The Father of Systems Dynamnics Dies: Jay W. Forrester

Jay W. Forrester has died at age 98, the father of systems dynamics.  He arrived at the MIT to study electrical engineering in 1939 and during WW II would develop the first human-computer interactive system, the flight simulator.  He would invent the magnetic core system of computer memory, long widely used for RAM in many computers.  In 1956 he joined the Sloan School of Management at MIT where he remained for the rest of his career and founded the Systems Dynamics Society that year, which still exists and of which I am officially a member, although I have not been to one of their meetings since the 50th anniversary one at MIT a decade ago.

The main principles of systems dynamics were laid out in his most famous book from 1961, Industrial Dynamics, which would be followed some years later by Urban Dynamics, and then World Dynamics, which provided the program (DYNAMO) and the basic model used by the Club of Rome group in their much more famous The Limits to Growth (Meadows, Meadows, Randers, and Behrens, 1972). Many of both the virtues and flaws of systems dynamics can be seen in that effort, with the benefits being use of an integrated system of nonlinear difference equations to model a large-scale system, along with a tendency to over-aggregation of important variables along with weak statistical or econometric foundations for posited relationships and equations within the model.  In the case of the Club of Rome book, this led to exaggerated forecasts of impending global ecological-economic doom that did not happen, even as many of the problems highlighted in that book remain serious.

My old friend and complexity colleague, Richard H, Day, has long been an advocate of Forrester's work, which also included models of long economic waves in multi-sectoral models, as an early part of complexity modeling via nonlinear dynamics.  Something that Forrester long emphasized, and which is correct, is that such systems of interrelated nonlinear differential equations (to be more general) can lead to "counterintuitive results."  Funny things can happen on the way to Broadway when the world is full of nonlinear relationships, especially when some of those involve positive feedback effects that can lead to destabilizing dynamics in such systems.  Forrester's work dating from the late 50s and early 60s was among the first to warn of such matters, which indeed must be taken seriously, whether we are modeling financial markets or global climate dynamics.

The only time I really had a chance to talk to Forrester myself at any length was a decade ago at the 50th anniversary conference of the Systems Dynamics Society.  He was then fully alert and sharp, despite being 88 years old, and also full of his many strong views of the world.  Something I had not realized until then was the sharp competition he engaged in intellectually with Norbert Wiener and his cybernetics approach, which from my view have many similarities.  I have long listed cybernetics as the first of the "Four C's" of complexity: cybernetics, catatstrophe, chaos, and agent-based complexity.  I have long argued that in effect the last of these, also known as Santa Fe complexity (or small-tent complexity), essentially replaced cybernetics, which also shared the problem of over-aggregation that its rival, systems dynamics.  In principle neither of them needed to.  This may be more a matter of timing and improved computer power over time, rather than something fundamental.  Our modern computers can handle the much greater complexity of modern agent-based models than the earlier computers that the old cybernetics and systems dynamics models ran on. 

I note a curious personal aspect of my relation to all this. The very first paper I ever wrote to get published in the early 1970s while still a grad student was a critique of certain parts of Forrester's model used in his book, Urban Dynamics, which he had used as the basis for some testimony before Congress.  He had testified that the government should not build low-income housing because it would just attract poor people to cities where it was built, which would then lower investment and growth.  His model simulations showed such an outcome, but looking at it closely it was clear that this was driven by an equation that had inter-urban migration being driven most strongly by relative housing prices, whereas virtually all the literature said that it was jobs and wages which were the most important driving variables.  My paper was rejected by the journal I sent it to, with very poor in my opinion referee reports. I expressed a desire to send a letter to one referee to point out his errors, but the editor made it clear in no uncertain terms that he would not let me do this.  As it was, I never sent the paper anywhere else, and it remains unpublished to this day.  Ironically, housing prices may in fact now be beginning to impact migration and employment growth as we see high housing price locations possibly facing growth slowdowns as it becomes too expensive to live in them.

In any case, I think that Jay Forrester's role in the development of modern complexity economics based on nonlinear dynamics has been insufficiently appreciated.  I mourn his passing and recommend his work for those more deeply interested in these matters.

Barkley Rosser