Not so long ago, the late Frederick C. Mosher (1980, pp. 545-47) observed that government has experienced a sea change in its responsibilities and its tactics. He concluded that these massive changes have rendered obsolete the rule-bound governance mechanisms that we inherited from our forebears. Nevertheless, according to budget expert Allen Schick (1978, p. 518) these changes in government responsibilities and tactics have been accompanied by massive growth in the scope and content of traditonal administrative controls. Schick warns us that we cannot afford to go on imposing direct controls over an ever-widening sphere of activities -- that new solutions to the problem of administrative governance must be sought. He also reminds us that in many cases individuals can be more effectively influenced to serve the public interest "by inducements which allow them to pursue their own interests than by constraints which try to bar them from behaving as they want."

Remarkably, many of the participants in contemporary debates over government management and operations are unfamiliar with alternatives to rule-bound governance mechanisms. In this article I describe the four basic management control systems designs that are available for influencing people to advance the policies and purposes of the institutions they serve: (1) outlay budgets, (2) responsibility budgets, (3) fixed-price contracts, and (4) flexible-price contracts. I then show how each of these mechanisms can be executed to enforce efficiency in the delivery of services and outline the circumstances under which each has a comparative advantage over the others. I also show what happens when these designs are misused -- and overused.

The design and implementation of control systems is a ubiquitous problem. Engineers, planners, and regulators as well as management controllers encounter it. The purposes of various kinds of control systems differ, as do the details of their execution, but all control system designers face the same key choices: what, where, when, and, in the case of human systems, whom to control. The choice of what and where to control is reasonably self-evident. Management control should be primarily addressed to the behavior of service suppliers (departments and agencies, other levels of government, and contractors), the efficiency with which they produce goods and services, and ultimately the efficiency with which they use the assets at their disposal.

The choice of whom to subject to controls and when to execute those controls is far less self-evident. In the abstract, a control system designer has four sets of options, comprised of two choices of subject and two of timing.

(1) The subject may be either an organization or an individual,

(2) Controls may be executed either before or after the subject acts.

Before-the-fact controls are intended to prevent subjects from doing undesirable things or to compel them to do desirable things and necessarily take the form of authoritative mandates, rules, or regulations that specify what the subject must do, may do, or must not do. The subjects of before-the-fact controls are held responsible for complying with these commands and the controller attempts to monitor and enforce compliance with them.

After-the-fact controls are executed after the subject, either an organization or an individual, decides on and carries out a course of action and, therefore, after some of the consequences of the subject's decisions are known. Because bad decisions cannot be undone after they are carried out, after-the-fact controls are intended to motivate subjects to make good decisions. Hence subjects are made responsible for the consequences of their decisions, and the controller attempts to monitor those consequences and to see that subjects are rewarded or sanctioned accordingly.

Combining the choice of subject with that of timing, we find that the control system designer must choose among four distinct institutional alternatives: individual responsibility, before-the-fact or after-the-fact, and organizational responsibility, before-the-fact or after-the-fact. In this article I will try to explain the significance of this choice, its relevance to management control, and the economic logic that should guide it.

The significance of these alternative institutional arrangements is partially reflected in the current debate over the merits of privatizing the delivery of public services. Proponents of privatization imply that we have a choice between rule-governed, often over-regulated, monopolistic public bureaucracies and freely competing private firms. They conclude that the latter will usually be more efficient than the former. If that is our choice, it is difficult to see how privatization could be wrong, since it resolves to a simple question of monopoly or competition. Clearly provision by competing private firms will almost always be more efficient than provision by a public monopoly -- except possibly where production of the good or service in question is characterized by increasing returns to scale, a high degree of lumpiness in production or consumption, asset specificities, or the absence of close substitutes.

However, the distinction drawn by proponents of privatization between provision by a public agency and provision by a private entity is inordinately simplistic. It fails to capture the full range of choices available to the management controller. It also fails to reflect all of the factors that are relevant to his or her choice.

First, although it is true that organizations produce most of the goods purchased with public money, effective control ultimately presumes individual accountability. The distinction drawn by the proponents of privatization between public and private provision ignores a controller's capacity to hold managers under his jurisdiction responsible for their behavior and to influence directly the rewards and sanctions that accrue to those individuals.

Controllers cannot possibly hold managers personally responsible where their relationship to the supplying organization is at arm's length and the structure of individual responsibility is veiled by the organizational form. The only way an organization can be rewarded (or punished) is by increasing (or reducing) its revenues. An organization’s revenues can affect an individual manager's welfare, but only indirectly.

The difference between holding individuals and organizations accountable or between direct, personal influence and indirect influence is quite straightforward. Take the following example: if the quality of services supplied by a public agency is grossly unsatisfactory, as controller, you can recommend the dismissal of the agency manager. Where government has an arm's length relationship with a service supplier and the relationship is unsatisfactory, all that you controller can do is recommend termination of the relationship. You can punish the supplying organization, but cannot punish the manager responsible for the failure -- although his or her actions might very well lead the organization's board of directors to so! Unfortunately, punishing your only supplier is like cutting off your nose to spite your face; rewarding one is like eating an eclair to celebrate staying on a diet. Consequently, where your supplyer is a monopoly, the capacity to influence managers directly will have considerable utility, particularly where you can stimulate and exploit competition between alternative management teams.

This claim can be verified by reference to the private sector. In the private sector, most real natural monopolies make intermediate products, i.e., goods that are used to produce consumer goods or services. Natural monopoly (decreasing costs as output increases) can usually be attributed to spreading large, lumpy investments in specialized resources -- technological know-how, product specific research and development, equipment. -- over additional output. Investment in specialized resources often inspires a process called vertical integration ("backward" if initiated by the consumer goods producer, "forward" if initiated by the intermediate goods producer). The new economics of organization tells us that vertical integration occurs because it permits transaction costs to be minimized, in part through the substitution of direct supervision for indirect influence (see Williamson, 1985).

In the jargon of transaction-cost economics, investment in specialized resources is called asset specificity. An asset is said to be specific if it makes a necessary contribution to the provision of a good or service and has a much lower value in alternative uses. The corollary of asset specificity is bilateral monopoly, a circumstance that provides an ideal environment for opportunistic behavior on the part of both the intermediate product supplier and his customer.

