The environments in which we live and operate—corporations, cities, even national economies—are far too complex in their structures and internal processes for any one individual to have a clear, coherent view of how they "work" as total systems—that is, of how particular actions can trigger a cascading series of ever-more indirect effects. At most, what we see and understand—often poorly, at that—are the workings of a tiny fraction of such systems. As a result, we regularly make decisions with little or no appreciation of their full range of consequences. And we also make decisions that, by virtue of their consequences, are not really ours to make. Indeed, lacking a systemic vision of causality, we have no way of knowing why policies—whether corporate or governmental, whether on low-cost housing or real-estate construction or production volumes or whatever—often produce results directly opposite to those intended. But we do know that this happens—all the time. The field of system dynamics provides an organizing framework for analyzing how such policies and decisions interact in complex and often unexpected ways. The field, originated in 1956 by Jay W. Forrester, Germeshausen Professor, Emeritus, at the Massachusetts Institute of Technology, makes it possible for senior managers to assess in advance the impact of different policies on the growth, stability, and behavior of their corporations. In other words, it gives them the tools to design organizational structures and policies that—both in theory and practice—are consistent with their overall objectives.

Before creating system dynamics, Professor Forrester led the research at MIT that produced the Whirlwind, one of the first high-speed computers. He also invented, and holds the basic patent for, random-access, coincident-current magnetic storage, which for many years was the standard memory device for digital computers. In addition, he used his knowledge of computers to guide the planning and technical design of the US Air Force's Semi-Automatic Ground Environment (SAGE) continental air defense system. Recently, he has been applying the principles of system dynamics at the level of national economies to examine the forces underlying inflation, unemployment, business cycles, and the causes of major depressions.

McKinsey: What is driving the current resurgence of interest in system dynamics?

Forrester: We are seeing not a resurgence but a natural consequence of exponential growth. Suppose the number of system dynamics practitioners doubles every three years. If you start with one person, in three years there will be two; in six years, four; and so on. In the 36 years since system dynamics began in 1956, there has been time for 12 doublings. Such exponential growth would bring the number today to over 4,000. Probably at least that many are now engaged in one or another aspect of system dynamics.

This kind of growth resembles that of the pioneering days of digital computers. My first exposure to the field of computation was as a research assistant sitting in a room with the only 30 people in the world interested in computation. (At that early stage, a "computer" meant a young woman at a mechanical keyboard calculating machine.) A few years later, I was presenting papers at other people's meetings on digital computers, then organizing sessions for meetings of growing size, and still later presenting keynote addresses at annual computer society meetings of 1,000 people.

Finally, I went to a computer convention of 20,000 people and found no one there I knew nor anyone who knew me. It was time to get out. So I moved to system dynamics.¹ Sometime in the future, if I go to a system dynamics meeting where I recognize no one and no one has ever heard of me, it will be long past time for me to have shifted to something else.

McKinsey: Is there any way to accelerate the growth curve of a field like system dynamics?

With any field, growth depends on the challenge it poses, its persuasiveness, and its importance

Forrester: With any field, growth depends on the challenge it poses, its persuasiveness, and its importance. Actually, interest in system dynamics is growing faster than the supply of skilled professionals. At present, the real bottleneck is in education for system dynamics expertise, which is being seriously impeded by the inherent conservatism of universities.

Faculties in existing departments tend to favor new appointees with the same backgrounds as their own. New fields, like system dynamics, which cut across the boundaries of existing fields but do not neatly lie within any one of them, lack a home and a supporting constituency. It is like the early days of physical science and engineering, which for a long time were excluded from the established universities. Schools of technology were created to fill the vacuum. Perhaps we should be thinking about new institutions set up especially for research and teaching in physical, medical, social, and economic dynamics.

McKinsey: What do you see as the key philosophical milestones in the field's development?

System dynamics is a process in which the objective is to help people alter and improve their own mental models. Only by changing mental models will one change decisions

Forrester: I do not know of any single milestone that I would call a "major" philosophical change—unless it is the transition of system dynamics from a process in which a consultant builds a model for a client to a process in which the objective is to help people alter and improve their own mental models. Only by changing mental models will one change decisions. People need to understand better how the structures and policies of organizations shape the behavior of corporations and countries.

Operational milestones include the progression of improvements in system dynamics software. Originally there was only dynamo from Pugh-Roberts, running on IBM mainframe computers. The more user-friendly stella and ithink were developed by High Performance Systems for Macintosh computers. For serious professional work with larger models, we now have professional dynamo and other powerful software applications, such as vensim, from Ventana Systems. I have recently been using the latter for my work on the national economic model.

McKinsey: Do you expect senior managers to use such software?

Forrester: Some are already doing so. Others will gradually join. The role of senior management, especially of chief executive officers, should be that of corporate designers and not of corporate operators. CEOs may need assistance with the details of simulations. But they will eventually have to be competent in the creation and use of models if they are to lead such planning activity effectively.

