The Limits of "Urban Dynamics"
Much has been written in the past two years about URBAN DYNAMICS, a book which applies systems analysis to the problems of growth and decay in our cities. Liberals have damned it because it questions the value of their cherished programs for helping the poor; some libertarians have championed it because the book's conclusions—though arrived at by different means—seemed to support their opinions about the usefulness of government programs. Although I will discuss these conclusions in this review, my primary focus will be on the author's methodology. The author, Jay W. Forrester, Professor of Management at MIT, initially applied his modeling techniques to corporations in INDUSTRIAL DYNAMICS and most recently to global problems of population and resources in WORLD DYNAMICS (see "Counterintuitive Behavior of Social Systems," REASON, July and August 1971). URBAN DYNAMICS deals with an area of intermediate scope, the urban community.
Notwithstanding well-intentioned criticisms of the whole idea of "modeling" and "social systems," carefully thought-out descriptions of socio-economic relationships (i.e., models) deserve attention in that they provide a systematic way to think about complex relationships and even to predict future developments. Dr. Forrester's approach (if not his specific methodology and early conclusions) gives promise of being developed into a useful scientific modeling technique. Thus, although much could be said pro and con about the urban model's conclusions, it would be more profitable to concentrate on improving the general modeling approach. Forrester acknowledges the preliminary nature of his work, in referring to URBAN DYNAMICS as "more an opening of a subject than it is a package of final results and recommendations. The primary objective is to improve our understanding of social systems." Despite this caveat, however, he then goes on to draw conclusions, citing his model as the authority.
MODELING
In brief, what does the model depict? The urban community is defined as a system of three interacting variables (or levels, as Forrester calls them): industry, housing, and people. The specific land area of the city, a limitless surrounding environment, and the city's relative attractiveness to people control changes in the levels of industry, housing, and people. The initial values of land area, family size, etc.—quantities routinely measured by social scientists—are assumed at the outset. In addition, the fundamental relationships by which the levels of the different variables change over time are also hypothesized, in Forrester's a priori approach.
Much as in the electromechanical systems Forrester originally studied as an electrical engineer, a set of initial values is assigned to the "level variables," and these, together with the structural relationships and parametric values, are used to compute the changes in industries, housing, and people over some period of time. The model's equations produce a series of changes in these variables over time until the values of industry, housing, and people stabilize, i.e., reach a constant level. Forrester calls this 250-year process "growth and stagnation" and claims that the model explains why our older cities are stagnating.
Having done this, Forrester then goes on to examine four typical government programs aimed at improving urban conditions, "implementing" them with the model by simulating their effects when applied to the model set at the stagnant condition. These programs are 1) creating jobs for the underemployed, 2) training the underemployed, 3) subsidizing welfare and education, and 4) building low-cost housing for the underemployed. All four programs give unsatisfactory results and, after some short-term improvements, make conditions worse than before. After exploring some other policies via the model, Forrester concludes that urban revival requires demolition of slum housing and its replacement with new business enterprise. (These conclusions should come as no surprise to readers oriented to laissez-faire and noninterventionist policies.)
Those who are interested in the details of these results should read URBAN DYNAMICS. The results have been reviewed extensively elsewhere (e.g., FORTUNE, November and December 1969); indeed, it has been the results that have attracted the attention and provoked the controversy, despite Forrester's admonitions that the results can be no better than the structure of the model and the data that went into it. Readers who wish to use these results in support of previously-held convictions must be familiar with the underlying assumptions and model structure if they are to be able to defend their position. (And it should be noted that Forrester's model is quite different from the relationships described by both Jane Jacobs and Edward Banfield, whose policy recommendations are nonetheless somewhat similar to Forrester's.)
METHODOLOGY
What, then can be said of Forrester's method? In all his work he uses, without improvement, the original closed-loop level and rate methodology developed for industrial dynamics 13 years ago. (One suspects that perhaps the investment in computer programming precludes changes that would significantly generalize the models.) The methodology stems from modeling techniques used for deterministic electromechanical systems. It deliberately does not employ strucutral and parametric data which could be developed by empirical observation, principally in the social science area. One might say that Forrester's philosophical approach is rationalist rather than empirical (or in sum: that reason amplified by the computer can explain the external world). Because of this, Forrester's work is open to strong criticism by social scientists, whose modern-day work is almost entirely empirical.
Forrester's reply to this criticism is that social scientists measure parameters (static values of isolated parts) rather than structure (dynamic relationships between parts), and since parameters do not control system behavior, as structure does, he is justified to some degree in ignoring their work. Relationships or structure can be quantitatively determined on an empirical basis if you know what you are looking for, namely the system dynamic model, not extant social science models.