For example, once an intermediate product producer has acquired a specialized asset, his customer may be able to threaten to switch suppliers to extract discounts from him. In that case, the supplier may find it necessary to write off a large part of his specialized investment. Or, if demand for the final good increases greatly, the intermediate product supplier may be able to use his monopoly power to extort exorbitant prices from his customer. Hence, where the relationship between the intermediate product supplier and his customer is at arm’s length, the threat of opportunistic behavior may be sufficient to eliminate the incentive to make what would otherwise be cost-effective investments. Vertical integration can eliminate this threat. Indeed, where the intermediate product producer provides homogeneous goods or services (i.e., outputs that are easily monitored), total production volume is specified, and technologies are mature, vertical integration permits a bilateral monopoly to be governed satisfactorily by unbalanced or two-part transfer prices.

Moreover, the proponents of privatization err in their implicit claim that responsibility can be vested in organizations if and only if the organization is private, and in individuals if and only if the organization is part of the public sector. The absurdity of this claim becomes clear as soon as it is explicitly stated; it is consistent with neither theory nor practice. For example, many state legislatures base their relationships with public entities such as universities or hospitals on arm’s-length relationships that are guaranteed by self-denying ordinances, which exempt the managers of these public entities from detailed oversight and direct control (e.g., see Blumenstyk, 1991). Similarly, the recurring procurement fraud cases show that the managers of private entities that supply services to the government can be held directly responsible when their behavior violates federal law.

Finally, most of the proponents of privatization implicitly presume that the services provided to or for government are homogeneous or fungible, which implies that the problem of identifying the most efficient supplying organization or management team resolves to a simple question of price search, an elementary control mechanism that reveals information about the "customer's" demand for the service. In fact, many of the organizations supplying goods or services to or for government supply bundles of more or less heterogeneous products&emdash;many of these products are hard to measure and costly to evaluate, some prohibitively so.

The proponents of privatization do, however, make one significant, unexceptionable claim: that the choice of institutional arrangements should depend on the cost and production behavior of the good or service in question. However, they frequently fail to carry this claim to its logical conclusion. At least two factors are relevant to this choice: the ease with which the consequences of operating decisions can be monitored and the desirability of interorganizational competition.

Most management control theorists believe that where consequences (that is, an organization or responsibility center's outputs) are easily monitored, control should focus on the consequences of the subject's decisions; where they are not, control should focus on their content (inputs). Because consequences are easily monitored where entities produce homogeneous outputs or where a responsibility center within an entity performs fungible activities, it follows that controllers should rely on after-the-fact controls where homogeneous outputs are supplied. In contrast, it follows that they should rely on before-the-fact controls where each item supplied is, from the "customer's" perspective, intrinsically unique. Furthermore, this view has been reinforced by recent findings in transactions costs economics and agency theory.

At the same time economic theory tells us that interorganizational competition is desirable only where costs are constant or increasing as quantity of output (rate or volume) increases. Where costs decrease as output is increased, monopoly supply is appropriate. Because responsibility can be effectively vested in organizations only where customers or their agents are ultimately indifferent to the survival of one or more of the supplying organizations, the implication of this line of reasoning is that controllers should vest responsibility in organizations only where inter-organizational rivalry is practical and likely to be effective; and in individuals where it is not.

Government uses all four kinds of controls. But is each appropriately employed? Before I can answer this question, I must first show how these designs are used and explain the practical logic of their implementation. My discussion will concentrate on the use of before-the-fact controls. This does not mean that I particularly like them. On the contrary, I believe that controllers should resort to before-the-fact control designs only where the cost and production behavior of the good or service in question makes their use the least objectionable alternative available.

I concentrate on the use of before-the-fact controls because it seems to me that their implementation is not well understood, especially by those who most rely on them. Many participants in the policy process believe that before-the-fact controls not only safeguard against abuse, but also, by reducing costs, improve mission performance. If failure occurs nevertheless, they tend to believe the solution lies in still more or better rules. One possible explanation for the persistent faith in the efficacy of before-the-fact controls is that its devotees just don’t understand how hard it is to execute them efficiently. For example, they appear to believe that the subjects of before-the-fact controls will comply with them simply because they are morally obligated to do so. Obviously, however, not everyone is inclined to respect moral authority, to respect the law, or to obey rules. It is necessary to monitor and enforce compliance with rules and to ferret out and punish noncompliance. It is also necessary to specify the content of before-the-fact controls -- to tell subjects what to do and what not to do in such a way as to find and enforce efficiency, which is no easy matter.

Before-the-fact controls are like after-the-fact controls in that they ultimately rely on incentives and sanctions for their effectiveness. The difference is that after-the-fact management controls are incentive or demand-revealing mechanisms, while before-the-fact management controls are incentive or demand-concealing mechanisms. As I will explain, this means that opacity is an essential characteristic of before-the-fact controls. The incentive aspects of before-the-fact controls are thus less clear than are the incentive aspects of after-the-fact controls. This means that their effectiveness is hostage to the skill with which they are executed. It also means that the incentive aspects of before-the-fact controls are easily overlooked, which might help explain why they are not better understood. In order to show how demand-concealing mechanisms work, I will first try to show how demand-revealing mechanisms work or, at least, what they are.

By demand-revealing mechanisms, I mean those in which customers (or their agents) declare their willingness to pay for various quantities of goods, services, or activities. Customers transparently reveal a demand schedule that fully expresses their wants and preferences to their suppliers. Then they let suppliers figure out how best to satisfy those wants and preferences. The classic demand-revealing mechanism is the competitive spot market, where customers buy from any number of anonymous firms. When many suppliers are disposed to satisfy customer wants, the customer simply chooses the best price and quantity combination offered -- the one that moves him farthest down his demand schedule. In so doing the customer rewards the organization that is willing to do the most to satisfy his preferences and implicitly punishes the rest. For example, he might order wheat from a broker, at the market price payable on delivery. In that case, there would be no formal contract. The customer would put no restrictions on the producer. In fact, the customer will probably not even know who grew the wheat. But the wheat farmer is nevertheless rewarded for his contribution. Government relies on spot markets when, for example, it purchases electrical components off the shelf.