I am told that the recently appointed president of a large European corporation was the first CEO in his company's history to have a personal computer on his desk, and that the first thing he wanted on that computer was vensim software for system dynamics modeling. There are not yet many CEOs who are serious about corporate design. In the future, however, the most effective CEOs will not be those making day-to-day decisions, but those who are designing their corporations—that is, structuring which information is available and to whom, establishing policies that govern decisions, and deciding which decisions should be made where.

Future CEOs will focus on creating corporate designs in which ordinarily competent people can succeed

Future CEOs will focus on creating corporate designs in which ordinarily competent people can succeed. Too often, we see people in corporate positions repeatedly blamed for failure and replaced when the fault lies in the situation into which they have been put. Much of the time, it is the design of the organization that is defective. Unnecessary and distracting information may be supplied while necessary information is missing. Indeed, corporate design often allows, or even specifies, that decisions be made at the wrong places.

Responsibility for decisions should, instead, be carefully made to fit the context of available information, authority, and systems impact

Responsibility for decisions should, instead, be carefully made to fit the context of available information, authority, and systems impact. For example, production managers should not decide production rates and sales managers should not set prices, because these decisions affect the corporate system far outside their own departments.

McKinsey: Why shouldn't a production manager decide production rates?

Forrester: How much to make is tied very closely to sales, pricing, advertising, and inventory—all of which lie outside such a manager's domain. Production managers should decide how to achieve efficient production, how to maintain morale, and how to make good use of machines. Such decisions are in their domain. But when the obligation to make decisions about production rates gets imposed on them, they cannot make them properly. They get frustrated, get criticized, and often get fired for making such decisions incorrectly. At a minimum, their attention is diverted from the decisions that they should make and that only they can make wisely.

McKinsey: Which department should decide pricing?

Forrester: Definitely not the sales department, which has the wrong motivation and lacks a balanced view of the corporate system. When sales of a product stagnate or decline, salesmen inevitably want to reduce price to make selling easier. But reducing price often does not revive sales, which can be limited by many factors other than price—unavailability of product with long delivery delays, say, or poor service or low quality.

Price setting is one of the system policies that ties together everything a corporation does. Because prices interact with production rates, field service, quality, profits, and product design, pricing policy should not be set by any operating department, but instead as a part of the overall corporate design.

The American automobile industry provides a good example of misjudging the role of prices. Detroit believed for two decades that Japan was selling automobiles in the United States on the basis of lower price. So Detroit reduced prices and gave discounts. But this did not increase market share. In fact, just the opposite happened. Market share continued to drop because Detroit took a long time to realize that customers do not buy cars on the basis of price alone. They buy substantially on the basis of trouble-free operation. However, with profit margins reduced by price discounts, the auto companies lacked the resources to fix the actual reasons for loss of sales.

McKinsey: What role do management information systems play in helping people make the decisions that are rightly theirs to make?

Forrester: People who run management information systems tend to make more and more information available in more and more detail to more and more people until everyone is overwhelmed by excessive information. Information pollution may be our most serious form of pollution, because it focuses scarce managerial attention on information streams that are not appropriate to the decisions being made. The whole area of MIS needs to be redesigned using system dynamics, so that simulation models can reveal the effect of having or not having certain information streams. The result, I am sure, would be to discontinue many of the present streams and to supply a few critical information inputs that are not now available.

McKinsey: Does system dynamics support only corporate-level design, or is it also useful in making decisions on a day-to-day basis?

Forrester: System dynamics is primarily for the design of policy. By policy I mean the rules by which decisions are made. By determining policies—that is, the rules for decision-making—system dynamics establishes how day-to-day decisions are to be made. A system dynamics model consists of a structure of policies and shows how the decisions resulting from those policies create corporate stability, growth, market share, and profitability.

People usually react too quickly and do too much. They forget that often the right advice is, "Don't just do something, stand there."

For example, production rate is based on backlog, inventories, production capacity, average past sales, profitability, corporate liquidity and other considerations. But how are these inputs to be combined and interpreted in arriving at changes in production rate? And how quickly should changes be made? The rules used to assign importance to various inputs and to determine speed of response strong-ly influence stability of em- ployment, market share, and profits. In my experience, an intuitive approach to adjusting production level is apt to be quite counter-productive. People usually react too quickly and do too much. They forget that often the right advice is, "Don't just do something, stand there." Only a dynamic systems analysis can accurately show the consequences of a proposed policy.

McKinsey: Could you provide some examples of policies that govern decisions?

Forrester: Policies exist at every point where decisions are made. Some policies cause harm; others contribute to success. In this electronic age, where customers are often repelled by electronic telephone answering with its long menu of touchtone steps before locating someone with whom to do business, it is refreshing to find companies with customer-friendly telephone policies. I know a furniture company in which the policy regarding customer satisfaction requires that every phone call be answered by a person within one or two rings. Furthermore, whoever answers the phone is authorized and expected to completely settle the issue raised by the customer. If people who normally answer phones are busy, those above or below them in the organization receive the call. A caller might get anyone in the company, and that person would settle the question or complaint. Such a policy builds loyal, repeat customers.

