1. A few definitions and categories of controls

Consider a public bureau set up to perform a certain function. Bureau managers have some budget that enables them to command the resources necessary to perform the function of the bureau. Depending on the type of bureau, this budget will be derived to a varying extent from two main sources: either a voted appropriation from tax revenue, or forecast revenue from the sale of goods and services produced by the bureau. Typically, however, bureau managers are not free to decide how they will spend their budget in meeting the policy objectives set for them. The extent of controls on bureaus, after they have received a budget, varies between countries, levels of government, and by type of bureau.

For purposes of exposition, however, such controls can be split into distinct categories, which include the following:

a. Price, quantity, and other controls on the use of labor;

b. Controls on spending in other categories of current expenditure;

c. Controls on capital expenditure;

d. Controls on using receipts (including those from the sale of existing capital stock) for other capital or current expenditure; and

e. Controls on management of cash.

For simplicity, we define the bureau as being associated with a particular function, although in practice a bureau may be charged with administering a number of functions, activities, or programs. The level at which the government chooses to define a bureau’s function is taken as given. The paper’s concern is not with, for example, whether the head of the education department should be free to substitute between expenditure on universities and expenditure on secondary schools, if the government chooses to define these as separate functions. Rather, the paper is concerned with examining the controls that governments place on how bureau managers may combine the factors of production--labor, other current factors, and capital--to produce a bureau’s output as defined by its function.

It is recognized that such an approach is not without its difficulties. In New Zealand, for example, money for teacher education is appropriated in the budget as a distinct program. However, it could be argued that expenditure on teacher training is really an input into meeting the objectives of other programs, such as providing secondary education. It makes little sense to examine a manager’s ability to decide on the mixture of the number and the quality of teachers in the secondary schooling budget if one accepts that expenditure on teacher training is fixed.

A further objection--and possibly a powerful explanation for some of the input controls we observe--is that the government may in practice regard factors that appear to be inputs toward its avowed output objective as output objectives in their own right. For example, while it might provide money to the navy for the purpose of maintaining naval forces at the level of operational readiness needed to uphold defense and security interests (to paraphrase the published objective in the New Zealand budget), the government might control naval capital expenditure separately if it has an additional objective of using the navy’s capital budget for the purposes of regional development.

The term “government” as used in this paper refers to the body that has the power to authorize taxation and public expenditure and to pass laws. It may delegate some of this power to bureaus, and effective power may reside in subsets of this body, but for our purposes, an abstract concept of government as the body with ultimate control over the bureaucracy is sufficient. The term “bureaucrat” as used in this paper is close to that used in Niskanen (1971): “the senior official of any bureau with a separate identifiable budget.” ^1/ Specifically, interest focuses on bureau managers, rather than on all employees of a bureau. When we talk of input controls, we are interested in whether the bureau manager must obtain approval from the government or a control bureau to spend money from its budget on, for example, stationery, not in whether the bureau manager chooses to put limits on stationery expenditure by divisions within the bureau.

2. Price, quantity, and other controls on the use of labor

Labor is generally a tightly controlled input in public bureaus. Many countries have a distinct control agency responsible for setting limits on the numbers and types of staff that bureaus may employ, and for setting remuneration for public employees. Bureau managers are also typically limited in their ability to hire, fire, and promote employees as they see fit.

In New Zealand, a control agency, the State Services Commission (SSC), is considered the employing authority for the public service. The Government, on the advice of the SSC, sets “staff ceilings” (maximum permitted staff numbers) for each bureau. Pay and conditions for public service-wide occupational classes are also centrally controlled. Some authority to make appointments and promotions is delegated to departments, although the SSC has the right to intervene. Senior appointments are made by a panel of civil servants outside the department (not by Ministers), and salaries for these positions are set by the Higher Salaries Commission, whose findings are binding on the Government. ^1/

Recent reforms in this area have been aimed at increasing state pay-fixing flexibility and increasing emphasis on recruitment and retention criteria in setting rates. Staff numbers as a control variable have become less important; the “sinking lid” on staff numbers has been removed, and managers now have some flexibility to adjust staff numbers within staff ceilings. A recently introduced State Sector Bill would, if passed, require the State Services Commission to “negotiate…conditions of employment that will enable each Department to compete effectively for the employees that are needed in order to achieve the principal objectives of that Department.” In part, such moves suggest that the problem with labor inputs in the public sector is not always one of too many bureaucrats getting paid too much, as some of the literature examined in Section III implies. Rather, the existing controls have exposed production inefficiencies in bureaus, owing to too many staff in some areas and not enough in others, and the inability of, and lack of incentives for, managers to do anything about the problem, even if they are trying to minimize costs. ^2/