After-the-Fact Controls Transparently Reward Measured Performance

The spot market is by no means the only demand-revealing mechanism that is used to govern relationships between buyers and sellers. There are many variations on the basic theme of reliance on transparent rewards. But all of these variations have one common attribute: rewards are provided after operating decisions have been made by the producer, after his or her asset acquisition and use decisions have been carried out and outputs have been monitored. Because they are executed after asset acquisition and use decisions have been carried out, I refer to them as after-the-fact controls.

A close analog to a spot market is seen where government uses prospective price mechanisms to reimburse free-standing service providers. The system used by the Health Care Financing Administration to pay hospitals for treating patients is an example. The enrollment-driven funding formulas used by some states to compensate post-secondary institutions for teaching students is another (Jones, et al., 1986). In both of these instances the subject is a freestanding organization, and the structure of authority and responsibility within the supplying organization is assumed to be a purely internal matter. The government or its agent, say a controller, announces a price schedule and specifies minimum service quality standards (or a process whereby these standards are to be determined) and the time period in which the price schedule will be in effect.

Under prospective pricing, all qualified organizations will be paid a stipulated per-unit price each time they perform a specified service, such as enrolling a full-time equivalent student or treating a heart attack. This means, among other things, that the government's financial liability is somewhat open ended. It depends on the quantity of service actually provided, although not directly on the costs incurred by the organizations supplying the service.

Another close relative of the spot market is the fixed-price contract. Again the government buys from numerous suppliers held at arm’s length. Frequent bidding contests are held and orders are shifted among suppliers chosen simply on price. Under a fixed-price contract, government may grant a selected organization a franchise to provide a specified service, perhaps at a specified location, for a fixed period of time -- garbage collection at a military base, for example. But when the contract is completed, the government again puts the franchise out to bid to all comers.

Under all of these demand-revealing mechanisms, the government relies upon interorganizational competition, combined in most instances with the profit motive, to motivate service suppliers to produce efficiently and therefore to make wise asset acquisition and use decisions. If interorganizational competition is effective, organizations that don't make wise asset acquisition and use decisions will fall by the wayside.

Demand-Revealing Mechanisms in Vertically-Integrated Organizations

In some cases, even where the cost behavior of the service in question renders vertical integration and therefore monopoly supply appropriate, demand-revealing mechanisms or after-the-fact controls can still be effectively employed. Where the supplier is part of the same organization as the customer, the organization rewards managers who do the best jobs of satisfying their customer’s preferences and explicitly punishes the rest. This is done in businesses and businesslike public sector organizations by holding a manager responsible for optimizing a single criterion value, subject to a set of specified constraints. This control mechanism is known as responsibility budgeting (Anthony and Young, 1988, pp. 365-386; Thompson, 1991). For example, under responsibility budgeting the manager of a cost center is given the authority to make spending decisions -- to acquire and use assets, subject to exogenously determined constraints on the quality and quantity of output -- and is held responsible for minimizing costs. Note that, in contrast to other demand-revealing mechanisms, under responsibility budgeting an organization’s financial liability will depend upon the costs actually incurred providing the service to the customer and not merely on its quantity or quality.

Under this control system design, the structure of authority and responsibility within the organization is of crucial interest to the management controller. The effectiveness of responsibility budgets depends on the elaboration of well-defined objectives, accurate and timely reporting of performance in terms of those objectives, and careful matching of spending authority and responsibility. Their effectiveness also depends on the clarity and transparency with which individual reward schedules are communicated to responsibility-center managers and the degree of competition between alternative management teams.

Before-the-fact management controls are demand-concealing mechanisms. Their distinguishing attribute is that they are executed before public money is spent. That is, they govern a service supplier's acquisition and use of both short-term and long-term assets, which means that the controller retains the authority to preview these decisions. Examples of before-the-fact management control include object-of-expenditure appropriations -- these govern the kind of assets that can be acquired by governmental departments and agencies; apportionments, position controls, and the fund and account controls that regulate the rate, timing, and purpose of public spending (Pitsvada, 1983; Schick, 1964 and 1978), and the similar rules and regulations that govern the behavior of private contractors (Goldberg, 1976; Kovacic, 1990).

Readers will recognize the combination of before-the-fact controls and individual responsibility in traditional governmental budgets. Most will also recognize the combination of before-the-fact controls and organizational responsibility in the so-called cost-plus contract -- the most notorious member of the administered or flexible-price contract family.

Traditional governmental budgets are basically spending plans. To distinguish them from responsibility budgets, I will use the term "outlay budgets." Under outlay budgets, supplying organizations are guaranteed an allotment of funds in return for providing a service for a stipulated period. They usually receive the allotment regardless of the actual quantity or quality of services provided.

Flexible-price contracts are basically production plans. They fully specify product or service characteristics and a usually a delivery schedule. Under flexible-price contracts, supplying organizations are guaranteed reimbursement (complete or partial) for any legitimate expense incurred providing the service. Hence, the prices they are paid for providing services are determined retrospectively according to settled cost-accounting standards and the specifics of their contracts.

To say that controllers primarily focus their attention on a supplier’s asset-acquisition decisions does not mean, however, that they ignore performance in executing outlay budgets or price in executing flexible-contracts. Controllers usually take account of information about the future consequences of a supplier’s decisions as well as information on its current and past behavior. Their attention to performance may be tacit, as it is in the execution of traditional line-item budgets, rather than express, as in the execution of performance, program, or zero-base budgets. But the consequences of asset acquisition decisions usually matter a great deal to controllers. What is crucial is that, under these control systems designs, attention to the performance consequences of spending decisions is necessarily prospective in nature. Controllers will not reveal a demand schedule that fully expresses customer wants and preferences to suppliers or leave it to suppliers to figure out how best to satisfy those wants and preferences.

Even under these control systems designs, the service provider, whether a department or an outside contractor, must assume some responsibility for managing output levels and delivery schedules, service quality, or price. Nevertheless, as I shall demonstrate, the logic of demand-concealing oversight requires supplier discretion to be carefully restricted. This means that suppliers must be subjected to fairly extensive, fairly detailed before-the-fact controls. A bureau’s outlay budget, for example, should clearly identify all the asset acquisitions that it is to execute during the fiscal year, specify their magnitudes, and make it clear who is responsible for implementing each acquisition.

