9—
Conclusion

Those who were best at planning and coordination at MAP were those who felt that complete overall planning and coordination by a central directorate was quite impossible.
—Ely Devons, Planning In Practice

In cases in which coordination can in fact be reasoned (where there are agreed and workable criteria), then it need not after all be central. . . . Thus when the situation is most conducive to successful centrality, centrality is dispensable.
—Charles E. Lindblom, The Intelligence of Democracy

Unnecessary Surgery

Physicians are censured for performing unnecessary surgery. Attorneys are discouraged from frivolous litigation. Should those who pursue unnecessary reorganization be looked upon any differently? Reorganization has no intrinsic value; it is used simply to fix something that is diseased. It follows that before any modification of an organizational system is attempted, the ailment to be relieved should be carefully diagnosed. Then specific cures for specific problems may be proposed, selected, and implemented. Careless tamper-

ing with a healthy organizational system may destroy its vitality and bring into being new and unexpected problems.

Successfully maintaining a loosely coupled organizational system is very different from decentralizing an existing system that has been organized as a formal hierarchy. It is more difficult to flatten a system that has been peaked than to maintain an existing loosely coupled system. This fact alone should make us think very carefully about consolidating loosely coupled systems; should we decide eventually that we have erred, we may not be able to retrace our steps.

Despite surface appearances of fragmentation and ill-coordination of services and resources, given the extent of its interdependence, the Bay Area transit system is highly coordinated and appears to be becoming more so. Integration is made possible by an extensive series of informal relationships between key actors in the system, founded on and buttressed by informal conventions, and supplemented by informal agreements.

These informalities develop because individual actors in the system seek ways to reduce uncertainty resulting from interorganizational interdependence. Uncertainty is produced because organizations, as open systems, depend on each other, but cannot control the behavior of others that affects them. Organizations, as problem-solving entities, operating under norms of rationality, seek to reduce uncertainty. In such circumstances, because absorption or annihilation of other organizations is not an option often available in the public domain, the principal avenue for reducing uncertainty is coordination.

In the absence of formal coordinative arrangements, informalities make coordination possible. Because of these informalities, fragmentation of formal organizational arrangements in Bay Area transit does not present a serious problem for coordination. In fact it is more accurate to refer to Bay Area transit as a loosely coupled organizational system, whose coordinative arrangements match its relatively low level of interdependence. Attempts to integrate formally the currently independent actors would surely disrupt the existing informalities that currently make coordination possible, without any guarantee of improvement.

The findings of this study, although preliminary and suggestive rather than conclusive, have a series of ramifications for organizational systems outside of public transit. Although by organizations I

am referring to administrative organizations in the public domain, with careful attention to the differences between them, these findings should also be applicable to other kinds of organizations. To make a definitive statement regarding the relative effectiveness of one organizational design versus another, some sort of census of the problems that the system faces, and a track record of efforts to deal with those problems, would be useful. This study did not develop such a census, nor did it have a set of experimental conditions in which two (or more) different organizational designs confronted precisely the same set of problems under the same conditions. The occasional rough comparison between the more centralized Washington Metro and the decentralized Bay Area transit system, however, was possible.

Most areas of interdependence that could be identified in the Bay Area transit system required only unilateral or bilateral action; relatively few demanded attention on a multilateral basis. The Bay Area transit system could be decomposed into its several components; intracomponent linkages were indeed much stronger than intercomponent linkages. In fact, the overall level of interdependence was lower than appearances indicated: organizations in the system could behave independently, for the most part, with little effect on each other. Where interdependence in the system was relatively higher, the corresponding density of informal channels was greater than where interdependence was lower, consistent with the hypotheses initially posed in this study.

It would be helpful to develop some way to ascertain more precisely the interdependence between any two (or more) organizations. What percentage of decisions made in any given organization must consider the behavior of other organizations in order to approach some minimal level of rationality? Those percentages could then be distributed into bilateral and multilateral categories. In theory, at least, thresholds at which it would be sensible to consolidate formally could be established, based on analysis of a number of cases with different such percentages and organizational designs.

It also would be useful to develop a distribution of all problems of interdependence requiring coordination solutions by their state of resolution, such as that shown in Figure 9. Which organizational design permits approaching the theoretical limits of coordination most closely? If it was found, for example, that a large proportion of the problems faced by an organizational system fell into the right side of

Figure 9
Distribution of Problems Requiring Coordinated Solutions

the figure, it might be appropriate to revamp the system. But by what criteria would the success of the system be evaluated? And what allowances should be made for variation in importance of the problems? Moreover, if it were possible to devise such a census, then there would already be sufficient knowledge to program decision making for the system, making choice among alternative designs a moot point: a programmed structure would be optimal.

