26
TECHNOLOGICAL FORECASTING AND SOCIAL CHANGE 9,309-334 (1976) 309 Mathematical Model Building and Public Policy: The Games Some Bureaucrats Play JACK N. SHUMAN ABSTRACT This paper analyzes the relationships between mathematical model building and decision-making. Special emphasis is placed on the problem of “contextuality”. The primary focus of this review is on the “entity” concept as it relates to this study. A major segment of the paper examines various quantitative procedures as they related to decision-making in the Vietnam War. Additionally, there is a comprehensive analysis of issue quantification as a cultural problem. Finally, several alternative system designs are presented as well as an evaluation of the pros and cons of using mathematical models in public policy. I. Intoduction The purpose of this study is to show that while mathematical models of social and economic structures of the real world can be constructed, the computations derived from these models are of extremely limited use for analyzing the current economic and social situation, and are of even more limited use for predicting the future state of social and economic conditions. This premise is based on an examination of the nature of mathe- matical modeling and the ability or lack of ability of a model to handle more than a very limited number of variables. Further, the accuracy of measurements of the variables used as imput to computation is examined. And, perhaps most important of all-before analysis of the output of the model-is an examination of the reality (and limitations) of our theoretical knowledge of how social and economic variables interact. The thesis of this analysis is that model building around systems with the many variables of social and economic conditions is an exercise in unreality, particularly when the inability to obtain meaningful measurements of many of these variables is considered. The author suggests that other approaches to solutions in these complex areas must be developed. This article (borrowing from the comments of one of the reviewers) is written for those individuals who experience the world as informational and statistical qualities, and in terms of the variables of economics, or “scientific” disciplines, rather than as over- whelming sensations and events which are hard to analyze and describe without losing the essence of experience. It is this group who experience surprises to their professional consciousness from reading this kind of material, or even more importantly, experience DR. SHUMAN is a Lecturer in Psychology at Georgetown University where he also received his doctorate. His area of interest is the social psychology of planning. 0 American Elsevier Publishing Company, Inc., 1976

Mathematical Model Building and Public Policy- The Games Some Bureaucrats Play

Embed Size (px)

DESCRIPTION

produtividade

Citation preview

  • TECHNOLOGICAL FORECASTING AND SOCIAL CHANGE 9,309-334 (1976) 309

    Mathematical Model Building and Public Policy: The Games Some Bureaucrats Play

    JACK N. SHUMAN

    ABSTRACT

    This paper analyzes the relationships between mathematical model building and decision-making. Special emphasis is placed on the problem of contextuality. The primary focus of this review is on

    the entity concept as it relates to this study. A major segment of the paper examines various quantitative procedures as they related to decision-making in the Vietnam War. Additionally, there is a comprehensive analysis of issue quantification as a cultural problem. Finally, several alternative system designs are presented as well as an evaluation of the pros and cons of using mathematical models in public policy.

    I. Intoduction The purpose of this study is to show that while mathematical models of social and

    economic structures of the real world can be constructed, the computations derived from these models are of extremely limited use for analyzing the current economic and social situation, and are of even more limited use for predicting the future state of social and economic conditions. This premise is based on an examination of the nature of mathe- matical modeling and the ability or lack of ability of a model to handle more than a very limited number of variables. Further, the accuracy of measurements of the variables used as imput to computation is examined. And, perhaps most important of all-before analysis of the output of the model-is an examination of the reality (and limitations) of our theoretical knowledge of how social and economic variables interact.

    The thesis of this analysis is that model building around systems with the many variables of social and economic conditions is an exercise in unreality, particularly when the inability to obtain meaningful measurements of many of these variables is considered. The author suggests that other approaches to solutions in these complex areas must be developed.

    This article (borrowing from the comments of one of the reviewers) is written for those individuals who experience the world as informational and statistical qualities, and in terms of the variables of economics, or scientific disciplines, rather than as over- whelming sensations and events which are hard to analyze and describe without losing the essence of experience. It is this group who experience surprises to their professional consciousness from reading this kind of material, or even more importantly, experience

    DR. SHUMAN is a Lecturer in Psychology at Georgetown University where he also received his doctorate. His area of interest is the social psychology of planning.

    0 American Elsevier Publishing Company, Inc., 1976

  • 310 JACK N. SHUMAN

    the agony of seeing their hypothesis more often than not failing to have any validity for predicting the real world. Undoubtedly all professions, including the humanities, have people of limited perception and even less vision. However this paper focuses largely on the narrowly disciplined economists, econometricians, mathematicians, and statisticians and those, particularly in public policy, who have an unquestioned high respect for them.

    II. Some Realities of Mathematical Modeling For our purposes, models are representations of processes describing in simplified

    form, some aspects of the real world. Mathematical models, quoting the Organization for Economic Cooperation and Development, consist of statistical magnitudes articulated into an organic system by means of a number of quantifiable relationships between these magnitudes.

    Mathematical modeling does afford us a unique way of gaining insight into the behavior of limited numbers of variables under rigidly specified conditions. Model analysis is based on Newtons Law of Similarity. In general, this law states that models must be similar to the structures or situations they represent.

    Mathematical modeling has been introduced into public policy largely through aero- space and defense technology, as a sort of universal panacea, a problem-solving tool which plays its most significant role in those areas of public policy involving mans interactions with the natural or physical world. Fields such as medicine, aerospace, and environmental quality obviously depend heavily on an expanding technological base that includes quantification techniques. Most of our pressing problems, including race rela- tions, income redistribution, public service access, and metropolitan development largely defy quantification, simply because they involve the enormously complex and unpre- dictable patterns of human relationships.

    A mathematical model confirms for us, in a highly refined and simplified language, what we already know or should know. The model improves on the accuracy, precision, predictability, and reliability of our knowledge. Consequently, a mathematical model may be much more important for telling us what we do not know about a problem, than for reconfirming our established notions.

    To attain rigor and specificity when we employ mathematical models, we tend to sacrifice variables which are not particularly amenable to any form of measurement or quantification. For example, the book, The Limits to Growth, has been criticized as being so oversimplified as to be fundamentally misleading. The authors base their model on purely physical parameters such as technology, natural resources and population. They largely ignore the price system, the political system, the evolution in values and customs, and other processes of social and economic adjustment. The authors measured only what could be measured. Although we might disagree with their conclusions, we do not yet have any means for conducting the required causal analysis to gain understanding into the complicated forces which mediate the process of existence. There does not exist at present a mathematical language sufficiently complex to effectively describe or predict social processes and relationships, or which can capture the randomness of human behavior.

    Most mathematical modeling, even in the social sciences, is based on the calculus. The calculus, as Peters and van Voorhis point out, offers the distinct advantage of dealing with the relation of infinitesimal increments of one variable to infinitesimal increments of another. However, meaningful measurement or quantification is always related to some sort of physical variable or mechanistic process. An especially interesting type of mathe-

  • THE GAMES SOME BUREAUCRATS PLAY 311

    matics for the analysis of social problems, which offers greater flexibility than the calculus, is the algebra of sets or classes, which is closely connected with symbolic logic.

    The term productivity in economics offers an excellent illustration of a purely quantitative approach to problem analysis. Productivity is partly defined as physical

    efficiency or, the ratio of useful output to total output. Admittedly, the concept is subjective; someone has to define useful. But because of the difficulty in measuring accurately the inputs of capital and natural resources to the processing of wealth creation, productivity-as a convenience-is measured most often in labor input terms, i.e., as output per man hour. The difficulty with measuring productivity in output per man hour terms is the tendency for false inferences from the measure to the conclusion that productivity gains are attributable to improvements in the productivity of labor itself. Gains in productivity occur in roundabout and largely non-measurable and non-quanti- fiable processes ranging from cultural behavior to basic science and educational policy.

    The processes of evaluation, interpretation, and judgment-in sum, understanding-are as at least as important, if not more important than the actual utilization of a specific mathematical procedure. One specific instance is the regression equation. When com- puted, a correlation coefficient can express the degree of interrelation or commonality among systems of variables.

    It is, of course, a necessary, but far too often not obvious precondition that the relationships between systems of variables be realistic and reasonable. For example, in the summer of 1972, the Department of Health, Education and Welfare was asked to spend $100,000 to develop what amounted to a series of regression equations that would indicate a high coefficient of correlation between housing abandonment in ghettos and social services. Clearly, the requisite mathematical relationships could have been devel- oped but they would have meant nothing. Any reasoned social analysis would show that housing abandonment is caused largely by owners who feel that their investments can no longer be profitable, and is unrelated to the quantity much less the quality of social services.