The Nordstrom department store chain has policies on customer service that have become legendary. One story involves a customer who complained because an automobile tire, which he said he bought at the store, had not stood up properly. The manager responded by giving a full refund on the tire, even though Nordstrom had never sold anything for automobiles! The story is worth many pages of advertising. The company has been conspicuously successful.

McKinsey: What are the situation factors that make it more likely for a corporate designer to succeed?

Forrester: First, management must believe the organization has a serious weakness. System dynamics can't help people who have no problems. Such people have no motivation for changing the way their organization operates.

Second, people must believe that their problems are largely caused by themselves. At least, they must believe they can change what is causing their problems, rather than blame their problems on others. Once people recognize that they are part of the system causing their difficulties, it is easy for them to become interested in the nature of systems.

Third, there must be a desire to understand what is happening and why, instead of a grasping for superficial quick fixes. To begin to understand "why," one needs a small group of people who have an inquisitive turn of mind and who really want to understand what is going on. The group should include at least one senior manager, preferably the CEO. These individuals must be willing to sit through sessions in which all their cherished assumptions are put on a computer screen and challenged.

Only through such computer simulations will they see that their assumptions often do not lead to the behavior expected

Only through such computer simulations will they see that their assumptions often do not lead to the behavior expected. The next step, of course, is to review and revise their assumptions to get better convergence between proposed policies and desired corporate behavior.

McKinsey: How often do you find managers who do the things you just described?

Forrester: Twenty years ago, very few managers thought in terms of designing a corporation. But the number is growing. Two reasons lie behind the increase.

Many senior executives see that their primary role is no longer to exert direct control, but to educate their people.

First, many organizations sense that things are no longer going well. Consequently, many senior executives see that their primary role is no longer to exert direct control, but to educate their people. They may not be sure precisely what education is needed, but they have stepped away from running everyday operations to lay the groundwork through education so that other people can make better decisions.

Second, many managers no longer find a computer frightening. They have had experience with computers and are willing to apply them in new ways. We are moving toward a time when increasing numbers of corporate executives will be ready to take on the role of corporate designers. To succeed, they must realize that redesigning a corporation will take even longer than designing, producing, and marketing a major new product. Corporate redesign can be a ten-year job.

McKinsey: What kind of companies do the CEOs who view themselves as corporate designers lead, and is there a good time for such CEOs to start a redesign process?

A good time for CEOs to play the corporate designer role is when their companies are very successful. But when a company is successful, people are less likely to become uneasy about the future

Forrester: It is not easy to specify a particular kind of company or the best time to start. One group of companies may not qualify: those that are in deep and increasing difficulty right now. They may not have sufficient time or freedom of action for redesign. In general, a good time for CEOs to play the corporate designer role is when their companies are very successful. But when a company is successful, people are less likely to become uneasy about the future.

A second appropriate time is when a founder-manager who has built a successful company realizes he is not sure why the company has been successful—and is doubtful about others carrying on after his retirement. A third is when a new company is being organized. Many venture capitalists do not understand the ensemble of policies necessary for success. Very often a company is doomed to failure by the initial incompatibility among policies regarding growth, financing, pricing, and product development.

McKinsey: Does the ability to look ten years ahead come easier to CEOs in, say, pharmaceutical and automotive companies, which have long decision cycles or lead times?

Forrester: Willingness to take a long-range view seems to depend more on the person than the kind of company. System dynamics is equally applicable to companies with long and short product cycles.

The automobile industry in the United States, for example, has long product lead times, but has persistently taken a short-range view of decisions. When Detroit resists recalling defective cars, that refusal undermines its long-term reputation for quality. Further, an auto company that identifies dealers as its customers and so produces cars to satisfy dealers, who make most of their profits from a high frequency of automobile repairs, will in the long run prompt its ultimate customers, the car owners, to turn elsewhere to find cars that do not require frequent service.

McKinsey: How would system dynamics help a Detroit CEO?

Forrester: A system dynamics model usually shows how management is creating the problems that a company is experiencing. Very often the policies adopted to solve a problem are actually causing it. Then, as matters get worse, pressures increase to apply the same policies more strongly still.

This leads to the corporate equivalent of the "dead man's spiral" in aviation, which can happen when a pilot without skill or instruments for blind flying enters a cloud bank. The pilot may feel the airplane is diving, so he pulls back on the controls. But the sensation of diving continues, so the pilot pulls back more. In fact, the plane is not really diving but is in a tight turn. The pilot continues to misinterpret what is happening until the stresses from the turn tear off the plane's wings. The same misinterpretation of data causes corporations and countries to reinforce policies that are the source of troubles.

McKinsey: Does a CEO have all the information needed to create a system dynamics model?

Forrester: Usually, everything needed to build an effective policy-oriented system dynamics model is available. The information is in the tremendous database in people's heads. They know who uses which information, which decisions are made at each point in the system, how they are made, and what has happened in the past. I've found that people have pretty clear insight into why they do what they do.

