In assessing change many observers are inclined to stress the more dramatic shifts in policy or even the stated goals of policy. They are perhaps especially prone to ignore certain costs—e.g., of impairing incentives, of neglecting uncertainties, of placing heavy demands on delicate communications networks—probably because these costs are so hard to measure. If such costs are neglected, people are in effect insisting that performance be improved or efficiency increased—no matter what the cost! However, in such a view there is an underlying fallacy. Unless we look at full costs, we can easily be penny-wise and pound-foolish.
The issue most critically susceptible to the neglect of full costs is the degree of centralization in decisionmaking. The apparent benefits (in future avoidance of errors experienced in the past) are such that the tendency over time is to acquire additional layers of controls. The benefits are immediately obvious, but the costs—in money, in time and energy, and ultimately in lost options—are far less apparent, especially in the short run. Since this issue of “how much control?” is crucial to the discussion of costs, it is necessary to consider certain aspects of centralization which bear on striking the best balance in distributing responsibility between OSD and the Services…
In any given circumstances, there is no precise balance between centralization and decentralization which is “correct.” In fact, there is an advantage in altering the balance from time to time simply to prevent organizational staleness from setting in. All that one can say is that wherever there is even a moderate risk that overall organizational objectives will be substantially undermined through “suboptimization” on the part of subordinate units, the case for sufficient centralization to reduce the risk of counterproductive clashes to an acceptable level becomes very strong. On the other hand, if there is little risk that a suboptimizing unit will substantially undermine overall organizational objectives, the opportunity for substantial decentralization exists. Action will depend on the effectiveness with which subunits could perform their duties. One should note that the lower the costs of operations relative to the costs of communication or coordination under such circumstances, the more powerful is the case for decentralization. Speaking more concretely, for the expensive and interrelated force-structure decisions, major control at the center appears essential, but for inexpensive and independent R&D decisions which need not damage overall objectives, the case for decentralization is powerful.
That was James R. Schlesinger in Defense Planning and Budgeting: The Issue of Centralized Control. It was written over 1963-1964 with Roland McKean, who coauthored a book with Charles Hitch which launched McNamara’s PPBS reforms of program budgeting and systems analysis. McKean quickly became less optimistic about the PPBS system. Though it wasn’t necessarily that way for McKean, he felt the PPBS could lend itself to centralized control, the suppression of alternatives, the neglect of costs and uncertainties, and so forth.
In 1965, the PPBS was installed across most of the other government departments. Within a couple years, Congress reviewed the PPBS concept in a remarkable series of hearings called the Jackson Committee hearings. By that time, Roland McKean left RAND to become a university professor. Schlesinger took the reins, and his part at the Jackson hearings skyrocketed his career. He became the director of strategic studies at RAND, then assistant director of the Bureau of the Budget in 1969, then onto chair the Atomic Energy Commission and Director of the CIA, all before becoming Secretary of Defense in 1973.
Here’s a good footnote to the first paragraph above:
Several illustrations may be given of how the single-minded emphasis on a particular aspect of performance or cost may result in reduced efficiency or performance overall. In portions of the value engineering effort, to cite one aspect of the cost reduction program, it is almost explicit: we aim to make contractors more efficient (in producing particular components) no matter how much it costs overall. Incentive contracts, while valuable in many applications, can bring heavy costs in others. For example, tying incentive payments to a delivery date results in what the Russians call “storming” —frantic efforts to meet the delivery date at the expense of anything not clearly constrained by the specifications (e.g., reliability, future deliveries, storage costs). Again the existence of bad effects does not tell one what to do—but it does suggest that we must keep struggling to look at the full costs and full effects.
I think we see this latter problem of incentive contracts continuing to occur today.
Leave a Reply