    Regression analysis is an effective method for validating hypotheses. For example, one case examines the factors behind the near-eradication of tuberculosis in the Western world. The National Institutes of Health would undoubtedly measure the diminution of tuberculosis against medical research. The correlation coefficient of medical research does show a high degree of interrelation among diagnosis, treatment, and ultimately cure. A more thorough study indicates that the eradication of tuberculosis correlates much more closely with the revolutionary redistribution of income over the past forty years.

    III. Entities An entity is a modern systems concept. The concept of an entity presupposes the

    coherence of a number of parts (constituent activities) of an organized analytical unit. These parts are characterized by a set of relationships which are thought to exhibit a degree of self-regulation, or to be capable of being made coherent through regulation.

    Entities are the consequence of statistical abstractions. When we speak of entities, we are talking about the design of information systems (statistical or otherwise) with a set of social relationships conceived as a social system or a behavioral entity. This is important because the concept of real world behavioral entities is matched with empirical representatives (however incomplete) of integrated statistical information systems. In Kuhns term, the paradigm that emerges is not so much concerned with the design of a

  • 312 JACK N. SHUMAN

    statistical or other informational product, as with the design and operation of a statistical or informational production process. The informational difficulties in the entity concept must not be minimized. This is particularly true for the collection of statistical data.

    One prototype case in point is the routine and systematic collection by the National Science Foundation of data on the Federal Governments research and development efforts for every three Fiscal Years. These data are submitted in response to question- naires to the Federal agencies. At the outset, this system is in difficulty. Agencies may and, usually, for a variety of reasons, do selectively filter whatever data and information they choose to report. In other words, there is no independent means for evaluating the accuracy or veracity of the submissions. Another serious difficulty with the use of this particular statistical compilation is the matter of interpretation. In the National Science Foundation procedures, each agency, in effect, interprets its own activities by classifying them. For example, in Fiscal Year 1972, the National Institutes of Health reported that their allocation for both internal and external basic research amounted to somewhat over $420,000,000. Yet, in February 1972, the then Director of the National Institutes of Health, Dr. Robert Marsten stated quite categorically and explicitly that the National Institutes of Health conducted no basic research whatsoever. The National Institutes of Health, by statute, are forbidden to conduct research which is not somehow mission- oriented. Thus, the need to resort to some vagaries in statements.

    The necessity for this dissembling does, however, possess serious problems for applying the entity concept. For example, if a decision-maker were concerned with a concrete biomedical research and development policy problem he would obviously want to know how much is basic and how much is mission-oriented. He would derive little useful data and information from the National Science Foundation statistical compilations.

    IV. The Validation of Mathematical Models The meaning of the term validation in social systems is concerned with assessing the

    capability of the system to somehow describe and predict behavior. In physical systems the criterion of validity means that the laboratory system is an accurate representation of a natural system. In social systems, validation refers to a reliance on statistical probability augmented by large overlays of analysis and interpretation. The problem of accuracy and meaning is a far more complex matter in social systems than it is in physical systems.

    One of the more interesting and important examples of the use of proper validation procedures in mathematical modeling is in the area of water resources management. This research represents a significant recent trend in research management by combining, explicitly, natural science, engineering, and economic variables.

    Robert A. Young and John D. Bredehoeft (Water Resources Research, June 1972) have developed one of the better modeling procedures in water resources management. The essential elements of the simulation model include a set of instruments or control- lable variables representing the alternative courses of action, an objective function permitting the ranking of these alternatives, a set of factors considered uncontrollable within the scope of the problem as it is defined, and an empirical model expressing as a set of equations the relations between the central variables and the consequences of their manipulation.

    An analysis of this study provides several useful insights concerning the application of mathematical modeling procedures:

    1. Simulation models do not directly provide an optimal solution to a problem. They

  • THE GAMES SOME BUREAUCRATS PLAY 313

    can, however, be designed to produce a numerical measure for each of a number of alternative courses of action.

    2. Where there are a large number of decision variables, a complete explanation of the response of the system must involve a large number of simulation runs. Sampling techniques can then be applied to select efficiently a probable optimal policy but expressed wholly in terms of the variables and data incorporated into the models.

    3. Simulation studies such as this generate results that always pertain exclusively to the system or systems investigated. One cannot extend these results or even their implications to other systems without extensive reworking. As every situation we encoun- ter is always by definition unique, so correspondingly are any models, simulations, or other attempts at explanation.

    The first matter of concern, from the standpoint of validity, is the internal validity of the model itself. Are there inconsistencies of such a magnitude that they might markedly diminish the effectiveness of the model?

    Next, the model refers to uncontrollable factors. If these factors are uncontrollable, are they equally unpredictable? We really cannot determine, from either the numerical or the verbal expressions of the model, the probable, much less the possible interactions between known and unknown factors. Thus, uncontrollable factors can really be treated in the model only in a hypothetical deductive fashion.

    Professor Julius Kane of the University of British Columbia makes an interesting and important distinction concerning this type of model. He defines it as being arithmetic, since it is concerned with the manipulation of a few select parameters. Without elabora- tion, arithmetic models are explicit, staff-oriented, present-oriented and insular.

    Professor Kanes converse concept of geometric models is that they are intended to gain insight into structural relationships, rather than precise numerical specifications, or predictions and control. Geometric models are implicit, policy-oriented, future-oriented, and global. The geometric model is non-linear, relational, and concerned with communi- cations and imaging.

    There is an additional crucial element in this discussion, which concerns the concep- tualization model. Max Weber, the founder of social modeling, originated the concept of the ideal type, the precursor of the present numerical model. The ideal type in Webers methodology refers to the construction of certain elements of society into a logically precise conception.

    Webers ideal type offered social scientists the use of logically controlled and unam- biguous conceptions directly removed from historical reality. For the optimum use of these ideal types, Weber added two predominant qualities, a feeling of personal accounta- bility or responsibility, and a sense of proportion. These elements are combined into an ethic of responsibility which requires that one account for the foreseeable results of ones actions. The ethic of responsibility is crucial to any individual involved in political or policy analysis, particularly in the development or use of procedures which are little more than sophisticated mental abstractions.

    The foregoing discussion might be taken to imply that models by themselves can never directly provide an optional solution to any public policy problem. The misapplication of numerical models (as expressly stated by the designers) by decision-makers for political purposes is becoming a far too frequent occurrence. There is undoubtedly much abuse by political leaders of what is, after all only, a limited and specific technique. But we cannot merely inveigh against this process, and leave the blame to unscrupulous politicians. To

  • 314 JACK N. SHUMAN

    understand the misuses or abuses of mathematical modeling in public policy, we must also see that model designers (in many instances with honorable motives) themselves are quite capable of leading a political leader into potentially dangerous policy actions, whether drastically curtailed population growth or unwise limits on economic growth. Both rely on our normal psychological tendency to think in terms of limits and our fascination with exponential curves. The basis of such a decision is, of course, data and information (in sum, knowledge), partly in numerical form, partly in verbal form.

    All of us, whether political leaders or private citizens, are fascinated by exponential curves even if we never agree precisely on what they mean. For decision makers, the elegant simplicity of these curves such as those in the Limits to Growth, offers a deceptively straight-forward way of explaining a way out of complexity by extrapolation.

    All of us, whether we are conscious of it or not, set limits to what are, in effect, exponential curves. Only, we never place a limit to a particular curve in exactly the same place.

    A limit is in actuality based on our interpretation and judgment of various types of data and information. We aggregate knowledge-behavioral, cultural, historical, social, economic, scientific, technological, etc.-according to certain patterns on a curve.

    It is essential that any mathematical model be valid in relation to its environment. The model must be isomorphic. It must be identical, similar, or possess some sort of contiguity with other shapes or structures. A model, to possess any intrinsic meaning, must be perceived as it relates to other analogous forms. These points are based on J. Stafford Beers definition of systems as coexisting and mutually supporting levels of models.

    V. The Role of Mathematical Models in Social Problem Solving Any model which begins with the supposition that social problems can somehow be

    forced to analytically tractable shapes and then solved automatically consigns itself to failure. Quoting Ida Hoos, the solution of social problems is never achieved. One does not solve the problems of health or transportation. Consequently, where we start or stop is somewhat artibrary and usually a reflection of resource availability and commit- ment. Public policy does not require that problems must always be solved.