Many times, I have had the experience of going into an organization with serious and widely-recognized difficulties, such as falling market share or high instability of employment. I talk to Manager A about the pressures and reasons for decisions at his location in the company. Then I talk to Manager B about Manager A, and the same picture emerges. In other words, people see themselves and each other in very much the same way. Repeating this exercise across several managers gives a good sense of how the organization operates, what its policies are. A system dynamics model based on those policies will behave like the corporation does and generate the same troubles.

McKinsey: How much does a CEO need to know about technical details of system dynamics to create a good model?

Forrester: Creating a good model requires a certain amount of professional art. A CEO without system dynamics skills should create a model along with someone who understands both the modeling process and the kinds of feedback structures that can create the problems that led to the modeling effort in the first place. Such advisors must be sensitive to the "little things" that indicate where political power lies, where the centers of influence are, and who is afraid of whom and why. This takes special skills. The average manager or someone out of an ordinary business school or consulting firm will simply not know which dynamic structures to look for in conversations with operating managers.

Such advisors must be careful, however, not to dictate what goes into creating a model: the model must come from the corporate participants and be done by them. This is delicate ground because advisors must also focus the managers on issues that matter so as to avoid cluttering the model with unnecessary details. In time, a growing number of CEOs will understand system dynamics modeling well enough to fill the role just described for professional advisors.

McKinsey: How do you know how to avoid the dead ends?

How do you know how to design a successful airplane? Or become a successful manager? Like all the professions, system dynamics is an art based on education and experience

Forrester: How do you know how to design a successful airplane? Or become a successful manager? Like all the professions, system dynamics is an art based on education and experience. There is no guaranteed way to develop a worthwhile model, just as there are no rules to guarantee that everyone can become an effective McKinsey consultant. Success in system dynamics comes from understanding some deep fundamentals about the structure and behavioral dynamics of feedback systems.

McKinsey: Can you provide an example of these fundamentals?

Forrester: There are two—and only two—kinds of variables in a system dynamics model: levels (or accumulations) and rates (or actions). Once you believe that there are only two kinds of concepts in a system, everything you look at has to be one or the other. The idea of two and only two kinds of variables should not be new to managers. Accounting reports are cleanly divided between the balance sheet and the profit and loss statement. Balance sheet variables are system levels; profit and loss variables are system rates, which cause the balance sheet levels to change.

The existence of only two kinds of variables—levels and rates—is true of all systems. Levels—that is, things like number of employees, reputation of the firm, degree of trust within a group, and quality of products—state the condition to which a system has arrived at any point in time. They are gradually built up or degraded over time by streams of actions. Rates, by contrast, are controlled by the system policies that describe how decisions result from system levels.

Every action happens as part of a structural loop in which each decision causes a result that influences future decisions

The critical point here is that levels and rates always exist within feedback loops. A decision causes something—like employment or product quality or debt—to change, and those changes affect the next round of decision making. There is no beginning or end. Every action happens as part of a structural loop in which each decision causes a result that influences future decisions.

McKinsey: Just how important are the system dynamics models for designing organizations?

Forrester: Almost all existing organizations have been created without them. But most people are unhappy with the results. Conventional methods of planning or designing organizations are just not able to deal with dynamic complexity. In mathematical terms, the task of building real-world organizations that work is much more complex than that of solving a tenth-order, nonlinear differential equation. No engineer or mathematician can "solve" such a system by intuition and debate. It is unreasonable to expect managers to do so. It may be my biased perspective, but I see the rigorous application of system dynamics as the most powerful methodology for designing organizations.

McKinsey: How is "systems thinking" related to system dynamics?

Forrester: A clear distinction should be made between general systems thinking and the discipline of system dynamics. The former, systems thinking, is becoming a popular phrase to describe talking about systems, thinking about systems, and observing that systems are important. But, in general, it does not refer to the quantitative and dynamic analysis that constitutes real system dynamics.

Systems thinking, for example, which includes management games tend to focus on decision making; system dynamics focuses on the design of policies that guide decisions

Systems thinking, for example, includes management games, which can demonstrate the existence of complexity and show people that they cannot get the best results from using mere experience or rules-of-thumb. But these games rarely, if ever, carry participants into an understanding of why dynamic behaviors occur as they do. Games tend to focus on decision making; system dynamics focuses on the design of policies that guide decisions.

Systems thinking can, of course, be a door opener for system dynamics. The danger lies in people believing that systems thinking is the whole story. It is not. It is just a sensitizer—something that calls attention to the existence of systems. Some people feel they have learned a lot from systems thinking, but they have gone less than 5 percent of the way toward a genuine understanding of systems. The other 95 percent lies in the rigorous system dynamics-driven structuring of models and in the simulations based on those models. Only these simulations—and nothing else—can reveal the deep inconsistencies within our mental models.

McKinsey: But how do you know that a model you have created to represent a system is right?