In spite of this problem, none of the 19 OECD countries (a sample that excluded New Zealand) examined in the OECD’s survey of government expenditure control practices appears to have formally devolved authority for setting pay to government departments (as opposed to state owned enterprises). ^1/ A number of countries however--notably, Denmark, Norway, and Sweden, and also Australia, Canada, and the United Kingdom--have been moving toward allowing substitution between staff and other current expenditure, and flexibility among different categories of staff. Even within this small group of countries, however, there is significant variation; the United Kingdom and Norway retain a control over staff numbers, whereas the control variable in Sweden is financial resources. Where substitutability has been allowed, it is significant that there is typically more freedom to substitute from personnel to other current expenditure than to use other current spending allowances to expand personnel expenditure. In practice, where controls on staff numbers have proved to be a constraint, departments have often been permitted to expand their labor inputs by means other than increases in the numbers of permanent staff; an example is the use of consultants in the United States and other countries. ^2/

3. Controls on spending in other categories of current expenditure

Spending on nonlabor current inputs is often controlled in budgetary systems, either by means of limiting particular items at the time money is appropriated, or by requiring managers to obtain approval from either control agencies or the government for expenditure after money for a program has been appropriated. The former type of control--line-item appropriations of such current expenses as travel, office accommodation, printing, stationery and supplies, maintenance, and others--is sometimes used, although many countries have moved away from such detailed specification of operating costs in the budget.

In New Zealand, nonpersonnel operating costs are appropriated as a single item, and the delegations to spending department managers as to the use of those funds have been substantially rationalized and liberalized.

The reduction in controls on current expenditure is probably the area where the most significant relaxation has taken place. For example, Australia, Canada, Denmark, Ireland, Sweden, and the United Kingdom have recently reduced detailed controls over current expenditure.

4. Controls on capital expenditure

Spending for the purchase or construction of capital assets is usually appropriated and controlled separately, in a particular program, from current expenditure. This is because capital assets typically both provide a flow of services and commit a flow of costs over a period greater than that for which money is appropriated (in an annual budget, for example), and since the cost of capital does not enter departmental budgets, it would be inappropriate to allow free substitution between current and capital expenditure. In addition, capital projects are often larger and more visible than individual current expenditures, and so are subject to control for political reasons.

As we shall see in Section III, a further rationale is provided by some models of bureaucratic behavior that hypothesize that bureaucrats have fewer incentives than private sector managers to take into account the consequences of their decisions. Under this hypothesis, bureau managers, given the freedom to do so, might spend more on capital assets than would be justified on the basis of net present values at society’s rate of time preference, even with the inclusion of the cost of capital in budgets.

Capital expenditure in New Zealand is subdivided for control purposes into expenditures subject to Works Program procedures, such as building and engineering projects, and “administrative” capital equipment--for example, office equipment, furniture, and motor vehicles. While the Minister of Finance has delegated some control powers to the Ministry of Works (which checks proposals for accuracy and feasibility of estimates), major capital works still require approval from the Minister of Finance, Cabinet Committee, or Cabinet, even after money has been included in the Appropriation Act. This is true only for government departments, however; with the establishment of many of the goods and service producing bureaus as public corporations, much of the public sector capital works program has been removed from annual budget control procedures. ^1/

No OECD countries allow departments to switch freely between current and capital expenditure. Within the capital allocation, countries are allowed varying degrees of flexibility. Ireland is typical of countries allowing some delegation, where, “in relation to capital expenditure, authority for expenditure on established programs such as public housing, hospitals, roads, and schools has been delegated to Departments within the overall parameters and targets for the program previously agreed with the Department of Finance. For certain major projects within such programs, however, and for all major one-off public sector capital projects, specific Finance authority for the expenditure must be secured.” ^2/

5. Controls on using receipts (including those from the sale of the existing capital stock) for other capital or current expenditure

Bureaus typically control some fixed assets, built up over time from annual capital expenditures. The resources represented by the opportunity cost of the capital in those assets are generally not freely transferable by agencies to other uses. In other words, government agencies cannot sell existing assets and use the proceeds to purchase other assets or to finance current expenditure. This restriction arises out of budgetary systems that prevent expenditure of public money without appropriation by the elected representatives. Devices such as revolving funds have been set up in some countries (including New Zealand) to enable some bureaus with quasi-commercial activities to spend receipts from sale of assets and from other sources. Most government departments, however, have no incentive to sell assets with low rates of return, as they face a zero cost of capital and are constrained from reallocating those assets to areas of higher return.