    Model building in itself adds to our existing knowledge base by providing new insights. It only provides us with varying configurations and perceptions of what we already know about a particular problem and limited views of the consequences of alternative interven- tion strategies.

    Modeling simplifies problems. It simultaneously raises the level of predictability and reduces the level of uncertainty about a limited problem. All too often, this is only in our own minds and not in reality. Models cannot always clarify. They can also lead to confused intervention strategies.

    As a case in point, there is little that we can do now, or in the future, meaningfully to model synergistic terror. Here we are dealing with unpredictable groups and individ- uals. Moreover, even while we might develop some reasonably accurate models of known terrorist organizations and individuals, these models would be useless in terms of those we know nothing about.

    The type of models we are discussing presupposes that an individual who commits a criminal act is rational in the sense that he has a plan or gestalt, which interrelates him, his act, and the rest of the world. What is unfortunately all too often developed is a set or sets of psychological profiles developed through psychometric techniques of varying credibility, with limited variables and data of dubious quality.

  • THE GAMES SOME BUREAUCRATS PLAY 315

    VI. Modeling as History Professor Matrin Shubik of Yale University, in his ascerbic but penetrating review of

    Jay Forresters World Dynamics, alludes to a commonality between mathematical model building and historical analysis. Responsible practitioners of both disciplines know that only certain elements or variables dominate events. Social scientists know that it is important to locate the few variables to which the system is sensitive and to measure where practicable the few relationships that count. Eric Jantsch suggests that even relatively small numbers of variables and few alternatives lead to problems of unmanage- able size-n variables with k alternatives yield kn combinations.

    Mathematical models are actually built upon bits and fragments of historical-in the form of data and information-knowledge. They are constituted on the basis of past magnitudes and relationships. Next, they are checked in terms of how well they work as representations of what has been happening.

    Mathematical models preferred by most engineers and scientists are static since they rely on historical data by definition. Used in a policy context, they, however, are becoming increasingly transformed into dynamic models, and as the re.cent profession of growth models these models only project narrow spectra of history into the future, all-too-often as self-fulfilling expectations.

    A major ingredient in the development of either historically-based conceptual models or mathematical models is what Alexander Christakis calls the process of defactualiza- tion. Our perceptions of problems are always grounded on a data and information base, in other words, a knowledge bank. The danger occurs when this knowledge is perceived through the filter of semi-analysis, analogies, and examples. We tend to think in terms of familiar but not discontinuous terms. From this discussion we must see that social problems themselves are thoroughly and inextricably woven into a set of world problems whose solutions are totally beyond our current concepts.

    The English anthropologist John Burton rightly cautions us that there is a significant distinction between an analogy and a model. A model-whether historical, intellectual or mathematical-is a simplification of reality and draws attention to those features in which the observer is interested. An analogy is a means by which some features of reality can be better understood. Consider the analogy, The behavior of states is like the aggressive behavior of individuals. There may be some relation to reality in that states include individuals and these individuals may be aggressive. But this association between states and individuals has to be demonstrated.

    VII. Responsibility: Some Further Thoughts Webers ethic of responsibility assumes an extremely important meaning within the

    context of this discussion: an ethic of responsibility in politics or policy analysis has nothing in common with the Judaeo-Christian ethic. The Judaeo-Christian ethic is purely a matter of an individuals view of the effect of his personal conduct on his relationship with the Almighty. The ethic of responsibility, on the other hand, is purely concerned with the domination of men by other men. Responsibility in policy entails simul- taneously an ethic of responsible ends and an ethic of responsible means. One must assume, of course, that a decision-maker who formulates the latter, is sane and of decent instincts.

    Both ethics refer to the optimal use of responsibility. The ethic of responsible means is optimal decision-making. On the other hand, the ethic of responsible ends, involves optimal analysis, particularly the selection of both appropriate and relevant methodol- ogies and reasonable alternative courses of action.

  • 316 JACK N. SHUMAN

    The essential basis for both ethics are mental or situational models. These are basically reasoned sets of assumptions about the real world and ones capability to change this world in a desirable way, and conversely to minimize actions that might cause unwanted situations to arise.

    The decisions that mental models are concerned with are the articulations and detailed specifications of realizable situations, e.g., desirable social conditions or states. These situations can be goals or objectives. However, our preoccupation with goals and objectives over the past several years has assumed obsessional proportions in many areas of national life.

    The mental model of the policy analyst is a match or translation of the decision- makers mental model. The analysts model must be more complex since it is largely technological.

    A hypothetical illustration of these ethics might revolve around strategic defense planning. Any President of the United States, by himself and through his Secretaries of Defense and State, constantly calculates and revises assumptions concerning the United States interest and capability for strategic intervention in any present or potential conflict situation. To an extent, these assumptions are mathematically derived scenarios. These scenarios can only be useful to a decision-maker if he endows them with his own mental model controls. His own strategic information picture must always be greater than that of his staff. It must reflect domestic as well as foreign considerations, and, economic, social, as well as military factors.

    The late President Johnsons decision to intervene in South Vietnam is an interesting illustration of the interaction of these two ethics. His decision is not arguable on moral grounds. From an analytic standpoint, President Johnsons decision was simultaneously responsible and irresponsible. He did articulate what he considered to be an achievable social state, a stable democratic South Vietnam in the context of real world relationships as he perceived them to be-namely, the foreseeable strategic balance between the United States, China, and the Soviet Union. Additionally, this decision was clearly a product of the prevailing theories of the time, e.g., containment, agrarian reform, strategic hamlet location, limited gradual response, etc.

    President Johnson failed to be responsible in two critical points. The first was most definitely within his control. It was his obligation to ensure that his adversaries were at all times fully appraised of the lengths to which he would go to attain his ends. By not doing so, he allowed doubt to creep into their minds and prolong the war.

    Secondly, and more tragically, was President Johnsons selection of some of the key civilian individuals to conceptualize the war, Robert McNamara and Alain Enthoven. What happened, of course, is that there was not a suitable supporting ethic of responsible means which these two individuals should have developed. As a result, the war degener- ated inevitably into an ethic of responsible ends.

    Alain Enthoven stands out as the eminence gris in the Vietnam entanglement perhaps even more than Robert McNamara. Enthoven developed his version of the sophisticated methodology known as systems analysis. By 1967, Enthoven had come to realize the limited state-of-the-art in systems analysis as indicated in testimony to a Subcommittee of the Senate Committee on Government Operations concerning the Planning, Programming and Budgeting System (Oct. 18, 1967). Enthoven views systems analysis as a quantitative approach to decision-making. Furthermore, he claims that systems analysis is an applica- tion of the scientific method-using the term in the broadest sense. The final ingredient to this methodology is a broadly based inter-disciplinary research program.

  • THE GAMES SOME BUREAUCRATS PLAY 317

    There are a number of inconsistencies in Enthovens thinking. First, many parameters of national policy can be articulated quite specifically, but if they are quantified to any extent, too much meaning is lost. One excellent case in point is morale, an individuals willingness to give his life for his country. At best, all that one gets in this context is a limited surrogate indicator.

    Second, Enthovens claim about the scientific validity of his methodology cannot be substantiated. The true scientific method checks hypotheses that are either freely ac- cepted or rejected in the light of analysis and evidence. Enthovens hypotheses are only objectives derived through the Planning, Programming and Budgeting System. AS such, his analytical exercise is far more Thomistic than it is scientific. One must initially accept his objectives and then only question the selection of the best alternative for their attainment.

    Third, Enthoven does not really delve into interdisciplinary research with any intellec- tualism. Interdisciplinary research in itself is not a science, it is merely an act. The selection of individuals with diverse backgrounds and personalities is a difficult tas. Properly selected and guided the group can deliver a useful product. Otherwise it becomes a Tower of Babel.

    Enthovens methodology is particularly dangerous in one crucial aspect. His insistence on quantification obscured the difference between the functions of systematic analysis and judgment-the functions of the decision or policy-maker. Judgment, of course, is based to a large extent on imponderables, e.g., values, the likelihood of future events, when risks should be taken, etc. Enthovens disservice was his application of a good deal of ingenuity that would often yield questionable ways of measuring and making comparisons and clarifications that were not available at the outset.

    Unfortunately, the cost for these benefits of clarification and comparison is meaning, or perhaps a better term-context. Enthovens insistence upon placing the decision- making process into a highly quantified and structured framework, particularly priority determinations and the concomitant allocation of resources, disguised the real nature of the budgetary process. As Professor Aaron Wildavsky persuasively argues, in The Politics of the Budgetary Process, the formulation of any budget, hence any policy, is always a political process; it is never an economic or technical process. Implementation on the other hand may be economic, technical, political, or any combination or variation of these three processes.