Forrester: There is no proof that Einstein's theory is right. There is no proof that Ohm's low in electricity or Boyle's law in gasses are right. There is only an experimental demonstration that such laws are useful for specific, limited purposes. There is no way of proving that a model or law or theory representing the real world is right. The same is true of a system dynamics model. It is a model of structure and behavior that purports to represent something in real life. There can be no proof that it is right. That is not a meaningful question. The relevant question is the degree of confidence one has in a model.

The alternative to a system dynamics model is almost always an inferior mental model that is already in use

The competition for a system dynamics model is not a comparison against perfection, but a comparison against other models that would be used. Remember, all decisions are made on the basis of models. The assumptions in a person's head are not actual systems, but assumptions about actual systems. You do not have a family or city or corporation in your head. You have mental models—often poorly and incompletely defined models—of these real-life systems. The heart of the matter is your relative degree of confidence in each of these models. The alternative to a system dynamics model is almost always an inferior mental model that is already in use. It should be quite possible to develop a system dynamics model of a national economy in which people have far more confidence than they do in the mental models in the heads of elected government officials.

McKinsey: Given the long lead times in designing organizations, how can system dynamics help CEOs in situations where accelerated change is required?

Forrester: Accelerated change is required only for an organization that is already in or near extreme difficulty. Such an organization may be desperate enough to act on models and recommendations that have not yet become part of management's accepted way of thinking. If there is a sufficient sense of urgency and desperation, management may say, "That looks reasonable. We do not have a persuasive alternative. We accept the model."

In some cases, where the conditions have already deteriorated far enough, there may no longer remain policies capable of reviving a company. Failure is inevitable. If so, the sooner the enterprise is terminated, the better. Many banks, airlines, and real estate companies are now going into bankruptcy with results that are less satisfactory than if their weaknesses had been fully acknowledged earlier. A system dynamics model can help identify policies for survival if they exist or clarify the timing for disbandment if they do not.

McKinsey: Is there a "best" entry point or a sequence of actions to make the process of accelerated change effective?

Forrester: We do not really know the best way to accelerate change, but some guidelines have emerged. First, for the most rapid results, the undertaking must have the enthusiastic personal participation of top managers. If change is urgent, there may not be time for a step-by-step sequence of bringing all levels of people into an understanding of the new policies. But even if top management understands and accepts the necessary changes, extensive education may still be required at lower levels before changes can be effectively implemented.

Important challenges are no harder to attack than unimportant ones—and the payoff from success is much greater

Second, start the modeling process with the most serious problem facing the company. Many beginners think that serious problems must be the most difficult and so should be deferred until experience has been gained with minor problems. I have found, however, that important challenges are no harder to attack than unimportant ones—and the payoff from success is much greater.

Third, one must realize that a company may not be able, at first, to identify its most serious problem or even to identify what constitutes a problem. People respond in curious ways to being asked to state their important problems. The answer may be, for instance, "The problem is to train my middle management," with no indication of why they should be trained or why they are not already adequately trained or what the symptoms are that suggest training is needed. One may have to go through several layers of management and many discussions before the real nature of the fundamental source of difficulty becomes clear. People often do not distinguish causes from symptoms or symptoms from solutions. Yet these are very different things.

Fourth, a wide range of people should be involved in the actual model building. Often this selection of participants cuts across departments, layers of personnel, and even company boundaries. Discussions may be needed with customers, bankers, and competitors. Such an extensive network of inputs may not involve a large number of people: the requirement is for capturing key inputs from all the critical parts of the feedback loops that are creating the undesirable behavior.

McKinsey: How does this process of identifying problems and structures actually work?

Forrester: In the early days of system dynamics, my practice was to carry on extensive discussions with people in a company, then go away to pull the ideas together into a model. But such working apart from managers has serious disadvantages. Managers see the result as the consultant's model, not their own. They have not worked through the considerations of what to include and what to ignore. Modern system dynamics practice gets much better response by keeping managers involved at every step of the process. It is even possible to bring a Macintosh computer and STELLA software to a first meeting and put ideas about structure and policies on the screen as people talk. This makes the discussion more explicit and disciplined. The group moves back and forth between discussing assumptions and seeing the simulation consequences. So they develop a much deeper understanding and offer more productive contributions. When a useful model emerges, managers know how it happened and are more willing to accept the relevance of the results.

Such participation does not remove all difficulties. Several kinds of people may identify important problems and structures. Sometimes it is the CEO who has the best view of how undesirable behaviors are being generated; sometimes, it is people below top management. In a company that does not have system dynamics expertise, it may take an experienced consultant to organize available information into an integrated picture that leads to an effective model. But even when shared understanding exists, established traditions and practices often keep a corrective policy from being adopted.

I am reminded of an insurance company in which the reason for loss of market share was understood and the proper corrective action had been identified, but decentralized managers resisted implementation. Claims-related costs were rising, and customers were dissatisfied because claims were not being settled quickly. Several concerned vice presidents saw that more claims adjusters were needed. Without them, under growing pressure for faster settlement, current adjusters were not inspecting ordinary automobile damage, but only phoning customers to ask what the repair cost would be.