Denmark and Norway have recently shifted focus to net funding as a control variable, allowing agencies to use “minor” additional revenues without prior authorization. Although primarily designed for additional revenues from user charges on goods and services, these “minor” revenues might in theory include receipts from asset sales, and, if so, provide some incentives for asset management.

6. Controls on management of cash

Having been allocated an annual budget, bureaus are not faced with a cost of capital. They are not free to lend from, or borrow against, their budget. ^1/ Even if they could, insofar as they are not permitted to spend receipts, they have no incentive to do so. Some countries control the rate at which the annual budget may be spent during the year. Ireland, for example, uses the agreed profile of monthly expenditure during the year as a set of monthly cash limits, the Minister of Finance must approve expenditure of more than the agreed amount, and monthly savings cannot be carried forward. Although New Zealand does not follow this system, nevertheless the rate of departmental spending is monitored so as to track actual against budgeted expenditure.

Aside from state owned enterprises, OECD countries have not allowed freedom to departments to manage cash balances outside the public account.

7. Summary of country practices

A common theme in budgetary reform in a number of countries has been a desire to decentralize authority for the use of inputs--to “let the managers manage”--in return for a greater focus on output and performance, and accountability for that performance. ^1/ The OECD notes that “some countries are delegating much more authority to departments to shift resources between various inputs to programs--salaries, capital, travel, consultants, etc..” ^2/

In practice, however, very few moves to delegate control have been made (with the notable exception of the state-owned enterprise area) in the following areas:

a. The ability to set wages and salaries;

b. The ability to increase staff numbers by reallocation of expenditure from other areas;

c. Capital expenditure, especially in the ability to substitute between capital and other expenditure; and

d. The ability to manage cash balances.

Part of the reluctance to delegate further control may be linked to difficulties that countries have experienced in measuring output and holding managers accountable for performance. However, there may also be features associated with the inputs subject to control that would discourage managers from using them efficiently if they were free to do so. We turn now to a survey of the economic literature to examine this possibility.

1. Introduction

The effectiveness of bureaucrats in meeting society’s demands for publicly provided goods and services can be seen as a subset of the wider theory of public choice, defined by Mueller as “the economic study of nonmarket decision-making.” ^1/ There are a number of stages, in translating the package of publicly supplied goods embodied in society’s social choice function into actual delivery of those goods, at which interactions between agents can affect the outcome. The kind of voting rules affect how social preferences are transmitted to elected representatives. The policy platform of the group of representatives may differ from policy preferences of individual representatives. Finally, the actions of the bureaucracy may not reflect the preferences of the relevant group of representatives that directs the bureaucracy. ^2/

This last stage is the primary concern of this section: the problem of whether, and under what conditions, utility-maximizing bureaucrats (specifically, bureau managers) will act in accordance with the preferences of their political masters. However, it should be noted that the relevance of the discussion for welfare-improving outcomes rests on the linkages between the political representatives and the social welfare function. Tullock, for example, has suggested that bureaucracy may expand beyond the welfare optimal level because bureaucrats express an influential preference for an above optimal-sized bureaucracy through the voting system. ^3/ If true, this suggests that existing voting rules fail to link society’s preferences to the policies of elected representatives. Since the rules governing the relationship between the elected representatives and the bureaucracy are themselves part of the “platform” of issues on which the representatives are judged, however, we make the simplifying assumption of looking for rules formulated as if representatives were acting in society’s interests.

In general then, the following analysis is not dependent on the type of government. While a system of political representation may allow governing politicians to pursue their own objectives at the public’s expense, there is no reason why they should cede this freedom to the bureaucracy. ^1/ A control system that gave independence to bureaucracies would have to be enacted by the very politicians whose abilities to pursue their own objectives at the public’s expense would thereby be thwarted. It would, therefore, be against their best interest. ^2/

2. Tullock and Downs

Gordon Tullock, in The Politics of Bureaucracy, ^3/ and Anthony Downs, in Inside Bureaucracy, ^4/ were among the first economists to examine bureaucratic behavior using the assumption of bureaucrats as utility maximizers. Both derive a number of hypotheses from this assumption, although the difficulties of devising empirical tests of these hypotheses may explain why their work has not received as much attention as it might have in the ensuing literature. ^5/

Tullock hypothesizes that an individual bureaucrat, if faced with the choice of pursuing his own goals and pursuing those of the organization as he sees them, will choose the former. Although not all bureaucrats have to behave this way all the time, Tullock and Downs both suggest that bureaus will come to be dominated by those that do, as they advance faster in the organization.