    The highly sophisticated planning methodologies developed by McNamara and Enthoven, as well as by Peck and Scherer in The Weapons Acquisition Process, and Charles Hitch and McKean in the Economics of Defense for the Nuclear Age, attempted to achieve one thing-the reconciliation of an abstract notion of control with its real world counterpart (in effect making the two identical).

    According to modern psychoanalysis, the gross discrepancy between perceived abstrac- tions and the real world leads to schizophrenia. In policy analysis and planning, assuming normal ranges of behavior, this conduct does lead to certain dangerous illusions. The principle actors in this situation had as their ruison d Ctre, the complete civilian control of the military establishment. Unfortunately, the best way to achieve this end is to keep the military establishment out of war.

    Similarly, the entire weapons systems acquisition process has little validity in a purely economic context. The military establishment is requirement-oriented. It requires those resources it needs to carry out a mission or policy, when it wants them, irrespective of cost. Clearly, what was established was the myth, not the reality of control. Effective

  • 318 JACK N. SHUMAN

    control in any organization entails an understanding of the communications process in complex organizations. It is, furthermore, based on the clear recognition of dual patterns of authority and legitimacy, and not bureaucratized and institutionalized modern systems management.

    This discussion may also reflect a somewhat accurate representation of what actually occurred in Vietnam. Rather than one war there were at least eight different wars. First, of course, there was President Johnsons war. Then there were the wars waged by Robert McNamara, the Department of Defense, Office of Systems Analysis, The Joint Chiefs of Staff war (not including the separate services), Walt Rostows War, Dean Rusks and the State Departments War, The Central Intelligence Agencys War and the separate ventures of the Department of Defenses intelligence services. Thus what appeared to the public as one war, was really a plethora of wars. Analytically and mathematically, communications between eight separate elements are impossible to network.

    VIII. Vietnam as an Externality The concept of external economics and diseconomies (externalities) treats the subject

    of how the particular costs and benefits that constrain and motivate a decision-maker may deviate from the costs and benefits of a larger organization. If this organization should happen to be a part of the public sector, these costs and benefits will obviously have a psycho-social connotation. Externalities in the public sector are as much con- cerned with political ecology as economic considerations. In fact, both go hand in hand. This factor is dominant whether the issue is domestic or international, war or peace.

    A social system, with its economic, political, psychological, and sociological sub- systems works in such a way that the wide diffusion of decision-making which is permitted is necessary if complex systems are to operate at all. The vital mechanism (and social institution) that facilitates such specialization is the price system. The price system is an information system that provides users with signals to guide their behavior.

    If David Halberstams arguments are at all valid, this highly specialized system broke down in at least two ways as far as Vietnam is concerned. First, the policy-makers prevented the entry of accurate cost information on the Vietnam war to be communi- cated through the price system. As a result, in the words of Thomas Schelling and Roberta Wohlstetter, one had, by definition, a less-than-perfect communication system, communicating both signals and noise. These were impossible to separate with the most evident dangers for an interdependent system.

    Second, there is strong evidence to believe that defense analysts may have used an externalities approach in attempting to fathom the corrupation problem in South Viet- nam. It would not be difficult to visualize them conducting extensive and sophisticated analyses of the South Vietnamese price structure in an attempt to discern if the system was capable of communicating corruption signals with concomitant strategic oppor- tunities to which they as policy-makers could respond.

    The externalities approach to Vietnam is the problem of determining those externali- ties which are politically or socially undesirable and should be counteracted by public policy measures. We usually think of externalities that operate in a fashion that thwarts the full attainment of broad social objectives and which justifies government intervention. Extensive lists of unwanted by-products may be drawn up in modem societies-from air and water pollution to traffic congestion associated with the automobile. Perhaps a war that was imperfectly thought out and naively fought can also be considered a technical diseconomy of sorts.

  • THE GAMES SOME BUREAUCRATS PLAY 319

    External economies and diseconomies are a manifestation of the fact that, in complex systems, one mans decision or behavior can often have an undesigned impact upon others. The trick in systems design is to establish arrangements by which the mutually interacting and dependent behavior of all decision-makers harmonizes so that the larger system operates in an optimal way.

    The kinds of problems and phenomena we have discussed are not unique to the operation of a private enterprise economy, although they have been most extensively treated by economists in such a context. A large multiproduct corporation, a government agency, a military service, or a university-organizations that may be characterized as closed systems and may be centrally managed to a high degree-have identical problems. Decision-making and authority are necessarily diffused (governments consist of departments and bureaus, armies consist of divisions and squads, etc.). The decisions of many must nevertheless result in some coordinated and mutually consistent behavior. Decision-makers must be constrained as well as motivated. Finally, they must be able to obtain knowledge about their constraints, their opportunities, and the impact of the behavior of others. In varying degrees, discussions of externalities focus on these funda- mental aspects of system organization design and management.

    The problems are basically those of specifying over-all system objectives, measuring effectiveness criteria, identifying and measuring the relevant cost concept, determining the relative merits of alternative information systems (with particular reference to the cost and worth of obtaining and communicating information), and specifying the appro- priate decision rules that should guide individual decision-makers.

    IX. Opportunity Costs: Vietnam War is, by definition, an act of, if not irrationalism, at least, nonrationalism. Even

    though war may be senseless, it still must viewed as one aspect of public policy. If war is governed by the phenomenon of what amounts to external diseconomies, it is also no less governed by the phenomenon of opportunity costs.

    Policy-makers and analysts concerned with the conduct of war are also concerned with the attainment of certain strategic objectives.

    These objectives are the main social benefits of war. The costs of war, like any other public policy, are calculated according to the concept of opportunity costs-the social value foregone when the resources in question are moved away from alternative activities into the specific project.

    According to Professor William Baumol of Princeton University, the actual oppor- tunity cost rate can be calculated for most public programs, usually according to the rate of return on long-term trends. In a narrow sense, Professor Baumol is correct in arguing that a decision to utilize a low discount rate, or to achieve the same results by grossly understating inflationary pressures can lead to wasteful employment of these resources in terms of low yield on investment. However, opportunity costs from a public policy context have more than an economic meaning. Public policy decisions are based on excess benefit over costs, with the benefits being expressed as value, or in this instance, strategic judgments.

    Opportunity costs really must be judged in terms of their contextuality. They are relativistic. In any venture, there are costs associated with any benefit. It is a matter of determining, through the perceptual filter of ones own values, a proper cost-benefit relationship in terms that may be largely non-monetary.

    The Answan Dam is an excellent case in point. It is poorly engineered; it has created

  • 320 JACKN.SHUMAN

    biomedical and ecological problems including significant crises in the schistosomiasis fluke; and its effects may cause Egypt to have serious political difficulties with her African and Mediterranean neighbors. Moreover, the rapid rate of Egypts population growth has already outstripped the irrigation and power output potential of the Dam.

    All these factors could have been visualized to a large extent when the Dam was first conceptualized. This analysis would have predictably indicated a low rate of social return over costs. Yet, into this equation, Gamal Abdul Nasser added the factors of his own personal prestige and what he believed to be the enhancement of Egypts national prestige. Thus, in Nassers view as the ultimate decision-maker, these two value judgments justified the transfer of resources into what was a dubious project. Clearly, investments particularly in the public sector, are not made exclusively accordingly to economic rationality.

    These same factors are also evident in the Vietnam war. On a purely rational basis there was little benefit in relation to cost. Yet, if a decision-maker calculated factors such as national honor, commitment to an alliance, credibility in the face of ones adversaries, and the confidence of his own population, the opportunity cost relationships begin to assume a new meaning.

    Finally, there is a case where, in the conduct of war, opportunity costs have no meaning whatsoever. This issue arises when a nation is fighting for its sheer physical survival. The most obvious case in point is Israel. When a nation fights for its very existence, costs are extraneous. The national leadership can only be concerned with the availability or accessibility of resources. For Israel, existence and survival constitute the ultimate social benefits.

    X. Vietnam: The Social Indicator Because war is a political process, decision-makers must possess accurate and reliable

    information on the conditions of the friendly society, which is the object of their intervention strategy, as well as on conditions within the adversary society. In the case of Vietnam, the information solicited was known as, in large measure, the social indicator. Even in domestic use, the term is not yet clearly defined. It refers to some measure of overall well-being or a good quality of life. It represents an attempt, in Christakis words, to describe with some precision and detail the condition(s) of a society in terms of particular activities and social groups.