Does paying the customer what he asks lead to satisfaction? No. The customer may ask for $700, and the adjuster may answer, "OK. We'll send you a cheque." But the customer is unhappy: he could have asked for $900. If the adjuster had had time to visit the customer and inspect the damage and discuss a settlement, $500 might have seemed reasonable to the customer.

The company's senior executives had written memos to field offices recommending more adjusters, but local managers were sure that more employees would reduce profits. So nothing happened. System dynamics advisors were called in to develop a model and an associated interactive computer game for allocating expenditures in a field office. The model clearly demonstrated greater success in the long run with a higher allocation of budget to adjusters, although in the short run, costs were higher until a reputation for better service could be established.

Even so, the powerful influence of corporate culture immediately became apparent. The executives who had written the memos recommending more adjusters and who had commissioned the computer game to drive home the point, played the game several times without trying the option of adding more adjusters. Faced with the psychological pressures generated by the game, they reverted to traditional company practices.

McKinsey: If you are clever in designing an organization around structures and policies, does everything else take care of itself?

Education for implementation and getting acceptance of the required policies may be a greater challenge than the design

Forrester: No. Education for implementation and getting acceptance of the required policies may be a greater challenge than the design. Implementation is especially difficult unless suitable measurements of performance and rewards are designed at the same time to support the recommended policies. All too often, however, measurements operate against efficiency.

A corporation may have a quarterly quota for production and sales. To meet the quota, the plant works overtime in the last two weeks of the calendar quarter. Then, at the beginning of the next quarter, factory operation is cut back because parts and materials are so depleted by the last end-of-quarter rush that production cannot be maintained. It would be far better to have a continuous, daily-computed moving average of production for the last 3 to 12 months as the goal to be exceeded so that there would be no discontinuities.

Often there are traditions, fears, and influences from outside operating management that inhibit the adoption of improved policies. I worked for several years as an advisor to a company where the symptoms were falling market share and severe instability of employment. My task was to help find a comprehensive system dynamics model that combined production, employment, inventories, customers, competitors, some important policy conflicts between production and finance, and ability to borrow in the financial markets.

The model, derived from information known within the company, showed that intentional, recognized policies were actually causing both instability and declining market share. Moreover, the model suggested that market share had not declined continuously, but had fallen in steps that occurred during recessions. A closer examination of company records showed that this had actually happened. (System dynamics models often reveal surprising behaviors that turn out to be true but had not been previously detected in the real-life situation.) Throughout the company, operating people were conditioned by a fear of being discovered with excessive inventories. As a result, when a business-cycle recession began to develop, they aggressively cut back ordering of materials and production rates in anticipation of falling sales. Those reductions exceeded the degree to which sales would have fallen. Delivery delays became longer, even though the factory was operating well below full capacity. Faced with these longer delays during recessions, customers shifted to other suppliers.

I do not know of anyone in that company who disagreed with the model's assumptions about how employment instability and declining market share were being caused or with the logic of the proposed policies for correcting the difficulties. But there was an almost insurmountable hurdle in the way of implementing the new policies. The corrective action called for reversing policies that three generations of top management had made public speeches about as the basis for their success. All three generations of former management were alive, in town, on the board, and stockholders. Logic has great difficulty prevailing in that kind of setting.

McKinsey: It's 27 years since you wrote "A New Corporate Design," where you suggested replacing the authoritarian power structure of corporations with an entrepreneurial, free-enterprise structure.² Is the elimination of the superior-subordinate relationship that you describe in that article applicable to today's corporations?

Forrester: I believe it is. Many managers argue that centralized control and the superior-subordinate relationship are essential in a large corporation. They assert that a corporation cannot be controlled except through a superior-subordinate chain of command. But our largest economic institutions do operate successfully without that kind of structure. The national economies of democratic, capitalist countries do not coordinate activity by having a superior-subordinate relationship between legal entities. There is no such relationship between a dentist and an automobile manufacturer or between a grocery store and a law office.

Business leaders make impassioned speeches about the advantages of a free-enterprise economic system while running some of the largest socialist bureaucracies in the world

Business leaders make impassioned speeches about the advantages of a free-enterprise economic system while running some of the largest socialist bureaucracies in the world. They have central planning, central ownership of capital, central allocation of resources, subjective evaluation of people, lack of internal competition, and decisions made at the top in response to internal political pressures. These are the fundamental characteristics of a socialist economy. The speeches by corporate executives are right; their practices are not.

Better educated people and computerized information systems make possible very different kinds of working relationships

Many large corporations are now showing the same strains that led to the downfall of the Soviet Union. In fact, corporations are the most pervasive training ground for socialism—they set the pattern of "big brother" taking care of everything, suppressing individual responsibility and initiative. The paper you mentioned needs to be updated, and expanded and filled out with detail, but it does show how a very different and more effective kind of internal organization could be created. Corporations that are already organized at least part way in the direction that paper suggests show the advantages to be derived from a sweeping reconsideration of how to organize economic activity in the modern world. Better educated people and computerized information systems make possible very different kinds of working relationships.

