Uncertainty in Organizations

by

John Day

27 Jan 87

revised 10 Dec 95

One of the primary effects on organizations is uncertainty. Very few organizations ever have complete knowledge about themselves, their competitors, or the environment in which they exist. This uncertainty must be factored into all aspects of planning.

Most traditional theories of organizations (especially those that try to be rigorous and prescriptive) assume perfectly rational behavior and complete knowledge, the typical simplifying mathematical assumptions. These deterministic theories might perhaps be useful as academic or experimental tools; however, they have been used to reorganize corporations, set world development policies, and organize ministries in third world countries, often with devastating results. The reason for the failure of these deterministic theories is simply that there are too many things for which we never have perfect, or even near perfect, knowledge.

This failure of traditional theories led eventually to the development of a theory that predicted the behavior of organizations operating under high degrees of uncertainty. The theory was orginally developed by Cohen, March, and Olsen to describe the behavior of task forces or committees. That work was then extended by Padgett to apply the same concepts to more traditional hierarchical organizations. Cohen, March, and Olsen, in what has become a classic paper on the theory of organizations, conjectured that organizations operating under a high degree of uncertainty or ambiguity can be modelled as a stochastic process -- or, in their words, as a "garbage can" process.

The garbage can model bears a strong resemblence to the random walk model of the stock market. The random walk model, rather than trying to identify trends and to make predictions from complex causal models by massaging mountains of data in great digital masticators, simply observes that there is sufficient uncertainty in the true value of a stock that its daily fluctuations in the market can be modelled as an independent random variable. (This theory has been tested by numerous studies reported both in the academic and popular literature, and tends to predict market behavior better than the more complex models.)

The garbage can model is similar, in that the behavior of an organization is subjected to a high degree of uncertainty and is modelled as a stochastic process. Cohen, et. al. characterize ambiguity in an organization by three attributes:

1. Problematic preferences -- "the organization operates on the basis of a variety of inconsistent and ill-defined preferences." The organization has difficulty in determining how to measure its progress toward whatever goals it might have. Problematic preference may also be characterized by frequently shifting preferences or goals.
2. Unclear technology -- the members of the organization may not understand the processes by which the organization maintains its existence. Such an ambiguity may be characterized by uncertainty in the relation of the organization to other organizations and generally to its environment, uncertainty in the use of technology of production, and the side effects that technology may have on the organization.
3. Fluid participation -- participation in choices and decision-making varies widely in the time and effort that can be expended by each participant. Involvement in choices varies from time to time, thus clouding organizational boundaries and structure.

In the model, a choice opportunity is seen as a garbage can into which four streams flow. These streams represent choices, problems, solutions, and energy from participants. The material from these four streams is dumped into the garbage can and is constantly mixed. The mixing eventually results in problems, solutions, and choices coming together with sufficient energy to make a decision. The streams into the can are characterized primarily by different rates and patterns of flows. Access structures are used to model the relation between problems and choices. Decision structures are used to model the relation between decision makers and choices.

The Cohen et al paper used a computer simulation to investigate the behavior for several organizational structures and loads. The simulations showed the following major trends:

1. Most decisions are made by either flight or oversight. Oversight is characterized by making quick choices. Choices are made when no problems are attached to them. Flight is characterized by delaying a choice until all problems have migrated to another choice. In both cases, no problems are really solved.
2. An increase in the load on the organization generally increases the problem and decision maker activity, decision difficulty, and the prevalence of flight and oversight.
3. There is a tendency for decision makers and problems to track each other through choices.
4. Important problems are more likely to be solved than unimportant ones.
5. Important choices are less likely to resolve problems than unimportant ones. Important choices are made by flight or oversight. Unimportant choices are made by resolution.
6. Choice failures occur among the most important and least important choices.

Subsequently, Padgett applied the garbage can model to a more realistic, less abstract organization. Padgett developed a simple stochastic model of a traditional hierarchical organization with less fluid participation than the pure Cohen model allowed for, but more than the deterministic models admit actually goes on. Padgett described a simple four tier hierarchy consisting of a President, Vice Presidents, Program Chiefs, and Analysts. In the model, decisions are produced according to standard operating procedures in a straightforward chain of command. The garbage can model is introduced by characterizing management as having problematic preferences and Analysts as having unclear technology. From this model, Padgett is able to give the following advice to the President of the organization:

1. Hire rigid analysts for old, unambiguous (low uncertainty) programs; but hire uncertain and insecure analysts for new high uncertainty programs. (You want someone for high uncertainty projects who will consider all of the options, and for low uncertainty projects someone who will do what he has always done, someone who has a cookbook.)
2. Attempt to organize program assignments to divisions in order to segregate high uncertainty divisions from low uncertainty programs. Never put a high uncertainty project in a low uncertainty division or vice versa.
3. Hire Vice Presidents more liberal than yourself to run high uncertainty divisions, and VPs more conservative than yourself to run low uncertainty divsions.
4. Intervention begets intervention. Intervention in a decision by a higher level during a particular time period increases the probability of intervention during the next time period.
5. Above all, never make any decisions yourself. Any form of tight control over the organization leads to catastrophe.
6. Even though there is a region of the solution space within which the President can make decisions and control can be exercised, the rather steep slope of the curve implies that if one erroneously estimates his position in the safe region, the subsequent loss of control can be quite severe.
7. The President of the organization should expend his energies manipulating VP rules of discretion (centralization policy) to balance off conflicting VPs and Program Chiefs.

In a large organization, these rules can be applied in a relative manner to higher and lower levels of the organization. This leads to a theorem for which 5, 6, and 7 above are corollaries: A manager must restrict his decision making to a level of abstraction commensurate with his position in the organization. In other words, detailed decisions should be made at lower levels with more and more abstract decisions being made at higher and higher levels until at the top, the task is to set the long term environment and policy of the organization. If the President has to make a detailed decision, something at a lower level has failed to work. Decisions and policies set at a higher level must not be too detailed, but must be described in abstract terms consistent with the level in the hierarchy. Detailed plans and decisions on how to accomplish the more abstract decisions must be made in a manner consistent with their environment.

As one moves up the hierarchy, the decisions should become more and more abstract, and the personnel must become more and more adept at effectively manipulating abstract concepts. Basically, Padgett's model characterizes the President's job as setting the environment for the organization, not making detailed decisions. The model advises dealing with uncertainty by pushing decisions down to the level at which the detailed knowledge exists. The degree of detail of the decision must be commensurate with the degree of knowledge at that level of the organization. If not, the decision must be shifted up or down to a level where they are commensurate.

1. Bunker, Stephen and Day, John. "A Garbage Can Model of Amazonian Development Bureaucracy" unpublished paper, 1980.
   
   
2. Cohen, Michael; James March, and Johan Olsen. "A Garbage Can Model of Organizational Choice," Administrative Science Quarterly , 17:1 - 25, 1972.
   
   
3. Padgett, James. "Managing Garbage Can Hierarchies," Administrative Science Quarterly , 1981.
   
   
4. Fama, F. Raymond. "Random Walks in Stock Market Prices," Selected Papers of the Graduate School of Business of the University of Chicago , 1965.
   
   
5. Goodman, George. The Money Game by 'Adam Smith' New York: Random House, 1967.