    Robert McNamara depended on social indicators or their variations as one of the quantitative guides for his Vietnamese intervention strategy beginning in the early 1960s.

    The use of those indicators were most noteworthy in their failure as meaningful data collection and reporting systems. For example, as early as June 1962, former Secretary McNamara visited South Vietnam and reported every quantitative measure we have shows that we are winning the war. There is, of course, the problem of defining winning. Did winning mean the total or at least the effective destruction of the Viet Cong or did it mean the greater economic, political, and social stability of South Vietnam based on national evolution toward some sort of participatory constitutional system.

    Winning was and always is a mental model rather than a formal mathematical model. In any case our adversaries did not play the same game. Western rules of war, including confrontations with the Soviet Union, are based on chess with a check-mate and, thus, an eventual loser and winner. The North Vietnamese and the Viet Cong war games conversely were based on Wei Chi, a Chinese version of the Japanese game of GO. Wei Chi constitutes the basis of the Maoist strategy. Unlike chess, which is based on an

  • THE GAMES SOME BUREAUCRATS PLAY 321

    end-state situation attained as quickly as possible, Wei Chi is based on a protracted struggle, with no dramatic win-loss but rather shifts in relative advantages and disadvan- tages, particularly in psychological terms. The obvious lesson to be derived from this knowledge is that intracultural wars are much more manageable. Most certainly, the communications gap concerning the rules is markedly lessened. It is simply not good public policy to wage wars where the combattants cannot understand each other.

    By themselves these indicators did give a partial but dangerously incomplete picture. These measures could have, and probably did include the ratio of South Vietnamese to Viet Cong, and consequently North Vietnamese desertions, ratios of equipment, drop or loss ratios, monthly changes in the Viet Cong and North Vietnamese infiltration rate, equipment and supply destruction in over replacement by category, comparative South Vietnamese-Viet Cong-North Vietnamese casualty ratios, etc.

    These ratios, of course, all assume an analytically acceptable ratio. We also know that these measures contained other various but critical social indicators, such as the number of hamlets shifting under the control of either side, changes in the numbers of villagers who assisted or cooperated with either side (loyalty), the success of the Viet Cong psychological penetration in terms of villagers being given individual tracts of land, and probably the conduct on popularity opinion, attitudinal, or support polls of sorts.

    It is also clear that the defense strategists developed an elaborate mathematical sampling system. This system probably included sampling rates of junks searched for contraband with appropriate margins of error as percentages of the estimated total junk population in both North and South Vietnam. These inferences are probably quite accurate in view of certain statements by Walt Rostow on the subject. Moreover, the parameters and boundary conditions of this sort of model can be easily deduced. This system probably also measured highly controlled samplings of hamlets and villages supposedly pacified by region as a fraction of the total. This system never answered two key questions. First, obviously, the quality of the data base itself. Second, the capability of the enemy to thwart or even use the data base for their own purposes.

    This criticism of McNamara and Enthoven should not be construed as denigrating them for being judicious or even fanatical collectors of social statistics. These data were vitally needed. However, unlike purely physical science statistical data, social statistics are not universal. By definition, they are inextricably bound within the matrix of a particular culture and its historical consciousness. Their tragic mistake was in neglecting the extreme conceptual and methodological difficulties in the cross-cultural interpretations of social statistics. In effect, they did not know what data they had.

    These abused statistical data were incorporated into the entire policy process, both in its analytical and decision-making processes, with which, at least in hindsight, are predictable results. First, as already noted, these data were taken and used out of natural context. Second, they were used in a predictive mode, particularly for trend extrapola- tions. The use of any form of social statistics to discern the future is fraught with hazards, because these data, by definition, are always historical. When a user is not even aware of what his data are, the danger is compounded.

    Physical systems, unlike social systems, are governed by certain universally accepted laws, which regularize experimental behavior and results. Thus, if a metallurgist designs an experiment on the superconductivity of metals and specifies rigid boundary conditions as well as the critical parameters, his experiment can be universally repeated with at most only very minor changes. F = ma is immutable in Newtonian mechanics. The data obtained in this experiment are unchallengeable.

    A totally different situation prevails in social systems. For example, an experiment

  • 322 JACK N. SHUMAN

    might hypothetically be designed to sample in real time the factors governing reading achievement in children. If brain measurements could be taken while the children were reading, psychologists might examine in partially quantitative form parameters such as auditory and visual response, perception, and motor skills. Presumably this data could be fed into a computer for instantaneous analysis.

    What useful data would really result? First of all, the experiment is not really repeatable, even minor changes in the boundary conditions or variables would entail different results. Secondly, the notion of real-time in this context is totally specious. Each participant in the experiment is a human being. He is bound by historical circum- stances, owed in part to his unique genetic characteristics, which partially shapes the cognitive and emotive aspects of his personality. The other aspects of his personality are conditioned by his sense of values, or his information coding matrix derived from his external environment in terms of culture and history. Therefore, unless those investi- gators were prepared to somehow account for these factors, their experiment would be questionable.

    What Secretary McNamara and Alain Enthoven did was to place these indicators in a PPB model. Without going into further discussion, each of these indicators was included in the PPB model as a variable. Furthermore, estimates of cost-outlays and proposed benefits were probably attached to these variables which were then, in turn, related to the strategic goals and programs enunciated by President Johnson.

    The fundamental difficulty with the PPB model is that it is based on an econometric type of thinking. Therefore, as a purely economic methodology, it is explicitly and intrinsically grounded in maximization. The problem in social systems is that one must assume that the entire population tends to maximize its perceived benefits from an individual rather than a group basis. Maximization thus becomes a series of individual strategies. The result is the inevitable social trap, brilliantly depicted by Garret Hardin and John Platt. PPB, unfortunately, offers both a superb closed methodology and the distinct disadvantage of facilitating the accelerated achievement of situational social traps.

    The entire PPB rationale offers a strong inducement for these entrapments. Its Jesuitical structure, in reality a hierarchy of goals and objectives provides an initiate with a surprisingly broad range of options once he accepts the truth and veracity of very general premises. This situation is actually what occurred during the course of the Vietnam intervention strategy. An individual in a decision-making capactiy at a fairly high level thought that because he interpreted the truth correctly, he could apply his policies in any way he saw fit.

    The PPB system, or any other quantitative system for that matter, was and still is incapable of directly measuring the all-important intangibles of national life. These systems could not measure the elan of the South Vietnamese population, their national consciousness, their inner-felt loyalty to their government, their willingness for self sacrifice in a common cause-in other words, the total national infrastructure.

    The best that could be obtained were dubious surrogates. For example, how could one measure the problem of corruption? In brief, elaborating some of the discussion in an earlier section of this paper, the following conceptual-mathematical model was probably developed. First of all, the term corruption presents cultural-linguistic difficulties. Corruption to an American may not necessarily have the same connotation as it does to a South Vietnamese. Therefore, some sort of corruption correlation coefficient had to be devised. Next, there is the problem of data, which are suspicious. They came by definition from the South Vietnamese. Next from what sources? If the price system of

  • THEGAMESSOMEBUREAUCRATSPLAY 323

    any national economy does function as an information system, it might be theoretically possible to gauge corruption pressures such as inflationary pressures in the price system as functions of changes in the price system. That is, if one knows how much of the price level is affected by corruption, a suitable methodology might conceivably be some sort of innovative one-dimensional Phillips analysis. It is easy to see how one can concep- tualize a PPB approach to the problem of corruption if its dimunition is given the status of a sub-goal.

    The problem here is one of incorrectly applying multiplexing procedures. It is necessary to simulate various channels-honest and corrupt-and sample the data travel- ling within each channel. We must assume that a major deterrent for the American situation was, of course, that the Americans and the South Vietnamese may not have been acting either in concert or as allies in this instance. Rather, they probably perceived each other in terms of an adversary relationship. How could we test the data being transmitted through such channels?

    Undoubtedly, some indicators, such as the ratio of arrests to convictions for corrup- tion, were developed. The increased ratio between equipment and supplies off-loaded and deliveries was another. Changes in the number of corrupt officials to honest officials may have also been used. It was also probably at least theoretically possible to simulate, using the PPB model, the degree of black market activity as a function of price structure.

    These various and sundry indicators were (or in reality always are) surrogates of many phenomena. Giving them meaning was and always will constitute the other two phases of the evaluation pro.cess-interpretation and judgment. One can only wonder what might have happened if the Spartans at Thermopylae had had a computer processing their situational models into PPB models.