McKinsey: What keeps you occupied these days?

Forrester: Three things: I do some consulting; the largest part of my time is devoted to completing books on economic behavior based on several years of work with the System Dynamics National Model; and I am getting involved in system dynamics as a foundation for a new kind of education in junior and senior high schools.

McKinsey: What kind of consulting do you do?

Forrester: I do some work with corporations when there is a very important challenge of special interest to me. I also occasionally work on dynamics of urban and governmental systems.

McKinsey: Do you find the same counterintuitive behavior and internal inconsistencies in national economies as in corporations?

Forrester: Yes. If anything, economic systems are even more complex and obscure than corporations. Many national policies that are adopted in the hope of solving problems actually make matters worse. As in corporations, the belief in popular policies is often not shaken by failures created by those policies. Instead, people continue to believe in the policies and conclude that the failures came from not executing them strongly enough.

For example, in the United States, low-cost housing programs were created as a response to the decay of old cities. Officials assumed that such housing construction would alleviate hardship. Real-life urban consequences have borne out our modeling of urban growth and stagnation,³ which implied that construction of low-cost housing is a powerful force for driving down the economic well-being not only of a city as an institution but also of its low-income residents who were presumably to be helped. Such housing occupies space that could have been used for industrial job-creating structures while at the same time attracting people who need jobs. It is a mechanism for creating, not alleviating, poverty.

McKinsey: What does your system dynamics "National Model" tell us about present economic conditions?

Forrester: Our work shows two different dynamic modes in an industrial economy—ordinary short-term business cycles and the economic long wave. Business cycles are familiar variations of economic activity with peaks 3 to 10 years apart. The economic long wave, also called the Kondratieff cycle, is much larger in amplitude, with peaks 45 to 60 years apart. These two economic disturbances arise from quite different processes in an economy.

Business cycles are driven primarily from over-production of consumer products followed by cutbacks and labor layoffs while inventories are brought back into balance. The rise and fall of inventories and production of goods occurs over a short span of a few years. Business cycles are small economic variations compared to the rise and fall of the economic long wave, which is caused by over-building of capital plant during several decades of economic expansion, followed by a depression lasting 10 or 15 years while capital-producing sectors collapse and the excess of factories, hotels, and office buildings are worn out and depreciated on the account books. The economic long wave creates—and is accentuated by—large changes in prices, debts, money supply, and real interest rates. I believe we are now in the early stages of a long-wave downturn. The late 1920s and early 1930s were the last time that conditions were similar to what is now happening in economies around the world.

McKinsey: How widely accepted are your views about the economic long wave?

Forrester: Most American academic economists do not believe in the economic long wave. Such skepticism is understandable because there has been no theory for how the long wave could be generated in an economy. The long wave is taken more seriously in Europe. Although, from our system dynamics perspective, the Great Depression of the 1930s was a characteristic long-wave depression, most economists have tried, rather unsuccessfully, to explain it as nothing more than an unfortunately severe business-cycle recession.

We feel that a simulation-derived theory for long-wave behavior now exists. When we started this work we had not heard of the long wave but first encountered it in behavior of the System Dynamics National Model. A huge oscillation with about 50 years between peaks resulted from interactions among ordinary, everyday policies existing mostly in the private sector. However, governments do participate in making the long wave more severe. For example, easy credit during the 1970s, and especially the 1980s, encouraged the over-building that is now leading to a collapse of real estate values in many countries.

We have identified several feedback loops, all related to the building of capital plant, that contribute to the excesses of the long wave. These coupled loops include ones involving inflation, real interest rates, the "self ordering" of capital wherein capital sectors order from themselves to cause further expansion of their productive capacity, and the increase in household income resulting from rising money supply, which in turn results from borrowing to buy capital plant.

McKinsey: Assuming confidence in your economic modeling, are there policies that could now avoid a downturn?

Forrester: The economies in most industrial countries are now so unbalanced that there is probably no magic wand to bring conditions immediately back to a healthy equilibrium. There is an excess of capital plant and production capacity, at least given existing prices and wages. Extreme levels of debt must be paid back slowly or be defaulted. Either alternative produces disruption. The best hope lies in not taking actions that will make matters worse. However, one can make several observations about dealing with the existing threats to economic stability.

First, there is no one to blame. The policies that created this long-wave downturn were supported—in fact, insisted on—by every segment of society during the last two decades. We are all in this together. If one sector of society vents its frustration by blaming another, it will only prolong the period of economic difficulty.

A long-wave downturn is a reaction to past economic excesses, but it contains the seeds for its own recovery. In the meantime, we should do everything possible to alleviate individual hardship

Second, it will help if the public can realize that the economic troubles will not last forever. A long-wave downturn is a reaction to past economic excesses, but it contains the seeds for its own recovery. In the meantime, we should do everything possible to alleviate individual hardship.

