way to go forward (something will come out of this, but I am not sure that "Manhattan Project" will describe it). I note that the workshop included some lectures (including one by me on May 3), along with some breakout group sesssions.

Anyway, rather than a blow by blow of who said what when, let me note some major issues and positions. So, one big thing that there was a lot of agreement on was the likely superiority of agent-based modeling in some form or other for modeling non-equilibrium economic systems. A leader there for this was Leigh Tesfatsion, who maintains the main website for collecting agent-based models in economics. She called for this being done in macroeconomics to at least supplement the current DSGE models that dominate the basements of the central banks. Doyne Farmer also seconded this strongly, and is apparently building one with Robert Axtell. A curious aspect of this debate was that I learned later from Leigh that someone in attendance for part of the time was the computable general equilibrium modeler, John Whalley. However, for whatever reason, he chose not to make any comments at all on anything, and did not wear a name tag either.

A related issue, which may well be a major focus of more immediate efforts coming out of this conference, was what to do about the fact that most agent-based models are written in different languages and that it is not easy to link up between them. This is related to broader issues that had some of the computer scientists there worked up, as well as the estimable Leigh Tesfatsion, about interfacing between different computer languages in general. Also brought up in this discussion was the problem of data availabilty, with some saying that crucial data really is publicly available, but unknown by most, while others disputed this. There may be some push in these areas coming out of this.

I would note that there were several presentations of specific agent-based models in the workshop. One that was rather nice was by Alexander Outkin from before the decimalization of the stock exchanges showing that this would not necessarily stabilize things as forecast.

There were a lot of discussions of econophysics, with Doyne Farmer of the Santa Fe Institute providing a good summary of the state of play in that controversial arena. More cogent to this conference is the argument advocated by Lee Smolin, drawing on more specific work by Eric Weinstein and Pia Malaney, along with some post-docs at the Perimeter Institute, about applying gauge invariance theory to economics. Gauge invariance is an idea that floats around in some of the efforts to obtain a general unified theory of cosmology, particularly putting together general relativity with quantum mechanics, with the Perimeter Institute being a center of those who question string theory for achieving this, and Smolin a leader of this group.

Anyway, this was the matter that I had criticized, initially quite strongly, over on marginal revolution. I still hold to some of my criticisms, but also feel that I have not seen or fully understood all that there is to this argument. So, part of it seems to be a rediscovery of the wheel, in this case, the theoretical superiority of Divisia indexes for measuring values over time when relative prices and quantities are changing, with the claim being made further that this can also apply to a world of changing preferences, with the ability to chart cardinal utility over time, assuming that it is meaningful to talk about that. My problem with the former is that most economists know this, but that it is applying Divisia indexes in practice that is the problem as they assume continuous time, whereas empirical reality for actual indexes comes to us in discrete chunks not always all that close to each other. I am unsure about the utility argument.

An application was also made to financial markets by Simon Vasquez. He and Simone Severini also presented what was supposed to be a model of non-equilibrium dynamics. I would not say that this achieved something that Weinstein says can be done, which is to use gauge theory methods for measuring the curvature of fiber bundles to measure the degree of out-of-equilibriumness of a system. I think this latter would be really useful, and maybe it can be done, but I did not see this being fully achieved yet from what was presented there.

A rather looser end that I did not see the end of, as the summary from the relevant out-session was not presented before I left, was the application of biological or evolutionary ideas to all this. One disappointment was that co-organizer Stuart Kauffman did not attend, who is a biologist associated with the Santa Fe Institute, and whose work has been applied at times in economics. One presentation in the main workshop was by Kelly John Rose of PI on interpreting input-output matrices from an ecological perspective due to Robert Ulanowicz, that of ecological ascendancy, with looking at economic sectors as ecological trophic levels. He also said that gauge invariance was relevant to this, although that was not shown clearly. However, more generally, the ecological or biological arguments tended to be more on the side of this workshop.

There was more discussion of them in the main presentations the first day. So, Andrew Lo spoke of the adaptive market and how investors change their views over time in a market so as to lead to instabilities, arguably a fancier updating with neuroeconomics arguments of Minsky and Shiller. Also, the noted evolutionary biologist, Richard Alexander, spoke about various issues in the evoution of human beings and relations between kinship and sexual selection. However, he added little to the main workshop discussions, and in my private conversations with him he expressed reservations about people inappropriately using the "language of evolution" outside of more strictly biological evolution. So, while he is fine with discussing how moral systems evolved with humans over long periods of time, he was not particularly happy with people talking about firms or technologies or market forms evolving.

I am going to close this by saying that it was one of the most stimulating conferences I have been to, real clashes of serious ideas while people were willing to speak with mutual respect. I am perhaps sorry that Whalley did not speak up to defend more orthodox approaches, but, well, I do not blame him for keeping his head down. I do hope that it does lead to some further developments, and I think the critics who think this is all going to lead to proposals for central planning or whatever are barking up the wrong tree (see the discussions by Bailey and others).