    XI. Responsibility-Domestic Style The need for synchronous situational models on the part of the decision-maker and

    the analyst is also quite evident in several of the Federal agencies having R & D contract and grant programs. Many of the contracts and grants given out by these agencies have been of rather dubious quality. The reasons for these awards can be explained in several ways, e.g., cybernetics-the quality of information in organizational communications channels, statistics-poor awards may just be random samplings.

    If random samplings in grant awards are accepted as a viable explanation, it may only prove the existence of some sort of higher or lower limits of acceptable competence on the part of some of these organizations. That is, even if they do give out sub-standard contracts and grants, the ratio may be lower than in some other organizations. Accord- ingly, they may actually be more cost-effective in terms of a lower competence ratio. Conversely, a random sampling approach could be used by an effective organization, to identify a reasonable ratio of high-quality contract or grant awards. However, whether one is concerned with establishing limits of competence or quality, by any method whatsoever, the key factor is still judgment, or the lack thereof.

    This paper, on the other hand, ventures to offer the suggestion that one should look at how these organizations view their roles. If their intent and practice are some sort of beneficial change or remedial improvement, then they are searching for large-scale, dynamic system-wide improvement-in one word-reform. If conversely, their intent is more limited, one can rightly accuse them of doing nothing more than of having an elaborate series of policies all of which are designed to assure their self-perpetuation.

    A contract or grant organization may decide that it is, in fact, some sort of problem-

  • 324 .JACKN.SHUMAN

    solving organization (whatever the term means, usually Ida Hoos content ot the term suffices). In this instance, it would endeavor to develop what one might call a First-Rate Information System (FRIS) for short.

    This system might conceivably attempt to articulate simultaneously achievable and desirable future social states. Certainly, the attainment of these states would have to be dependent on technology in its broadest context. Mental models would essentially be sets of assumptions of the scope and need of technology. Technical models of various sorts would establish key variables, structural relationships and intervention strategies. In between these two modeling states-other conceptual models could be used to develop simulated projections of the relevant state-of-the-art, and profile matches with indi- viduals and organizations against particular technology areas. Most importantly, this roughly sketched system could provide scales of performance expressed linearly or logarithmically.

    Judgment is necessary both for the array and presentation of quantitative data and its interpretation. A computer simulation model cannot determine what is first-rate. This determination is the province of the analyst and equally or, even more, of the decision- maker who alone can really determine first-rate, as an arbitrary but hopefully analyti- cally substantiated value judgment.

    The process of establishing, if one desires to have one, a second-rate or even third-rate information system (SRIS, TRIS) is identical with a first-rate information system. The only variables are the quality and base of the analysis and the quality of the individuals making the appropriate value judgments.

    Unfortunately, the latter type of information systems abound in public policy. They may occur at times as deliberate machinations. However, more often than not, they are just managerial manifestations of John Platts networks of systems-wise social traps, which are essentially the end result of mans preoccupation with self-generating chains of short-term rationality.

    It is clear from Platts analysis that supposedly new innovations such as the National Institute of Education in the Office of Education will be ineffective in optimizing American education in the. sense that whatever the Institute does will not achieve long-term basic structural reforms. If for no other reason NIE like most organizations takes a macrosystems approach without first understanding microsystems inter- actions. Rather, NIE will probably pursue mainly short-term quick-fix, routines and standardized policies. Some of its grants and policy decisions are already based on SRIS and TRIS systems.

    Of course, NIE does suffer from two difficulties. First, it is part of the Office of Education. It has had to assume many of the existing R & D programs, contracts, and grants of the Office of Education as well as personnel from that office. These factors alone could ruin any organization. However, to compound these difficulties, the National Institute of Education is modeled on the National Institutes of Health, making its tasks even more difficult. Whatever NIE does, it is not science.

    XIII. A Behavioral Overview The intent of this section is to give a behavioral overview to the type of individual who

    both possesses and puts to practical policy-relevant uses, the narrow restricted outlook described in other sections of this paper. It is obviously difficult to characterize this personality in terms of a single or even various traits of character. However, for the purposes of this paper the term policy Philistine will suffice.

  • THEGAMESSOMEBUREAUCRATSPLAY 325

    Because policy Phil&tines exist-whether as individuals or as large aggregates, from the standpoint of cultural anthropology, they can be visualized as forming a cultural sub- group, as do members of any ethnic groups. Members of any sub-group always ahere to the following common features:

    1. Myths 2. Symbols 3. Initiation rites 4. Kinship relations

    Futhermore, being a separate sub-group, policy Philistines can and do clash with other sub-groups. Obviously, it is these clashes between sub-groups which causes internal strife and instability, which in their most exaggerated form produce civil war-the most destructive type of conflict. It is quite probable that differing sub-groups can have so much cultural and historical variance that true communication between them is no more possible than it was between the Americans and the Vietnamese.

    Professor Richard Gambino of Queens College observes that the exaggerated, exces- sive, and extra-systematic or systemic-destructive behavior of any sub-group is a form of tribalism. In his study, Professor Gambino was alluding to the phenomenon of ethnic tribalism. Our rather peculiar sub-group may be said to indulge in quantitative tribal- ism.

    The great pity is that few cultural anthropologists pay any attention to contemporary urbanized society. Rather, they seem to spend most of their time studying what they regard to be primitive societies. In their own narrow, distorted way, policy-Philistines, who collectively are quantitative tribalists, are at least as primitive as the most remote tribe in New Guinea.

    Whether or not these people possess abnormal character traits or personality disorders is beyond the scope of this paper. Clinical psychologists or psychiatrists such as Freud, Binschager and Strauss would, of course, look for signs of neurotic obsessive-compulsive behavior. Similarly, social psychologists might look at this over-attentiveness to quantifi- cation as a form of territoriality. What this paper attempts to point out, is that a policy Philistine does not need to exhibit pathological behavior. It is only necessary that the results of his endeavors be pathological in terms of their social disbenefits, especially when one thinks of these persons as being a collectivity.

    Thought and action are based on three perceptions. These are the perception of knowledge, the perception of reality, and finally the perception of the relationship that occurs when knowledge is applied to reality. In any individual, there is a certain natural degree of distortion among these three perceptions. This sort of distortion is perhaps more accurately a divergence between the mental image or the ideal and the true.

    Recently, perhaps in the past 20 years, a second element of distortion can also be observed-the computer. The present state-of-the-art of computer technology is not readily adaptable to any natural language, whether linguistic or symbolic. For example, a mathematical model must be placed into the context of an artificial computer language before it can be tested on a computer. This procedure clearly results in additional distortion.

    Fortunately, we can (or at least we should) recognize and compensate for these distortions. In this sense, however, they are not pathological. As an hypothesis, but one possessed of considerable validity, it is probable that the most severe distortion is within the mental processes of this rather odd collectivity-the policy Philistines or quantitative

  • 326 JACKN.SHUMAN

    tribalists. Their analyses have been inaccurate and misleading, perhaps because of an abnormal divergence between their perceptions of both knowledge and reality, and by extension, the application of this knowledge to real world problems. It is this extreme divergence which makes their collective behavior pathological in terms of its results.

    The difficulty of meaningful policy analysis is largely cultural-linguistic. We have not yet devised a suitable language for performing this task. A new language, of course, can be either verbal or numerical-mathematics is as much a natural language as English.

    The way we attempt to comprehend social systems offers an excellent illustration of this linguistic deficiency. We derive our conceptualization from the successful results of the laboratory control of experimental variables. And, the discovery of the laws of classical mechanics have given rise to the belief that physical science methods can help discover the basic laws of social systems. In physical sciences, the criterion of system validity is that it works, which means that the laboratory system is an accurate representation of a natural system. In other words, the system possesses the capability of describing and predicting behavior, or validation in some other sense. The complex interrelationships of politics, history, economics, psychology, and sociology have nothing like the empirical justifications of the phenomenology of the classical atoms and mole- cules of physics.

    Let us take two concepts of systems analysis to demonstrate this conceptual difficulty. First, is the boundary conditions or limits of the system. In social systems these concepts have little, substantive meaning. One can say that the domains of any social system are characterized by complexity, diversity, entropy, and resonance (or its con- verse, dissonance). These dynamics condition the behavior of a social system but they do not circumscribe its behavior, as we are conditioned to think about the meaning of limits in physical systems.

    The second term is the parameter. In physical systems the parameter is a constant since these systems are bounded. In social systems on the other hand, there are many boundary conditions. Therefore we are concerned with the variables which are amenable to at least some quantification.