Of course, constraining managerial discretion is not the only function that before-the-fact controls perform. If it were, it would be hard to claim that they ever represented a least-objectionable alternative, let alone explain their widespread use. Rather, as I will explain, constraining managerial discretion is chiefly a means to an end, not an end in itself. To show how subjects can be told what to do in such a way as to enforce efficiency, I will outline in detail the logic of employing the two basic before-the-fact control system designs: flexible-price contracts and outlay budgets.

There is a difference in the role that competition plays under fixed and flexible-price contracts. The difference is not that it takes place before the production of the service in question. (Economists refer to such a competitive regime as competition for the market, to distinguish it from competition in the market.) The recipients of fixed-price contracts often receive exclusive franchises prior to the delivery of services.

The difference between the role played by competition under fixed and flexible-price contracts is that, under flexible-price contracts, competition cannot be relied upon to keep prices low, let alone to enforce efficiency. Once a flexible-price contract has been signed, the supplier is free to dip his hand into the customer’s pocket. Because the supplier is spending somebody else’s money, the normal incentives to cost effectiveness largely disappear. Decisions that affect cost, service quality, or price (i.e., asset acquisition and use decisions) must be made during performance of the contract, but once the contract is signed, the supplier can no longer be fully trusted to make them. This conclusion holds especially where the customer ignores information regarding the performance of incumbent suppliers on earlier contracts or cannot (will not) promise to award future contracts based on good performance. Even where fixed-price contracts are concerned, the refusal to take past performance into account discourages supplier loyalty and eliminates any incentive to improve the quality of the product delivered (see Kelman, 1990).

Why, then, would a customer ever sign a flexible-price contract? Why not simply write fixed price contracts? The answer is that a fixed-price contract is the mechanism of choice where controllers know precisely what their principals want and there are several potential service suppliers who know how to meet those preferences. Under those circumstances, service quality attributes offered, promised delivery schedules, and bid price allow us to evaluate proposals satisfactorily. Regrettably, these conditions are likely to obtain only where the service supplied is fairly simple and relatively standard -- garbage collection, for example.

Technological Uncertainty and Financial Risk

In other cases, neither the controller nor the service supplier will have enough knowledge of the value of product attributes or production processes prior to performance of the contract to employ a fixed-price contract. It is a simple fact of life that considerable experience is usually required to manage to a narrow range of outcomes; where specialized or unique services are involved, no organization is likely to have the required experience. Consequently, any organization that agreed to produce a unique service, according to a specified schedule, at a fixed price would incur a large financial risk. This risk can be shifted, but it cannot be eliminated.

Government can often bear financial risks better than supplying organizations. This is the usual case where the federal government is concerned, because of the size of the assets it commands and its ability to pool risk. Consequently, the cost to government will often be lower if it assumes a portion of the risk associated with acquisition of the service. Flexible or retrospective pricing is one way for government to assume this risk. Moreover, the preferences of the government may change during performance of a contract. Under a fixed-price contract, it might not be possible to secure desired changes in service attributes if they involve increased costs for the vendor.

My point is that customers should prefer flexible-price contracts to fixed-price contracts where it is cheaper for the customer to deal with uncertainty than it is for the contractor to do so or where the customer is more concerned with the ability of the contractor to provide a product that works than with price. The question is: can before-the-fact controls be used to insure that the seller retains an interest in cost effectiveness?

Using Before-the-Fact Controls to Enforce Efficiency under Flexible-Price Contracts

Execution of a flexible-price contract begins with a fully specified project spending plan detailing work to be performed, personnel, material, and equipment to be used, input quality standards, and scheduled milestones. This plan provides a basis for the enforcement of efficiency through bargaining and negotiations carried on during the performance of the contract.

This process can be compared to a repeated prisoner's dilemma game, in which both parties have a common interest in reaching agreement but also have antagonistic interests with respect to the content of agreements. In this game, the customer tries to get as much of what he wants as he can at a given price, and the supplier tries to get the highest possible price for providing the service (Hofstede, 1967). Bargaining power in a prisoner's dilemma game depends on the information available to each party. In particular, the customer's power is greatest where the customer (or his agent) knows the supplier's true cost schedule, but can withhold full information as to his preference or demand schedule (Morgan, 1949). In a repeated game the information available to the customer (or his agent) will depend upon his ability to control the sequence of moves and counter moves that comprise the game. Public choice theorists refer to this condition as agenda control (Hammond, 1986).

Given comprehensive before-the-fact controls, under which changes can be made only with the prior approval of the other party or his agent, the party suggesting or initiating a change must necessarily reveal valuable information to the other. This can work to the advantage of the customer or the supplier, or both. For example, consider the following situation:

. . . contracts and specifications are drawn for . . . a ship and agreed to . . . . The contractor discovers he can do the welding of some plates less expensively by another means. About that time the client decides that some room on the ship should be larger.... The contractor can plead that he cannot easily change the room size: however, if the client will permit the altered welding maybe a deal can be struck (Stark and Varley, 1983, p. 132).

But when flexible-price contracts are appropriately employed, there is every reason to believe that the service supplier will initiate most change proposals. Competition for the market provides an incentive to potential service suppliers to promise more than they can deliver, since contracts are usually awarded to the service suppliers who promise the most. Consequently, very few contract winners can make good on all their promises, especially where their managerial discretion is severely restricted by a full set of before-the-fact controls. This fact will usually become evident to the service supplier during performance of the contract. The service supplier will also learn of the tradeoffs between cost, service quality, and delivery schedule available to it and will eventually want to (or in some cases have to) change its promises or its plans.

Under a full set of before-the-fact controls, such changes are contingent upon prior approval. To secure that approval the service supplier must reveal information about its capabilities and trade-off possibilities. As a result, power to enforce the preferences of the government will over time passed to the purchasing officer -- but only if she knows what she is doing and how to make it happen.