More generally, I suggest that coordinative arrangements be adjusted to the interdependence characterizing any particular system, using the general rule that no more machinery be used than is absolutely necessary to provide a satisfactory level of coordination. On this basis alone, other things being equal, informal mechanisms are to be preferred to formal coordinative arrangements for any given set of interdependence problems. The demands they place on the participants, both cognitively and politically, are less than those imposed by formal systems of coordination. They are also less prone than formal mechanisms to increase artificially the interdependence in the system.

However, informal mechanisms are proposed neither as a panacea for all the ills of formal organizational arrangements nor as appropriate devices for coordination in all cases. Certainly, they are not the pathological developments that they were once thought to be, nor are their beneficial effects merely incidental. Informal channels, conventions, and agreements are capable of performing many of the functions of formal mechanisms, as well as some functions formal de-

vices cannot. They do persist apart from formal structures, and they are far more effective coordinative devices than previously supposed, even where organizations actively compete or where significant resources are involved.

Although informal coordinative systems surely appear "messy," with inconsistency, conflict, competition, and duplication and overlap as major characteristics, coordination occurs where needed. And, as Simon has demonstrated, much problem solving that ultimately leads to superior results is very messy in the process.^[1] Coordination as an end product should not be confused with coordination as a process. There is, indeed, more than one way to skin a cat. Coordination takes place through many processes, some of which are not usually identified with it. The lesson is clear: bearing in mind always their peculiar operating conditions, organizational arrangements should be judged by their results, and not by their apparent rationality . Substance is to be preferred to procedure: in considerations of coordination, concerns about superficial consistency, neatness, and the elimination of redundancy are badly misplaced.

Limits to Informal Coordination

Aside from their considerable instrumental value, informalities do have limits, but ironically those limits have not been as systematically explored as those of formal mechanisms. In this study I explored those limits of informalities that are related to asymmetries of costs and benefits in coordination solutions. Because no organizational design is without costs or tradeoffs, that informal mechanisms have costs attached to them is no more an indictment than it would be of formal mechanisms. The coordinative capacity of informal systems can be enhanced by an organization that plays the role of facilitator in situations where asymmetries of costs and benefits render coordination improbable among formally independent organizations. This organization could also represent systemwide interests. Such a hybrid system compensates for the weaknesses of informalities without exhibiting the pathologies of formal, centralized systems of coordination. Certain conditions that increase both the capability and opportunity of actors to interact informally were identi-

fied and suggestions as to how to create and manipulate those conditions were made.

The central focus was on a multiorganizational system whose component organizations were characterized by similar technologies, common professional backgrounds, and generally similar goals, all of which facilitate the development and maintenance of informal mechanisms. In multiorganizational systems composed of organizations employing different technologies, whose members have divergent backgrounds and whose goals are dissimilar, can a similar quality of informal coordinative arrangements be expected? Can sanitation districts interdependent with highway departments reach informal accommodations as effectively as interdependent transit agencies? Can police departments informally cooperate with recreation agencies as well as BART and Muni work together? Will the same sorts of informal mechanisms develop, and if so, will informal coordination result? I believe they will, wherever there is interorganizational interdependence and some common interest can be established between the organizations involved.

I hypothesized that problems of interdependence involving only two organizations could be more easily resolved than those involving more organizations, irrespective of the type of coordination process. Informal mechanisms, for various reasons, were seen to be particularly well suited to coordinating bilateral interdependence. Their record in areas of multilateral interdependence is less clear, but there appears to be reason for optimism. Even though Bay Area transit is a multiorganizational system, it is characterized primarily by a series of bilateral interdependencies, which existing informalities handle effectively. The system's ability to coordinate the multilateral interdependence that does exist is improved by a voluntary association that easily permits needed action. For multiorganizational systems characterized primarily by multilateral interdependence, informal mechanisms may be less effective. However, Seidman has documented the relative impotence of formal interagency coordinating bodies for handling multilateral problems, and the importance of informal mechanisms for the same.^[2]

As Landau and Stout have pointed out,^[3] many problems endemic to large-scale organizations are mistakenly attributed to faulty or inefficient implementation of "good policy," when they actually result

from levels of value disagreement or lack of knowledge that would be very difficult for any organizational design to contend with. That is, it is the end product of coordination that is difficult to achieve under the particular circumstances and the character of the problem confronted, regardless of what coordination process might be employed. However, instead of taking these difficulties into account, organizations are frequently redesigned to increase "rationality," to the point of overdetermination. Ironically, this lowers capacities to manage and results in less effective coordination. This action rests on two mistaken assumptions: better coordination is always possible, and it is possible only through centralized formal arrangements.