    Social systems may have parameters. However, the presence of constants in a phe- nomenon as chaotic, dynamic and unpredictable as a social system is a matter of question. Most certainly, there are variables present in social systems, and these can be partially quantified. The process of meaningful quantification in social systems, which are by definition unstable, is extremely difficult. It calls for creativity, imagination, patience, diligence, and above all, the most profound skepticism on the part of a policy analyst. Here, the analyst acts in an intuitive fashion, as did most of the great figures in the history of science.

    Individuals accustomed to studying physical systems are used to two basic procedures. First, determine the boundary conditions or limits of the system. Second, identify the major parameters.

    This second procedure is quite difficult as far as social systems are concerned. For example, in education, an obvious quantifiable parameter is the school age population. A non-quantifiable parameter would be pupil attitudes. It is difficult at best to know which parameters are quantifiable and which are not.

    Unfortunately these parameters do not exist in an ordered way in social systems. Physical scientists are accustomed to visualizing neat, well-bounded systems, possessing highly structured systems of parameters. Such is not the case in social systems.

    The ferment in the Arab world offers an excellent illustration in point of isolating

  • THE GAMES SOME BUREAUCRATS PLAY 327

    parameters. The Arab world is presently going through an extraordinary social revolution which has essentially two interrelated aspects-modernization and secularization. The former can be quantified to a large extent by including variables such as increases in per capita income, greater educational attainment, improved longevity, industrialization, etc. Secularization on the other hand is virtually impossible to quantify. It represents a radical transition from the value structure of a 1,000 year old scholastic and traditional society to the value structure of a technological society. This aspect of the Arab social revolution is the most wrenching since these two value systems are totally incompatible.

    Social systems do, of course, possess parameters, some of which are quantifiable, and some of which are not. However, in social systems these parameters are linked almost as closely as the sections of the doublt helix of the DNA and RNA molecules. The point of this discussion is not to draw useless analogies. Clearly, there is no comparison at all between these two phenomena. Rather, the intent of this discussion is, in fact, sensory. If innovative policy thinking is to be predicated on Thomas Kuhns notion of the paradigm shift, then a sense of vivid but disciplined mental and visual imagery is critical to devising an innovative focus for intersensitive and intercultural problems.

    The mental and visual image of a doublt helix may also be important in another respect. If social systems in any way resemble double helices, then their self-organizing characteristics may in fact be templates. This conceptualization might well go far in explaining social inertia and cultural lag.

    Again, we are faced with a severe semantic difficulty. We know that part of this double helix comes under the general nomenclature of a parameter. How then do we meaning- fully describe the other parts of this double helix? They are certainly not moral imperatives. More than likely they fall between behavioral dynamics and institutional vestments. These factors include inherent powers, futurism, populism, and moral urgency. To these we can also add diffusion of power, pragmatism, the relativism of values, gradualism, institutionalism, process, procedure, legality, and technicalities. These vestments may and obviously do vary in nature depending on a particular social system. However, their existence in whatever form is about the nearest thing there is to a constant in social systems analysis.

    John Warfield of the Battelle Laboratories wisely suggests that in the search for any kind of methodology, it is valuable to seek historical perspective. One of the more useful works in terms of its impact rather than its intellectual ability is Thomas Kuhns 7he Structure of Scientific Revolutions. This work, as Warfield points out, shows how at any given point in the development of science there is a set of paradigms that has temporary acceptance. Most of science, including its application, is carried out under the prevailing set of paradigms.

    There paradigms are also twofaced. On the one hand, they furnish the knowledge and hence the freedom to conduct scientific research and science-based problem-solving in a professional way. On the other hand, they constrain us because most of us have to live with the limitations of the paradigms.

    Kuhn is correct as far as his analysis goes. Most of the physical scientists such as Maxwell and Einstein effected only one revolutionary paradigm shift in their careers. Unfortunately paradigms in social policy shift radically. In a creative and innovative policy analysis several paradigm shifts take place simultaneously-namely the continual recontextualization of knowledge. Using Kuhns ideas we can gain some insight into certain of the relatively simplistic quantitative modes of policy analysis now in vogue.The most well-known, or at least advertised, of these procedures is the growth exponential.

  • 328 JACKN.SHUMAN

    As Daniel Bell points out, any growth that is exponential must at some point level off or we would reach a point of absurdity. In the measurement of the growth of any phenomenon that shows patterns of saturation, the questions revolve around the defini- tion of that saturated state and the estimation of its arrival date.

    The exponential pattern described is a sigmoid or S-shaped curve in which the rate below and above is often quite symmetrical. Because this is so, it lends itself easily to predictions, since one assumes that the rate above the mid-point will match that below and then level off. In fact the beauty of this curve, as Bell characterizes it, has seduced many statisticians into believing that it is the philosophers stone for the charting of human behavior. The two most recent examples of this behavior pattern are Forrester and Meadows, i%e Limits to Growth and the Zero Energy Growth (ZEG) scenario in Exploring Energy Choices of the Ford Foundations Energy Policy Project. No curve of any type ever fully takes into account the growth of the human spirit, the development of technology, social complexity, evolution, mans ever-present capability to control his own destiny.

    The difficulty with any proposed saturation of social change is that such curves are plotted for at best, a few (and usually only one) variables and presume a saturation. But what may be true of beanstalks, or yeast, or fruit flies, or similar organisms where logistic growths have been neatly plotted in fixed ecological environments, may not hold for social situations where decisions can be postponed (as in the case of births) or where substitutions are possible (as in the case of bus or subway transit for passengers) so that the growths do not develop in some fixed way. As Bell argues, it is for this reason that the use of logistic curves may be deceptive.

    It was suggested earlier that meaningful paradigm shifts in social policy are the result of a process of continual recontextualization of ones internal knowledge base. The purpose of the paradigm shift is to gain perspective, imagery, proportion and above all validation. Another way of stating this purpose is that a paradigm shift, in essence, controls a shift in our perceptual filters and hence our value structures.

    A policy Philistine can belong to any intellectual discipline-the humanities, the physical sciences, or the social sciences. Some of the more interesting policy Philistines can be found in economics and its offspring econometrics. Economists and econometri- cians have evolved some complex and intricate mathematical systems for explaining economic change.

    The difficulty occurs when some of these economists and econometricians lose sight of the limitations of their methodologies. They attempt to extend their generally accurate short-term models into long-term models of economic growth. There are a number of difficulties in this transition.

    First, it is difficult, if not impossible, to assign quantitative relationships to the impact that education, science, and technology have on economic growth. Second, many economic theories in current vogue are static in the sense that the technological and social framework within which fluctuations in quantities, prices, and incomes takes place is taken for granted, or regarded as a residual factor.

    Penetrating the facade of residualism, we can see that education, training, innova- tion, research, and development are even by themselves only part of an infinitely diversified, and complicated process, which is at best only partially quantifiable. To arrive at even a partial explanation or interpretation of economic growth, we are forced to combine the standard methods of economic research with those of psychology, sociology,

  • THEGAMESSOMEBUREAUCRATSPLAY 329

    anthropology, history, philosophy, decision, and organization theory, and political science.

    The term residual does allow the economic-type Philistine the opportunity to play out his scenario. Where he can identify or isolate a quantifiable variable, he can develop a differential equation. Physical scientists are aware that even a limited physical model is so complicated and diverse that they can never hope to develop the full range of differential equations to characterize their system. By definition, social systems can only be ex- pressed by a virtual infinity of differential equations. Hence, the policy Philistine resorts to the cul-de-sac of the exogenous or residual factor.

    Economic or econometric Philistines and allied model builders are often accused of having no sense of humanity in their calculations, This accusation is only partly true. These Philistines can easily handle humanity. It is just individuals and people with which they cannot deal. Their abuse of the Phillips curve is one example.

    There are relationships among prices, wages, and employment at least as far as demand-pull inflation is concerned.

    There are also serious definitional questions inherent in the Phillips curve as well as the availability of data. In large measure, the Phillips curve is a specification problem in the econometric sense of the word. That is, the act of defining the variables to be considered in the Phillips curve also has implications of how the equation in question is to be specified regarding the appropriate sets of weights to be assigned to each variable.

    The Phillips curve if used by itself in policy planning can lead to serious complications, since it tends to give a simplified view of inflation and unemployment. In human and political terms, higher rates of unemployment never constitute an acceptable tradeoff.