Third, to the extent possible, a long-wave depression is a good time to rebuild the infrastructure of a country—roads, bridges, schools, and public buildings. However, some countries, like the United States, have lost the freedom to run a government deficit to pay for such rebuilding because of the debts that were accumulated during the prior economic boom. Excessive expenditure during good economic times is one of several processes by which governments can make the long wave more severe.

The United States has actively engaged in pro-cyclic policies—that is, government actions that accentuate swings of the long wave. By incurring large fiscal deficits during the expansion, the additional stimulus made the expansion more extreme, necessitating a deeper contraction to correct the imbalances. Furthermore, the accumulated debt resulting from the sustained deficits is creating financial and political pressure for a balanced budget at the very time when a fiscal deficit could cushion a major downturn. Other countries may be more fortunate by having maintained fiscal discipline during the expansion of the last two or three decades and by now being in a position to use increased deficits to combat a depression.

McKinsey: Why will the peak of a long wave in the United States not occur at the same time as, say, a valley in Europe, with the peaks and valleys cancelling each other out in the world economy as a whole?

Forrester: Such independence of timing is impossible to maintain. In all dynamic systems, similar modes of behavior lock together with even very slight coupling between systems. The technical term is "entrainment." Very small signals will entrain or synchronize systems that have approximately the same dynamic character. For example, my friends who collect antique pendulum clocks report that, if you have several clocks in a room, all adjusted to keep accurate time, there will probably be enough flexibility in the structure of the room to synchronize the swings of all the pendulums. Indeed, I have been told that when the National Bureau of Standards used pendulum clocks for the basic time standard, entrainment was enough of a concern that two clocks had separate concrete foundations down to bed rock and were then turned at right angles to one another to keep synchronizing forces from being transmitted between them.

In the world economy, there are tremendous forces that lock together the long waves of different countries. For example, daily international money flows are greater than the annual GNP of most countries. One also sees entrainment within a single economy. There are theoretically dozens of fluctuating modes in an economy, but only two are clearly visible—business cycles and the economic long wave. Only these two dominant modes appear because all modes with periods between 3 and 12 years tend to be entrained into the single business cycle, while modes with periods from 40 to 70 years tend to be entrained into a single long wave. Thus, one should expect the expansions and depressions of the long wave to occur approximately at the same time in all countries.

McKinsey: You mentioned involvement in junior and senior high school education. Have you decided to shift to educating the younger generation rather than trying to influence corporations?

I think we will have only limited success in achieving widespread understanding of business and social systems until the ideas take root very early in a person's education and are nourished and reinforced continuously thereafter

I think we will have only limited success in achieving widespread understanding of business and social systems until the ideas take root very early in a person's education and are nourished and reinforced continuously thereafter. System dynamics deals with concepts that have been almost entirely absent in education, even though those concepts underlie everything that happens. We have found that the same teaching materials and methods can be used anywhere from junior high school to chief executive officers.

The systems viewpoint is a paradigm, a frame of reference, a way of looking at one's surroundings, that takes a long time to internalize

The ideas and the way they can be learned are equally accessible, new, and relevant to all age groups. But the systems viewpoint is a paradigm, a frame of reference, a way of looking at one's surroundings, that takes a long time to internalize. Developing such a systems perspective takes less time with a young, inquisitive, and open mind than with one that has already been conditioned to see the world in terms of unidirectional cause-and-effect.

McKinsey: Have any academic institutions adopted systems dynamics as a basis for education?

Forrester: Some dramatic experimental successes have already been achieved in using system dynamics as a foundation to the teaching of mathematics, physics, social studies, biology, economics, environmental change, and even literature.⁴ System dynamics should not be taught as a separate subject but as a common thread running through all subjects. In the United States there are dozens of schools making significant progress with system dynamics and probably several hundred doing something. At a recent conference on system dynamics in pre-college education, 200 people attended from many parts of the country. I have a book of exercises developed in Germany for using system dynamics and the STELLA software for teaching physics in high schools. The Scandinavian countries have a coordinated program for introducing system dynamics in pre-college education.

McKinsey: Can system dynamics be integrated into a traditional class room?

Forrester: The greatest success is coming from a combination of system dynamics and "learner-directed learning," which defines a classroom environment in which students, even as early as age ten, work in teams of two or three on significant projects. The teacher is no longer a lecturer or the source of all wisdom or an authority figure. Instead, the teacher becomes a coach and advisor. The atmosphere is more like that of a research laboratory. Problems are encountered before necessary information has been presented. The project creates the motivation to find the background information that is required.

McKinsey: But is not such an education suitable for only the most able students?

Forrester: It appears that the prior academic ranking of students does not correlate with how well they do in a school organized around system dynamics and learner-directed learning. Some of the students who have traditionally ranked at the top are there because they can repeat information from lectures and books. But they may not truly understand its significance and so may do less well in dealing with dynamic relationships. On the other hand, some of the students who ranked at the bottom in conventional terms because they see school as irrelevant, may have keen insight into how the world around them is working, and be inspired by the way system dynamics allows them to build on the mental models they have already acquired.