A similar logic (Wildavsky and Hammond, 1965) applies where outlay budgets have a comparative advantage -- under decreasing costs to scale over an array of specialized or unique services. Outlay budgets can help to keep prices low and to encourage efficiency where large, lumpy investments in specialized resources are needed in order to provide services, where each problem, client, or task performed is in some sense unique, and where the most serious problems are supposed to be dealt with first. Many organizational units in government have these attributes. They supply outputs that are heterogeneous, hard to define, and nearly impossible to measure. As a consequence,

[s]uch bureaus seem always to be near the beginning or end of a comprehensive dismantling and restructuring since there is usually a sense that performance is not all that it might be. The performance of such bureaus can only be improved by budget augmentation. And, of course there are no guarantees in budget augmentation alone (Thompson and Zumeta, 1981, p. 43).

Under outlay budgets, the purchasing officer retains the authority to review all significant asset acquisition and use decisions. Presumably, therefore, he would like to know as much as possible about alternative choices and their consequences before the manager of an administrative unit decides or acts. That is, the controller would like the service supplier to reveal a comprehensive menu of all possible actions and a price list identifying the minimum cost of performing each action under every possible contingency. But wishes are not horses. There is no way to compel the manager of an administrative unit within an organization to reveal her unit's true production function -- even if she knows what it is (and in most cases, she won't).

Consequently, the controller must usually settle for a practical approximation of this ideal. Here too, the controller's authority provides a basis for the enforcement of efficiency through bargaining carried on during the execution of the budget. If the controller is skillful, if he plays his cards right, his principals' preferences may be approximated, if not fully satisfied. That is, over time, he may be able to compel the supplying organization to address the "most important" problems and to address these problems at a reasonable cost.

The more pressured the unit, the faster its movement. But, here too as with flexible-contracts, the impetus for change must come from the operating manager. That is, the responsibility center manager must have an interest in increasing his budget. Otherwise he will be indifferent to circumstances in which low priority problems drain resources from problems that are of greater importance to his superiors or legislative sponsors. And here too, a full set of before-the-fact controls must be in place. At a minimum this means that controllers must specify when, how, and where assets are to be employed and how much the subordinate can pay for them. In addition, money saved during the budget period from substituting less costly or more productive assets for more costly or less productive assets must revert to the treasury. Money lost in failed attempts to improve operations must be found elsewhere and new initiatives requiring the acquisition of additional assets or reallocation of existing assets must be justified accordingly.

These constraints are necessary because they prevent the operating manager from overstating asset requirements in high priority areas to get resources for use elsewhere, thereby creating a precedent for higher levels of support in the lower priority area. They are also necessary to force the operating manager to seek authorization to make changes in spending plans and, therefore, to reveal hidden preferences, capabilities, and trade-off possibilities.

Where these conditions obtain, where a budget maximizer is subject to tight before-the-fact controls, the controller can enforce efficiency during the budget period by requiring affirmative answers to the following questions: Will a proposed change permit the same activity to be carried out at lower cost? Will higher priority activities be carried out at the same cost? Will the proposed asset acquisitions or reallocations of savings support activities that have lower priority than those presently carried out? When operating managers know and understand these criteria, the controllers will approve most changes in spending plans that the managers propose -- because managers will propose only mutually advantageous changes.

Paradoxically, to say that before-the-fact controls are needed to reinforce the controller’s bargaining power where outlay budgets are called for does not mean that the controller must administer before-the-fact controls directly. Under certain necessary and sufficient conditions, authority to spend money, transfer funds, and fill positions can be delegated to operating managers. The threat that direct controls might be reimposed can be sufficient to insure that the operating manager asks the right questions of himself and gets the right answers to those questions before he takes action &emdash; which should go a long way toward insuring that the manager’s behavior corresponds to the customer's preferences.

The necessary conditions are: reimposition of controls must be a credible threat; the gain to the operating manager from delegation must more than offset the associated sacrifice in bargaining power (the manager of an aging agency in the stable backwaters of public policy, for example, may have nothing to gain from relief from before-the-fact controls, if the price of such relief is a change in business as usual); and the controllers must be confident that their monitoring procedures, including post-audit, will identify violations of "trust."

The sufficient condition is that the controller and the operating manager trust each other.14 Trust requires mutual respect and understanding and a common sense of commitment to a joint enterprise. In this context, its corollary is a willingness on the part of both the controller and the operating manager to eschew opportunistic behavior that would be costly to the long-term well-being of either the operating unit or the organization as a whole, including a willingness to forego opportunities to exploit events for personal advantage. Trust in a bargaining relationship can be poisoned by a single lapse of honesty or fair dealing, by contempt on the part of one of the parties for the abilities, judgment, or ethical standards of the other, by an excess of zeal or an overtly adversarial or confrontational approach, or by a simple lack of communication. In other words, the kind of trust that is needed to realize the best possible outcomes under a spending budget, or under a flexible-price contract for that matter, can be threatened by the very same conditions that threaten a business partnership -- or, more familiarly, a marriage.

All long-term buyer-seller relationships must ultimately rely on incentives, even those governed by outlay budgets and flexible-price contracts. As we have seen, the difference is that when these control system designs are employed, the incentives are deeply embedded in the process of budget or contract execution. Consequently, they are often overlooked. External observers fail to understand how they work; they also fail to understand how hard it is to make them work well. Effective execution of demand-concealing control system designs -- flexible-price contracts as well as outlay budgets -- requires a great deal of skill and savvy on the part of the controller. The skills required to execute demand-concealing control system designs properly are certainly far rarer and more remarkable than are those needed to design and execute after-the-fact controls, for which a modicum of technical expertise will suffice. It usually takes years of training and practical experience, combined with a lot of horse sense, to manage the complexities of bargaining in this context.

All long-term buyer-seller relationships rely to a degree on standards and rules. Even, where government uses prospective price mechanisms to reimburse free-standing service providers, quality standards must often be specified and enforced. But demand-concealing control system designs require considerably higher levels of reliance on before-the-fact controls and also on monitoring and enforcing compliance with them than do demand-revealing designs. At the very least, adoption of one of these control system designs means that controllers must take steps to ensure that suppliers fairly and accurately recognize, record, and report their expenses. This, in turn, requires careful definition of costs and specification of appropriate account structures, bookkeeping practices and internal controls, direct costing procedures, and the criteria to be used in allocating overheads.