REDUCTIONISM
In emphasizing structure over parameters (i.e., dynamic relationships among parts resulting from thier interconnection rather than static measures of the parts) Forrester breaks with the prevalent characteristic of most modern scientific disciplines: reductionism. Reductionist scientists examine parts separated from the whole or parts existing in a constant or controlled environment. In the real (nontextbook) world, parts do not exist separate from one another or in unchanging environments; they are connected to each other and changes in one part affect others, causing them to change also. (A bit of reflection demonstrates that parts related in one way behave differently from the same parts related or connected in other ways: if you took your TV set apart and put it back together connected differently, you could hardly expect it to operate the same way as before.) In this sense there is more to the whole than the sum of a set of parts. What also exists is the method of interconnection, and there can be many different possible sets of connections for any given group of parts, each set becoming a separate system. The set of connections, as well as the parts, gives the system its properties and its characteristic behavior.
Forrester states that "complex social systems bring together many factors which by quirks of history have been compartmentalized into isolated intellectual fields." Although he never refers to reductionism by name, it is that that he and a few other pioneers are challenging (see, for example, the lead editorial in SCIENCE, 9 July 1971). Isolated intellectual fields are not quirks of history; they are representative of an operational philosophy of science which is widely held and will be staunchly defended. Since these isolated fields constitute "intellectual territories" (cf., Robert Ardrey's TERRITORIAL IMPERATIVE), they will be defended by interested specialists who will resist generalists' integrating them into structured system models. This may help explain Forrester's reluctance to use social science data in his work, preferring to be criticized for antiempiricism rather than be accused of intellectual claim-jumping. However, he goes on directly to attack reductionism by saying "the barriers between disciplines must melt away if we are successfully to cope with complex systems.…The interactions among these are often more important than the internal content of any one alone. Yet if these separate disciplines are isolated in our study and in our thinking, the interactions will never come into view."
FEEDBACK LOOPS
What does Forrester propose as the structure of social systems? He depicts variables, such as industry, related to one another in complex "loops" wherein a change in one variable causes changes in other variables, and so on throughout the system, until eventually there is an effect on the original variable. Hence the term "feedback," as each change eventually feeds back upon itself. Forrester finds that complex systems have the following characteristics:
1) counterintuitive behavior,
2) insensitivity to change in many system parameters,
3) resistance to policy changes,
4) possession of a few influential pressure points which can cause changes in system behavior,
5) compensation for externally-applied corrective efforts,
6) long-run reactions opposite to short-run, and
7) tendency to low performance.
An understanding of most of these characteristics can be gained from a simple example familiar to all of us who live in a developed culture. Consider a building served by a powerful heating and air conditioning plant controlled by a thermostat. Suppose an intelligent individual from an underdeveloped culture found himself in the building and noted that it was too hot while the weather outside was cold. He would take the obvious action and open a window. Suppose further that the window is near the thermostat and the resulting draft of cold air strikes the thermostat (of which he is totally unaware). Now the powerful heating plant comes on and brings the temperature in the vicinity of the thermostat up to its setting. The rest of the building will get hotter instead of cooler. Our individual will find the complex building system behaving counterintuitively by internally compensating for externally-applied corrective measures. Its long-run reaction will be in opposition to the short-run, and our individual will be at a loss as to what to do next.
In the terms of Professer Forrester, the room temperature is inside the loop and any changes are removed by the homeostatic action of the heating and air conditioning system. It is necessary to find the "influential pressure point" which, in this case, is the thermostat. In general this is a variable which is not inside the loop, i.e., one which when changed affects other variables but which is not affected by the changes in these variables. In our simple case a change in room temperature can be caused by a changed thermostat setting, but the room temperature does not in turn cause the setting to change. The setting is outside the loop.
Forrester's models, while far more complex than the one just described, contain no variables outside the loops. (In every case a change in a level changes a rate which in turn sooner or later changes the level.) Thus, they are like modeling the temperature-controlled room without considering the fact that the temperature is controlled by the thermostat setting. It follows that the behavior he discovers must have the characteristics that it has. Epistemologically, the complex system characteristics noted by Forrester are characteristics of the model rather than of the real-world system. The control features of the system have yet to be found and added to the model in order to find a way to make it change its performance to one that is more desirable, rather than just reacting to maintain its present equilibrium. Forrester's present models are only a start, as they show only the performance of the variables in the loop.
There is reason to believe that the variables outside the loop which control the behavior of social systems can be determined on an empirical basis, by using computers to analyze the performance of actual social systems to determine the system structure. Once this is done the real-world variables outside the loops can be located and the necessary action taken to modify the situation from what now exists to one that is more desirable. (A description of such an approach is beyond the scope of this review.)
Thus, future extensions of Forrester's work offer the hope that we can determine the changes in our social and economic institutions necessary for individuals to lead lives of their own choice in acceptable environments. Much work of a fundamental nature remains to be done. However, a significant start has been made to exorcise the old tradition of reductionism and to see the social interconnections which affect all of our lives and the changes of our environment. The next step is empirical determination of many of these interconnections and identification of the variables or relationships that can act as pressure points to bring about desired changes in our social systems. Hopefully, libertarians rather than statists will be the first to discover and utilize such pressure points, to bring about a more free society.
William P. Patterson is a management consultant with many years of experience in systems engineering, marketing, planning, and organizational design. He holds a BS in mechanical engineering from Lehigh University, and both MS and PE degrees in electrical engineering from MIT. In addition to his consulting work, he is currently president of Terraqua Products, Inc., in San Pedro, California.