In a similar manner, it is a mistake to assume that problems afflicting individual organizations will be solved once those organizations are merged. In fact, the probability of solution may be diminished because the attention of top managers will be stretched further than before, or because of problems of institutionalization and dominant coalitions. Yet it is precisely this assumption that is currently being made with respect to the woes of individual transit operators in the Bay Area.^[4] Poor decisions may be made in any form of organizational arrangement.

Only some coordination can take place by central directive, or indeed need take place by centrality. In fact, Arrow has argued that (formal) organizations are only a means of "achieving the benefits of collective action in situations where the price system fails."^[5] And even a price system cannot realize the benefits of coordination without "a well-designed system of rights and liabilities"^[6] —for example, informal norms and conventions. The failure Arrow finds absolutely central to understanding organizations is uncertainty about cause-effect relationships. In part, uncertainty derives from the inadequacy of information resulting from the absence of reliable channels of communication.^[7] Informal channels help to redress this problem.

Where centralized coordinative arrangements are in place, their effectiveness often depends on largely informal processes. Lindblom has even suggested that where the conditions essential for successful central coordination exist, no central coordination is necessary.^[8] Where agreed and workable criteria exist, central coordination is at its best. "At the same time, under these circumstances, any competent mind will be brought by application of the criteria to the same decision as would be reached by any other competent mind employ-

ing the same criteria."^[9] In principle, at least, multiple actors (or organizations) should be as satisfactory as one for the problem of coordination.

While the coordinative capacity of systems of mutual adjustment has been made clear by Lindblom and others, for the most part only the processes have been specified, not the mechanisms through which they take place. This study's findings lead me to conclude that informalities provide the foundation that makes possible those processes of mutual adjustment described by Lindblom. Their influence is pervasive in such systems, just as it is in centralized structures. Informal channels facilitate the exchange and acquisition of information essential to coordination through mutual adjustment. Personal trust developed through informal relationships acts as a lubricant for mutual adjustment. Informal conventions preclude some behaviors, establish commonly understood coordination procedures, legitimate the use of sanctions if those boundaries are crossed, and frequently define appropriate sanctions.

It is equally important to note that much, perhaps most, coordination of human behavior occurs through mechanisms other than formal hierarchies. Human behavior regularly is coordinated Without conscious thought, through habit. Conscious or unconscious obedience to rules or laws has a similar effect. As Tocqueville noted more than one hundred fifty years ago:

Nothing is more striking to a European traveler in the United States than the absence of what we call the government, or the administration. Written laws exist in America, and one sees the daily execution of them; but although everything moves regularly, the mover can nowhere be discovered.^[10]

Sometimes general rules and specific understandings can be combined to promote coordination in several ways. They reduce problems of trust, fraud, and extortion. They reduce the costs of selecting from among a set of several preferred outcomes of coordination—in part, through various informal devices that share the normative insistence on certain patterns of behavior. Rules and laws also reduce the costs of communication and identification.^[11] In multiorganizational systems, informal mechanisms, including conformity to group pressures and the norm of reciprocity, contribute significantly to the development of such agreed upon rules of behavior.

In a similar way, the findings of this study complement the work of public choice theorists on problems of coordination. I do not believe the approach taken here competes with that of public choice theory. Complex problems are rarely adequately explained by single-variable theories; coordination is no exception. Where public choice theorists have focused principally on issues of motivation and interest, and alternative coordination solutions that address the differences and commonalities of interest among interdependent actors, this study has centered on the informal processes and mechanisms that make possible those solutions, paying particular attention to the opportunities and abilities to develop informal relationships afforded the actors involved. With Axelrod, I have emphasized that meeting more than once has special effects that tend to make formally independent but functionally interdependent actors more prone to develop stable norms of behavior and to find coordination solutions that satisfy all of the relevant actors.

Ultimately, choosing a centralized, formally integrated organizational system over a "messy," decentralized, informally coordinated multiorganizational system depends less on careful, pragmatic evaluation of relevant conditions and actual results than it does on a world view that emphasizes order, consistency, and apparent certainty and cannot tolerate overt disagreement and ambiguity. Such a world view also apparently values command and coercion over negotiation and voluntary agreement as mechanisms for bringing about coordinated action.