But accurate accounts will not guarantee efficiency. Even if -- as is unlikely to be the case -- the service supplier's financial and operational accounts completely and accurately present every relevant fact about the operating decisions made by its managers, they will not provide a basis for evaluating the soundness of those decisions. This is because cost accounts can show only what happened, not what might have happened. They cannot show the range of asset acquisition choices and tradeoffs the supplier considered, let alone those that should have been considered but were not. As I have noted, under outlay budgets and flexible-price contracts, asset acquisition decisions must be made, but the supplier cannot be trusted to make them efficiently. Consequently, suppliers must be denied some discretion to make managerial decisions.

A fundamental question is how much; that is, to what extent should government customers or their agents replace or duplicate the supplier's managerial efforts? It is necessary to pose this question because before-the-fact controls are costly -- both in terms of out-of-pocket monitoring and reporting costs and in terms of opportunity costs, benefits lost owing to the customer’s inability to exploit fully the supplier's managerial expertise. The government or its agent, the controller, will very seldom be more competent to make asset acquisition decisions than the supplier. The answer to this fundamental question is obvious: the minimum necessary, given the motivations of the service supplier and the incentives confronting her. Sometimes, "the minimum necessary" is a great deal indeed. How much depends on circumstance and the controller's skill in exploiting the opportunities that are created by the supplier's response to institutional constraints.

The problem of figuring out how much constraint is necessary is, perhaps, best expressed in terms of minimizing the sum of the costs that arise out of opportunistic behavior on the part of suppliers (that is, to use the language of public discourse, waste, fraud, and abuse) and the costs of control, both direct and indirect. Economic theory tells us that this optimum is to be found where the marginal costs of controls equal their marginal benefits (Breton and Wintrobe, 1975).

The benefits produced by administrative controls are characterized by diminishing marginal returns. This is simply an abstract way of saying that those controls that produce the greatest payoffs in terms of waste, fraud, and abuse avoided should be executed first. In contrast, the costs of control (the sum of direct and indirect cost of their execution) are characterized by increasing marginal costs. This assertion is, of course, debatable. So far as we are talking about the direct cost of controls -- the out-of-pocket search, bargaining, monitoring, and enforcement costs that they impose on buyer and seller alike, it might be more reasonable to presume constant marginal costs. However, it seems to me that the indirect costs of control -- those which take the form of stifled initiative, dulled incentives, and duplicative effort (Marcus, 1988) -- do probably increase at an increasing rate as the quantity of controls is increased.

These claims indicate that it almost never makes sense to try to eliminate abuse entirely. If the sum of the costs of opportunistic behavior on the part of suppliers plus the direct and indirect costs of controlling their behavior is minimized, some abuse must remain simply because it would be dreadfully uneconomical to eliminate it.18 The point is that controls contribute nothing of positive value; their singular purpose lies in helping us to avoid waste. To the extent that they do what they are supposed to do, they can generate substantial savings. But it must be recognized that they are themselves very costly.

How much more efficient would government be if control system designs were carefully tailored to circumstances? Unfortunately, I don’t have an unambiguous answer for this question. The theory outlined here states that both the ease with which the consequences of operating decisions can be monitored and the desirability of interorganizational competition matter. But most empirical studies overlook the distinction between the subject and the timing of controls. Hence, they don’t actually relate the cost of supplying services to the choice of governance mechanism. Moreover, I would distinguish the costs of mismatching controls from the costs of over-control or micromanagement. The nasty consequences of micromanagement are far more frequently denounced than measured. Nevertheless, my reading of the evidence suggests that mismatched controls may add five to twenty percent to the real cost of supplying services -- over-control can add far more.

Some of this evidence goes to the efficiency of privatizing various services, including custodial services and building maintenance, the operation of day-care centers, fire protection services, hospitals and health care services, housing, postal services, refuse collection, security services, ship and aircraft maintenance, waste water treatment, water supply, and weather forecasting. Because these are common, homogeneous services that do not require large, lumpy investments in extraordinary assets -- indeed, most have direct commercial counterparts, the logic outlined here indicates that they are appropriate candidates for a combination of organizational responsibility and after-the-fact control.

Not surprisingly, the evidence shows that shifting from individual responsibility and before-the-fact controls to organizational responsibility and after-the-fact controls does reduce the cost of delivering these services. In his evaluation of the navy’s commercial activities program, Paul Carrick (1988) of the Naval Postgraduate School found that the introduction of competition reduced service cost in eighty percent of the cases studied, with average savings of nearly forty percent -- the greater the number of competitors, the greater the average savings. Carrick also found that navy teams won over one third of the competitions carried out under OMB Circular A-76, achieving productivity improvements of thirteen percent on average. In these latter instances, the only significant change in governance relations was the shift from a demand concealing to a demand-revealing control system design, since the winning in-house teams were usually the incumbent suppliers.

In a second relevant study, Scott Masten, James Meehan, and Edward Snyder (1991) carefully analyzed the determinants of control costs, holding production costs constant, in naval shipbuilding. Looking at 74 components -- 41 "make" items and 31 "buy" items, classified using benchmarks similar to those outlined here -- they determined that control costs represented about fourteen percent of total costs -- about thirteen percent for make components and seventeen percent for buy components. They also determined that the proper choice of governance mechanism permitted control costs to be substantially reduced. Making the right decisions resulted in control costs that were a third less than if all components had been made internally and half what they would have been if all components had been contracted out.

Several analysts have found that, where appropriate, the substitution of after-the-fact for before-the-fact controls produces similar productivity gains. David J. Harr, for example, reports that replacing standard outlay budgets with responsibility budgets in the Defense Logistics Agency depots were associated with efficiency increases of ten to 25 percent (Harr, 1990, p. 36; Harr and Godfrey, 1991, pp. 68-9). Other analysts make even stronger claims about the significance of the nature and timing of controls. Gordon Chase, for example, asserts that "wherever the product of a public organization has not been monitored in a way that ties performance to reward, the introduction of an effective monitoring system will yield a fifty percent improvement in the product in the short run." (Allison, 1982, p. 16). Productivity increases of this size are not, in fact, unheard of. One frequently cited example of such an increase is the central repair garage of the New York Sanitation Department, which replaced its standard municipal outlay budget with a well designed responsibility budget. Robert Anthony claims that this reform increased productivity by nearly seventy percent -- from a high of 143 percent in the machine repair center to a low of 19 percent in the motor room (Anthony and Young, 1988, pp. 356-7).