    While the Phillips curve offers an explanation for demand-pull inflation, it does not provide an explanation for cost-push inflation. Hence, the Phillips curve cannot account for the present situation of high-inflation and high unemployment. Most importantly, the Phillips curve does not take into account the changing nature of the labor force. A Phillips analysis would only treat this sort of factor as a shift in a curve.

    As an alternative approach, it would be far more preferable to marry the Phillips relationships with a broad and expanded manpower development policy, which would reach the entire work force. In the long run, there is only one strategy offering truly a comprehensive view of coping with unemployment. According to this strategy, the surest way to reduce the natural rate of unemployment is to eliminate the reasons for it, which are primarily the workers uncertainty about alternative job opportunities and their inability to retrain for new jobs.

    There is no cheap, non-complex or simple way of increasing any countrys production or employment. The inflation side of the Phillips curve seems to offer an answer, but that is illusory. At best, the extra output is temporary. It is much more likely to be offset by less than normal output and unemployment sometime in the future.

    There are most certainly quite a few economists who have recognized the need to build models of varying levels of aggregation out of the information processed by individuals. These models reflect both diversity and heterogeneity. Diversity is unfor- tunately suppressed in most econometric models as a price for data consistency.

    Edgar Dunns Resources for the Future statistical entity approach built on cybernetic techniques is one such innovative approach. Similarly, Kenneth Bouldings classification of one form of knowledge as folk knowledge, is quite useful as he relates these categories to specific social problems. John Wartield suggests (using Boulding as well as

  • 330 JACK N. SHUMAN

    Dunn and John Burton) that it is helpful to think about the international system as being dominated by folk knowledge processed by individuals acting out multiple and simulta- neous roles. To paraphrase Warfield, the chief merit of these thinkers is that they challenge man to enlarge the dimensions of the social sciences.

    Some outstanding applications of model building have recently been made by Dr. Alexander Christakis of the Academy for Contemporary Problems. Using Warfields mathematical matrices and sets, Dr. Christakis has developed a system that analyzes the linkages of a contemporary problem set. While his structure is hierarchical in terms of its heuristics, it can also be used in a system of non-hierarchical heuristics. Problems affect each other and are linked to each other as complex communication or feedback chains which we do not always know to be hierarchical. All we really know is that they are interdependent, sometimes in hierarchical aggregates and sometimes not.

    Warfield has pioneered in the creative application of mathematics as a language in policy analysis and planning. His set and matrix form is based on hierarchical principles, based on his belief that there are subordinal relationships in social systems. One does not necessarily have to fully accede to this opinion to see that Warfields use of what he terms a system matrix offers a means of visualizing complexity in social systems.

    At first glance, the approach of Warfield and Christakis seems a deceptively simple use of mathematics. They give a meaningful and, for a policy-maker, useful perspective on social complexity. The simplicity is their avoidance of the calculus, which would only give a mechanistic and deterministic, model, which they have assiduously sought to avoid. Symbolic logic can be, and is, an equally useful mathematical procedure for establishing complex and systemic configurations.

    XIV. Econometrics: A Form of Cultural Destruction Several years ago, the Organization for Economic Cooperation and Development

    requrested a number of leading economists to develop econometric models for the educational systems of some less developed countries including Greece, Italy, Spain, Portugal, Turkey and Yugoslavia. The group was headed by Dr. Jan Tinbergen of the Netherlands who was later to become a Nobel Laureate in economics.

    The group used a Keynesian multiplier model. As its study pointed out, Keynesian models have the advantage of the clarification they bring to some basic properties of the mechanisms in the economic system, particularly supply and demand. In essence, the approach was an attempt to apply the method of input-output analysis to educational planning.

    To be fair, Tinbergen himself noted that erroneous results are obtained unless one clearly understands that macromodels only provide a first approximation to reality. Unfortunately, this caveat is not only lost in too many instances by too many econome- tric&s, but more importantly, decision-makers are often oblivious to its implications particularly for long-term planning.

    Tinbergens model suffers from the key conceptual and methological flaw common to

    1 The author is thoroughly familiar with the work of the Organization for Economic Cooperation and Development in educational planning. His doctoral dissertation, An Analysis of the Science Policy Programs of the Organization for Economic Cooperation and Development (OECD), focused on OECDs ability to deal with the relationships among economics, education, and science and technology. In view of his research as well as his continued issue-tracking of OECD, the author is qualified to comment on OECD in this critical fashion.

  • THE GAMES SOME BUREAUCRATS PLAY 331

    mathematical models. Only those parameters which can be estimated statistically have been introduced into the various models in this study. Even certain theoretically refined concepts and relationships, e.g., teacher-pupil feedback relationships which bear directly on the models, have not been included because they cannot be translated into numerical estimates.

    The problem is not that these models do not work. On the contrary, in their own quite limited way, they work far too well, with results that are often unfortunate. For all of its supposed clarity and sophistication, this sort of thinking is unrealistic.

    Perhaps one of the best ways to illustrate this sort of muddle headedness is to translate one of these models into a so-called word problem. This word problem obviously is simplistic, but it does demonstrate the logic or illogic of Professor Tinbergens model.

    Question: If it takes one painter eight hours to paint a room, how long will it take thirty-two painters to do the same job. 7 Answer accepted as correct within the context of Professor Tinbergens model: Fifteen minutes. Simple algebra. The true answer is that thirty-two painters are going to get in each others way. Professor Tinbergens models foster a kind of thinking which seriously holds that if technology in any form including econometric models causes problems, then more technology will solve these problems.

    Professor Tinbergen claims that these models are a large part of the basic long-term decisions in educational planning in terms of labor-market policies. His models give the impression of being dehumanizing. No one denies that educational planning must have a sound economic basis. However, in their approach, Professor Tinbergen and his colleagues are oblivious to the fact that education throughout history has an intrinsic, inner value to an individual apart but yet linked somehow to economic gain.

    The work of Professor Tinbergen and his colleagues in Greece offers ample illustration not only of the lack of realistic applicability of these models, but equally, the supreme benefits of having a receptive audience. Professor Tinbergens group began much of its indepth work in Greece about the time of the coup detut and continued throughout the rule of the military junta. Moreover Professor Tinbergens work in this phase shifted from hmited concern with education to a broader involvement in Greek economic and social planning and development.

    This change of emphasis is crucial in terms of Professor Tinbergens approach. Paraphrasing the Swiss economist Professor Eugen Beuler, economists such as Professor Tinbergen are grounded in linearity. Being extrapolative, linearity is regular, pre- dictable, rational, and mechanistic. The product of this form of thought is a plan which isolates and quantifies the major aspects of reality according to the scientific method. Later deviations of real processes are either not distinguished or not noticed from planned ones.

    Psychologically or methodologically, Dr. Tinbergen would have little to offer. A tough but open-minded social planner or thinker would reject this approach outright. Con- versely, a young, nationalistic but limited Greek colonel, who nevertheless regards himself as a social planner or thinker of sorts would find Professor Tinbergens planning meth- odology to be highly acceptable for what he considers to be its cardinal virtue, its simplicity, without any regard to its deceptiveness.

    Professor Tinbergen and his colleagues themselves admit that their proposed develop- ment model for Greece depends only on the validity of the economic postulates and the

    * This word problem was suggested by Dr. Arnold Mysior, Georgetown University.

  • 332 JACK N. SHUMAN

    accuracy of the equations which characterize the Greek educational system. They do not answer the question of how this system can be characterized in ways other than by equations.

    Basically, this group developed a capital/output ratio growth model. The main assump- tion in the model is that there is a systematic relationship, which they do not adequately define, between certain aspects of manpower stock and national income. They did mention, but glossed over, the difficulty of obtaining reliable statistical data, which dominate the validity of any model.3

    Governments too should be able to face up to the Tinbergen models discrepancies, particularly the role of qualitative factors in non-measurable factors, if they are at all concerned with second-order effects. First-order effects are by definition economic. The economy always first feels the effects of change. Second-order effects, which are psychological and sociological in nature, appear later on, but are usually of longer duration. Clearly, ill-conceived economic planning can and will produce the most disrup- tive second-order effects.

    XV. Conclusion This paper has attempted to demonstrate the complexity of determining cost-benefit

    relationships in public policy and some of the conceptual methods used to elucidate these relationships. Clearly, one of the most widely used conceptual methods, but rather primitive and rigid in view of our present state of intellectual development, is the mathematical model.

    While the mathematical model is not as useful in policy analysis as one might wish neither is it thoroughly useless. To put the issue in simple terms, one does not throw out t