Alternatively, selection of a hierarchical coordinative structure may be a function of a desire to impose one's own values on the rest of an organizational system, a difficult feat in a system composed of multiple, independent organizations. In this case, (proposed) organizational arrangements have less to do with achieving coordination in the face of interdependence than they do with enforcing behavioral conformity to a given set of values or decision premises to be imposed from the top down. Related to concern about imposing values is a devaluation of the importance of individual autonomy and pluralistic pursuit of divergent values.

Another component of this world view is an infallible belief in the human capacity to devise institutions artificially. As Simon describes it: "One kind of optimism, or supposed optimism, argues that if we

think hard enough, are rational enough, we can solve all our problems."^[12] Hayek, among others, also sounds a cautionary note about this capacity:

The fact, however, that in spite of all the advance of our knowledge, the results of our endeavours remain dependent on circumstances about which we know little or nothing, and on ordering forces we cannot control, is precisely what so many people find intolerable.^[13]

Referring to the economic system, he says:

The delusion that advancing theoretical knowledge places us everywhere increasingly in a position to reduce complex interconnections to ascertainable particular facts often leads to new scientific errors. . . . Such errors are largely due to an arrogation of pretended knowledge, which in fact no one possesses and which even the advance of science is not likely to give us.^[14]

To be sure, reliance on informal mechanisms and processes of coordination in lieu of formal mechanisms (particularly hierarchical ones) requires trusting that they will develop in the relevant areas. We know that formal devices can be there when needed; we have constructed them before after all and usually they have some form of legal charter. The issue is whether they will be effective for coordination. Informal mechanisms depend, however, on voluntary actions of many actors, and have no legal recognition, making their ultimate trustworthiness suspect. However, the evidence in this study indicates the regular self-conscious development of stable, enduring, and reliable informal mechanisms for instrumental reasons, particularly for coordination.

Problems, Expectations, and Satisfactory Coordination

Sciences, such as physics, are equipped to understand recurrences of similar problems. Physicists and mathematicians have developed powerful techniques of probability theory and statistics to contend with problems of "disorganized complexity,"^[15] in which

the number of variables is very large, and one in which each of the many variables has a behavior which is individually erratic, or perhaps totally unknown. However, in spite of this helter-skelter, or unknown, behavior of all the individual variables, the system as a whole possesses certain orderly and analyzable average properties.^[16]

However, a whole range of problems does not exhibit these properties,

problems which involve dealing simultaneously with a sizable number of factors which are interrelated into an organic whole. They are all . . . problems of "organized complexity."^[17]

The variables are interrelated in a complicated, but not random, fashion. These problems cannot be handled with the statistical techniques effective in describing average behavior in problems of disorganized complexity.^[18] Most problems confronted by social, economic, and political systems possess characteristics of "organized complexity." There are many factors involved, related in systematic, nonrandom ways, and whose first-order effects on each other are not easily ascertained, let alone their second- and third-order effects. Certainly most multiorganizational systems exhibit the characteristics of organized complexity.

Frequently, problems of organized complexity facing organizational systems are only "ill structured," as opposed to "well structured," where structure is understood to be a function of knowledge and understanding of the patterns of interrelationships among the components. Problems only become well structured when we are able to impose a structure on them. Most problem-solving effort is directed at structuring problems. Only a fraction of it is devoted to solving problems once they are structured.^[19]

Problems begin to acquire structure when they are decomposed into various problems of component design—that is, into a series of smaller problems that can be well structured because they are sufficiently simple to be comprehended. This results in a problem "well structured in the small, but ill structured in the large."^[20] The problem is solved by confronting each component individually, with the result that interrelationships among the various well-structured subproblems may be neglected or underemphasized. "However, there are ways of dividing the whole problem into parts that do less vio-

lence to those interactions than other ways of dividing it."^[21] And the interrelationships among the smaller well-structured problems can then be treated as separate problems.

If organizations are conceptualized as extensions of our problem-solving capacities, then an analogy is drawn between problem-solving method and organizational structure, and problem structure becomes extremely important. Such an analogy suggests that coordinative mechanisms that permit the decomposition of problems into manageable components be employed. Centralized, programmed structures are poorly adapted to contend with ill-structured problems, in part because they are not suited to assimilating new information and integrating it with long-term memory. This approach is consistent with Thompson's and Tuden's (and Lindblom's too, for that matter) contentions that different organizational structures should be linked with different decision strategies. Well-structured problems involve computational decision strategies.^[22] Similarly, centralized, programmed structures rest on assumptions of near-complete knowledge of cause-effect relationships and value agreement.