William Turcotte’s classic matched comparison of two state liquor agencies reports even larger productivity differences caused by the substitution of after-the-fact for before-the-fact controls (Turcotte, 1974). The organizations studied by Turcotte ran sizable statewide programs featuring large numbers of local retail sales outlets. Furthermore, both defined their missions in identical terms -- maximization of profits from the sale of alcoholic beverages to the public. According to the theory outlined here, this situation called for the use of a rather simple, straightforward responsibility budget to govern local retail sales outlets. One of the states (Turcotte refers to it as state B) did in fact adopt this approach to governance -- treating each outlet as a profit center, holding the outlet’s manger responsible for meeting a profit target, and granting her the operational discretion needed to meet it. The other state (Turcotte refers to it as state A) relied on standard outlay budgets and a comprehensive set of before-the-fact controls. Turcotte reports that one consequence of the difference in the control strategies used by the two states is that direct control costs were twenty times higher in state A than in state B. The indirect costs of control were somewhat less disproportionate, but absolutely far greater in state A than in state B. Furthermore, individual stores in state B were twice as productive as stores in state A. Operating expenses for each dollar of sales in state A were 150 percent higher than in state B, administrative expenses were 300 percent higher, and inventory costs 400 percent higher.

However, both Anthony and Turcotte appear to conflate the choice of governance arrangements with their intensity. New York’s garages and State A’s liquor stores were subject not only to the wrong kinds of controls but probably also to an excess of controls. One of the more melancholy properties of before-the-fact controls is their propensity to proliferate -- excess controls cause failures, which leads to more controls, and then more failure. I would not be surprised if two-thirds of the productivity differences reported here were due to over control.

The evidence also shows some goods are unworthy candidates for after-the-fact controls. The case that has been given the greatest amount of attention by industrial-organization economists is where customers artificially maintain rival suppliers where a single supplier could more efficiently supply the entire market (Anton and Yao, 1990). There is, however, a more interesting case. Consider what can happen when rivals are invited to bid on a fixed-price contract to supply an advanced and, therefore, highly risky or uncertain technology. They will likely respond to such an invitation in one of two ways:

1. If they bid at all, they will bid high to protect themselves against the risk of failure; this means that the price of the service to the customer will be excessively high; or, even worse,

2. One or more of the bidders will underestimate the difficulty of the contract (or overestimate his capacity to meet its terms). He will often be the low bidder, of course, and win the contract. If he is not very lucky, he will then be cursed by his victory. When he fails to deliver, as he mostly will, or threatens to slide into bankruptcy, his customer may have to step in to rescue project and, in some cases, the contractor as well.

Alas, open-bidding contests tend to select suppliers for their optimism (or their desperation), since the bidder with the most optimistic view of a project’s feasibility will usually win the contract. Unfortunately, the most optimistic (or most desperate) bidder is unlikely to have the best understanding of the contract’s technical feasibility. Indeed, he may overestimate its feasibility precisely because of his incompetence to carry it out! This likelihood probably doesn’t matter very much where all of the bidders have the experience needed to manage to a narrow range of outcomes. In that case, either comparative advantage will trump optimism or, if not, the advantage will usually be borderline. But this likelihood is crucial where bidders lack the experience needed to manage to a narrow range of outcomes -- as will usually be the case where advanced and, therefore, highly risky technologies are concerned.

Indeed, where a product or service is highly specialized, a single organization is often uniquely qualified to produce it. Identifying the right supplier is, therefore, frequently the key to getting the best product on time and at a reasonable price. In the private sector this process is often fairly informal. Firms tend to rely on experience and reputation to pick suppliers. A decision to invite a proposal is usually tantamount to a promise to do business. And proposals are more often than not jointly developed.

In the public sector, the process tends to be more formal. Potential suppliers must appear on a list of qualified vendors. Customers must usually request proposals from more than one organization. Requests for proposals (RFP) are supposed to provide detailed explanations of what proposals should include and how they will be evaluated. Evaluations tend to be highly ritualized, with each section of a proposal assigned an explicit numerical score and its overall evaluation based upon the weighted sum of these scores. Only after evaluators have identified the best proposal will the government’s representatives engage in ex parte conversations with the vendor to work out contractual details and nail down a best and final offer.

Nevertheless, these processes have similar aims and, I believe, more often than not produce similar outcomes. Doubtless, these sham battles are wasteful and add to the costs of executing before-the-fact controls, but there is much less waste than when the contract is awarded solely on the basis of price and the winning contractor turns out to be incompetent.

Unfortunately, this happens sometimes. It is generally acknowledged, for example, that the worst defense procurement fiasco in recent memory, Lockheed’s default on the C-5A program and the subsequent Department of Defense bailout, occurred because Lockheed misread the difficulty of designing and building the C-5A. Consequently, Lockheed submitted a bid on a fixed-price, total package procurement contract to design and deliver 150 C-5s that was fifty percent less than Boeing’s, the next highest bid. Evidently, even if Lockheed had known what it was doing, which as it turned out it didn’t, its bid would have been half-again too low. By the time the Department of Defense and Lockheed discovered magnitude of their error, they were in too deep to get out (Gregory, 1989, pp. 107-117). Something similar happened recently with the navy’s A-12 program. Fortunately, when the A-12 development team got into trouble, Department of Defense decided the A-12 was expendable and canceled the contract, thereby avoiding the worst aspects of the C-5A case. Nevertheless, this was evidently a near run thing. In the mid-1980’s, Boeing took a bath on a series of fixed-price development contracts that it sought and won despite lack of expertise. Again fortunately for Boeing and ultimately for the taxpayer, Boeing’s civilian profits were sufficient to make good its military losses.

