Vertical decentralization is concerned with the delegation of decision-mak- ing power down the chain of authority, from the strategic apex into the middle line. The focus here is on formal power—to make choices and authorize them—as opposed to the informal power that arises from advis- ing and executing. Three design questions arise in vertical decentralization:
- What decision powers should be delegated down the chain of authority?
- How far down the chain should they be delegated?
- How should their use be coordinated (or controlled)?
These three questions turn out to be tightly intertwined. Let us con- sider first some evidence on selective decentralization down the chain of authority. Dale (cited in Pfiffner and Sherwood, 1960:201) and Khandwalla (1973a) found that corporations tend to delegate power for manufacturing and marketing decisions farther down the chain of authority than they do power for finance and legal decisions. Lawrence and Lorsch (1967) found that power for a decision process tends to rest at that level where the necessary information can best be accumulated. For example, in the plas- tics industry, research and development decisions involved very sophisti- cated knowledge that was at the command of the scientist or group leader in the laboratory but was difficult to transfer up the hierarchy. Hence, these decisions tended to be made at relatively low levels in the hierarchy. In contrast, manufacturing decisions tended to be made at higher levels (plant manager), because the appropriate information could easily be accu- mulated there. Marketing decisions fell in between these two.
These findings, in effect, describe the organization as a system of work constellations, our fourth overlay of Chapter 1. Each constellation exists at that level in the hierarchy where the information concerning the decisions of a functional area can be accumulated most effectively. Com- bining these findings in Figure 5-2, we come up with four work constella- tions overlaid on our logo—a finance constellation at the top, a manufac- turing constellation below that, then a marketing constellation, and finally the research and development one. Thus, selective vertical decentraliza- tion is logically associated with work constellations grouped on a func- tional basis. (Note that the decentralization in this case can be horizontal as well as vertical; staff groups at different hierarchical levels are shown involved in the top three constellations, and the fourth is exclusively staff.)
But such selective decentralization leaves important interdependen- cies to be reconciled, which raises the question of coordination and control. Direct supervision may be used to some extent, specifically by having the decisions of each work constellation authorized, and therefore coordi- nated, by the managers at the strategic apex. But too great a reliance on this form of coordination would be tantamount to recentralizing the decision processes and thereby canceling the advantages of selective decentraliza- tion. The same is true for the standardization of work processes or outputs, since that transfers power over the decision processes from all the con- stellations to the technostructure, which amounts to horizontal centraliza- tion instead of vertical decentralization. So although it may make some use of activity planning, in the final analysis, the organization that is selec- tively decentralized in the vertical dimension will coordinate its decision making largely by mutual adjustment. Specifically, it will place heavy emphasis on the use of the liaison devices.
The situation is quite different for parallel decentralization in the vertical dimension. This kind of decentralization does away with decision interdependencies: power for the different functional decisions is focused at a single level in the hierarchy, specifically within units grouped on the basis of market. This is the structure known as “divisionalized” in the corporate sector. Each unit or division is decoupled from the others and given the power necessary to make all those decisions that affect its own products, services, or geographical areas. In other words, parallel vertical decentralization is the only way to grant market-based units the power they need to function in a quasi-autonomous manner. (Of course, such vertical decentralization must always be somewhat selective. That is, some decision-making power is always retained at the strategic apex. The divi- sionalized corporation typically delegates marketing and manufacturing decisions to the divisions but keeps finance and acquisition decisions at the strategic apex.)
Figure 5-2. Selective decentralization to functional work constellations
With the extensive autonomy of each market-based unit, there is no need to encourage mutual adjustment or action planning to coordinate work across them. What is important is to ensure that the autonomy is well used, that each market unit contributes to the goals considered important by the strategic apex. So the strategic apex faces the delicate task of control- ling the behavior of its market units without restricting their autonomy unduly. Three coordinating mechanisms present themselves for such con- trol—direct supervision and the standardization of skills and of outputs. (The standardization of work processes would obviously be too restrictive.)
There is some room for direct supervision, notably to authorize the major expenditures of the units and to intervene when their behavior moves way out of line. But too much direct supervision defeats the pur- pose of the decentralization: the strategic apex comes to manage the unit instead of its own manager. The standardization of skills, through training and indoctrination, can also be used to control the behavior of the manager of the market unit. He may, for example, be carefully indoctrinated and then sent out to run it with considerable autonomy. But there typically remains the need to monitor behavior—to find out when it is out of line. And that is typically left to the performance control system. Parallel de- centralization in the vertical dimension (to market-based units) is regu- lated primarily by performance control systems. The units are given per- formance standards, and as long as they meet them, they preserve their autonomy.
But does parallel vertical decentralization to market-based units con- stitute “decentralization”? In the corporate world, the terms “divisional- ization” and “decentralization” have been used synonomously ever since Alfred P. Sloan reorganized General Motors in the 1920s under the maxim “decentralized operations and responsibilities with coordinated control” (Chandler, 1962:160; see also Sloan, 1963). Faced with a structural mess left by William C. Durant, who had put the legal entity together through a series of acquisitions but had never consolidated it into a single organiza- tion, Sloan established product divisions with some operating autonomy but maintained tight financial controls at headquarters. A number of large corporations followed suit, and today the divisionalized structure is the most popular one among the largest American corporations. But does divi- sionalization constitute decentralization? Not at all; it constitutes the vest- ing of considerable decision-making power in the hands of a few people— the market unit managers in the middle line, usually near the top of it— nothing more. That is, divisionalization constitutes a rather limited form of vertical decentralization. These managers can, of course, delegate their power farther down the chain of authority, or out to staff specialists. But nothing requires them to do so. To paraphrase Mason Haire (1964:226), “decentralization” can give a manager the autonomy to run a “cen- tralized” show!3 Thus, we should not be surprised when the same struc- ture in a different context—the communist economy—is called centralized. A structure—capitalist or communist—in which a few division managers can control decisions that affect thousands or even millions of people can hardly be called decentralized, although it is certainly more so than one in which these decisions are made by even fewer managers at the strategic apex.
Source: Mintzberg Henry (1992), Structure in Fives: Designing Effective Organizations, Pearson; 1st edition.