Conversely, a decentralized, flat organizational structure permits the handling of smaller, well-understood, well-structured problems by programmed structures, and the ill-structured interrelationships among those problems to be dealt with by flexible, adaptive informal mechanisms, which place minimal cognitive demands on the actors involved. Thus, AC and BART can each be operated by formal, centralized arrangements because the problems each faces are well structured, whereas their less well-structured interdependence can be addressed by informal devices. This approach corresponds with my earlier arguments concerning "localizing interdependence." It is also consistent with Lindblom's arguments about simplifying problems by excluding some factors, the difficulties arising from that exclusion to be solved as separate problems.

Little reason exists to employ coordinative arrangements that obstruct the decomposition of problems or that create linkages among problems where none previously existed. Tighter interdependence increases the cognitive difficulty of the problem. It also requires that more actors be involved in the solution to each problem. Involving more actors magnifies the political complexity of the problem, makes coordination more costly, and reduces the probability of an

expeditious solution. Using smaller formal hierarchies to handle those problems sufficiently simple and known to be categorized as well structured, and informal mechanisms to coordinate where interdependence arises among them—ill-structured problems—seems to me an appropriate if radical proposal, given prevailing trends in the public administration. The use of informal mechanisms, constructed by many different actors at different times, follows Dewey's notion of problem solving—that issues be dealt with as they arise in the course of doing one's work, not comprehensively or ahead of time according to some overall master plan.^[23]

Even in the unlikely event that we should possess complete cause-effect knowledge, and face only well-structured problems, we might still choose an organizational design that provides less coordination than a competing alternative. Few organizational systems are characterized by a unitary value or a unitary ordering of values: in theory we can devise a welfare function; in practice, rarely. In such cases, issues of accountability and representation assume vital importance. Although frequently the argument is made that only a unitary centralized organizational structure can provide accountability, I contend quite the opposite: multiorganization is more likely to promote accountability and representation, directly and indirectly.

Profound consequences would follow from the elimination of the fragmentation of authority that is so frequently proposed by reorganizers of the public administration. It would "imply the creation of a sovereign authority as the locus of an ultimate authority in society that can coerce others to obey its decisions but cannot itself be coerced by others."^[24] Limits on authority, as specified in the U.S. Constitution, are only practicable in a system where shared powers permit veto by one organization of another organization's actions.^[25] "The device of popular election may not be sufficient to hold governments accountable,"^[26] and so the different components of government must be set against each other. This was the wisdom of James Madison and the Founding Fathers. They sought to institutionalize—not reduce—conflict among the branches of government, recognizing that it certainly would promote accountability in ways no other institutional arrangements could.

The costs of coordination may be cast in terms of tradeoffs with other important values. If improved coordination can only be achieved

Figure 10
 Costs of Coordination

by eliminating multiorganization, accountability will suffer, along with flexibility and reliability. The Constitution itself, with its separation of powers, "testifies to the willingness of even bold men to sacrifice efficiency in the interests of avoiding risks and to accept instead the price of tedious processes of persuasive coordination."^[27]

And while I have no direct evidence, it seems to me that the cost-coordination relationship is probably not merely linear: it quite possibly takes the form shown in Figure 10. In Section B of the curve, each additional increment of coordination costs substantially more than each additional increment in Section A. Where Line 1 intersects the curve is the theoretically satisfactory level of coordination, defined by an acceptable marginal cost. Line 2's intersection with the curve marks the theoretical limit to coordination, which may also be considered the optimal level of coordination. It simply costs too much to move from Line 1 to Line 2 on the curve: additional increments of coordination may not be desirable in light of the required tradeoffs with other important values, and with the cost of search.

Similar satisfactory levels of coordination can be produced by rather different organizational designs; thus, reaching for optimal levels of coordination may not be particularly desirable: "If the general magnitude of the available search effort is known in advance. then following an optimizing procedure until the cost of additional search exceeds the expected gain is not equivalent to employing the best satisficing procedure."^[28] It may be quite reasonable, therefore, to aim simply for a satisfactory level of coordination. If it is true that "for the satisficer, the unique solution is the exception rather than the rule,"^[29] in a situation of multiple satisfactory solutions, other than by tossing a coin, how can a decision be made among organization designs? Inclusion of other values permits choice among nonunique satisfactory coordination solutions to a problem—the expression of preferences in style. More generally, as relevant conditions vary. it may be appropriate to adjust our expectations of what can be achieved and the sorts of mechanisms that should be used.

Simple faith in our ability to devise formal, artificial structures that will efficiently solve all problems of interdependence simultaneously leads us to expect too much of our public organizations, to reorganize frequently with little real gain but with substantial cost, and to neglect other powerful tools at our disposal—informal mechanisms.