The lesson suggested by the example of the C5-A is that the total costs arising from mismatched controls are asymmetrical in their composition: if other things were equal, it would probably be far more prohibitive to rely on after-the-fact controls where before-the-fact controls are called for than vice versa. This lesson is reinforced by Masten, Meehan, and Snyder’s (1991) finding that, although making "buy" components would have caused control costs to be about seventy percent higher than they actually were, contracting out "make" items would have caused control costs to increase even more -- nearly two-hundred percent, from thirteen percent of the total value of the items to over thirty percent! But other things are not all equal. Not only are after-the-fact controls easier to use, they are also self-limiting. Where the purchaser relies on demand-revealing controls, over-control produces negative feedback in the form of higher prices or reduced output that causes controls to be cut back. Before-the-fact controls often produce positive feedback that leads to their multiplication. Hence, their costs are subject to no natural limits.

Carrying Legitimate and Necessary Controls to Self-Defeating Extremes

Organization theorists have long understood that failure induces certain predictable responses -- and that these responses, in turn, produce certain equally predictable consequences. Pradip N. Khandwalla (1978), for example, observes that threatening situations always generate pressures for direct controls: standardization of procedures, institution of rules and regulations, and centralization of authority. Michael Crozier (1964) argues that a failure to meet expectations almost inevitably produces a cycle of rule making, more failure, and then more rules. Anthony and Young (1988, p. 562) claim that detailed rules result from encrustation: an abuse occurs, someone decides that "there ought to be a law," and a rule is promulgated to avoid the abuse in the future; but such rules often continue after the need for them has passed. No one who has the power to rescind the rule may ever consider "whether the likelihood and seriousness of error is great enough to warrant continuation of the rule." Jack H. Knott and Gary J. Miller observe that stricter rules and tighter oversight often produces positive short-term results, but that they also exacerbate the factors that cause organizational failure. Furthermore, extra supervisors giving more orders and monitoring effort more closely may make subordinates "even more resentful of their status than before." In other words, the inclination to respond to abuses with calls for more and better rules is normal, as is responding to repeated failure with calls for ever more inflexible and comprehensive rules, greater oversight and closer supervision.

The propensity to devise inflexible and comprehensive rules is, perhaps, nowhere more irresistible than where military procurement is concerned. Consequently, military procurement generates more than 250 million hours of paperwork a year, 90 percent of the federal government’s procurement paperwork (Weidenbaum, 1990, p. 153), and the Department of Defense employs 100 thousand men and women, uniformed and civilian, and spends between 5 and 10 billion dollars each year to buy (not to pay for) the weapons, materials, and supplies it uses. Nearly 50 thousand these employees (including 26,000 auditors) are paid to monitor and enforce compliance with before-the-fact controls. As an example of this propensity to devise inflexible and comprehensive standards, consider the MIL-F-1499 (Fruitcake), 250 tons of which were recently purchased by the army. To preclude abuses on the part of unscrupulous bakers, to make sure there really were some candied fruits and nuts in the fruitcake, to guarantee adequate shelf-life and resistance to handling, and to insure palatability in all the far-flung places of the world where the American army celebrates Christmas, the specifications for the MIL-F-1499 (Fruitcake) were 18 pages long. Plastic whistles take 16 pages of specifications, olives 17, hot chocolate 20, chewing gum 15, condoms 13, and so on (Adelman and Augustine, 1990, pp. 126-7).

This level of detail may be ludicrous, but it is not evidence of over-control. Evidence of over-control requires information on the benefits as well as the costs of control. What about the benefits of control? Well, one agency, the Defense Contract Audit Agency proudly claims that it saved the American taxpayer about $7 billion in 1988 and cost only $1 billion. Its criminal investigations generated an additional $300 in fines and penalties million and cost only $84 million (Dunnigan and Nofi, 1990, p. 368). This sounds like a pretty good deal, even if one allows for the source of the claims. However, it is a generally accepted rule of thumb that monitoring and enforcing regulations imposes private costs of about $20 for every dollar spent by the government (Weidenbaum and DeFina, 1978). Since these costs are ultimately borne primarily by the regulated firm’s customers and since in this instance the customer is the Department of Defense, this multiplier implies that Defense Contract Audit Agency regulation imposed costs of $21 billion to save $7 billion, in the first instance, and $1.76 billion to save $300 million in the second -- in other words, it cost an average of $3 and $6 to save $1, which is consistent with marginal costs of $6 and $12! Evidence that marginal costs are greater than marginal benefits, let alone twelve times greater, is prima facie evidence of over-control.

There is also evidence that the marginal benefits produced by some before-the-fact controls are actually negative. Alfred A. Marcus, for example, shows that increasing the number of safety rules governing the operation of nuclear power plants, together with greater oversight and closer supervision, actually had the effect of degrading reactor safety (Marcus, 1988). Anecdotal evidence suggests that this is often the case where procurement is concerned, especially where demand-concealing governance mechanisms are called for, but where a plethora of rules deny the controller the authority to trade off costs, schedules, and performance (Weidenbaum, 1990).

Finally, excessive reliance on rules often produces organizations that are simultaneously over controlled and out of control. Turcotte (1974, p. 69), for example, found that the managers of retail stores in state A were subject to many more rules and far stricter executive and legislative oversight than their counterparts in state B, but, even so, were far less responsive to the wishes of their political masters. Evidently the managers of retail stores in state A were subject to so many rules that none of them mattered very much. Consequently, over-control led straightaway to loss of control.

Fortunately, public administration is changing, albeit slowly. Most students of public administration have accepted Mosher’s challenge to look outward more, inward less -- to understand a wide variety of institutional arrangements: regulation, incentives in the form of loans and taxes, contracting, and quasi-governmental enterprises. Despite their efforts, however, much of our knowledge remains equivocal. What is the reason for our uncertain progress? Since we have a satisfactory framework for institutional analysis, I believe that it is due largely to an inability to look beyond superficial institutional dissimilarities to their common structural elements -- an inability to see that the entire spectrum of institutional arrangements is put together from a common set of materials and that, to design effective institutions, the materials used must fit together harmoniously. This essay neither promises nor provides a complete answer to the question of how institutions are put together, let alone a complete parts list. It is, I hope, a step in the right direction.