Downs’ bureaucrats have a wide range of goals-he gives as examples “power, income, prestige, security, convenience, loyalty (to an idea, an institution, or the nation), pride in excellent work, and desire to serve the public interest.” ^6/

Downs develops a model of the bureaucracy using the motivations of bureaucrats, and derives a number of hypotheses relevant to the problem of controlling bureaus--for example, that attempting to control large organizations tends to generate other monitoring organizations, and that the amount of control by these monitoring bureaus tends to increase over time, regardless of the extent or nature of the activity being monitored. Downs also suggests that a bureau has incentives to evade or counteract attempts to control it, so that attempts to increase control may not result in improved actual control. ^1/

It is difficult to rigorously test Downs’ theories, but the idea of the self-generated nature of controls, and the tendency for them to become more complex and costly over time, is perhaps reminiscent of Mancur Olson’s (1982) work on the correlation between growth and the power of interest groups. If plausible, this idea suggests that a reduction in controls could save resources both in bureaus doing the monitoring and in those being monitored (which put resources both into complying with controls and attempting to evade them), and that these savings might outweigh the “waste” that would occur in the absence of controls.

While we do not attempt to prove or disprove this hypothesis, ^2/ it does provide a possible counter-argument to the application of principal-agent theory (Section 6 following), which suggests that observed monitoring and control mechanisms survive in different organizational structures because they represent the most efficient solutions (given information and transactions costs) to the problem of ensuring that bureaucrats act in the government’s interests.

3. Niskanen and the bureaucrat as budget maximizer

Niskanen’s model of bureaucratic behavior hypothesizes that those variables that enter the bureaucrat’s utility function lead to a proxy maximand of budget maximization. ^3/ The variables Niskanen lists include salary, perquisites of the office, public reputation, power, patronage, ease of managing the bureau, and ease of making changes. He contends that all of these variables (in a static framework) are a positive monotonic function of the total budget of the bureau. This assumption, according to Niskanen, does not imply “a cynical interpretation of the personal motivation of bureaucrats.” ^4/ Rather, bureaucrats who develop expertise in particular fields will acquire a dedication to those fields, which they will come to identify with the public interest.

The other critical assumption in Niskanen’s model is the idea that bureaus, in exchanging their total output for a fixed budget, can act as monopoly suppliers. While the political executive is a monopsonist purchaser of those services, leaving the indeterminate case of bilateral monopoly, Niskanen suggests that the sponsor’s unwillingness to forgo the services supplied by the bureau, together with the bureau’s superior information regarding its costs of supply, puts the bureau in a stronger bargaining position, enabling it to extract nearly all of the consumer surplus generated by its output.

The budget-maximizing assumption, subject to the constraint that the budget must be equal to or greater than the minimum total costs at the equilibrium output, leads to larger budgets than would be demanded by the government for that level of output. The government’s demand function for the bureau’s output provides an upper limit to the size of budget it will approve. As Migue and Belanger point out, however, the goal of budget maximization rests on the proposition that bureaucrats maximize output subject to their budget constraint (as long as the marginal value of output to the government is positive), and thereby leave themselves no fiscal residuum to enjoy as perquisites. The inconsistency in this model limits its value in helping devise optimal controls on inputs. In the words of Migue and Belanger, “to the extent that bureaucrats play no role in the allocation of resources, there is no place for an economic theory of bureaucracy.” ^1/

While the inability to build production inefficiency into Niskanen’s initial formulation of his model limits its usefulness, ^2/ a number of writers have found his assumption that a bureaucrat’s utility may be positively related to the budget of the bureau a valuable starting point. We go on to examine some models that use this assumption while at the same time allowing for production inefficiency.

4. Models allowing production inefficiency and factor bias

5. Criticisms of the budget maximization assumption

The assumption of budget maximizing behavior by bureaucrats has been criticized by a number of writers on various grounds. A common theme of these criticisms has been the link between the bureaucrat’s utility function and the budget size. Margolis, for example, suggests that “to understand the behavior of the group of ambitious and energetic persons who reach the top of bureaus we should consider their lifetime aspirations and, therefore, the reward structure should be specified for a sequence of posts and for behavior which is welcomed by the bureaucrats’ superiors…. Our growing society may reserve its largest rewards for the expansionists, but where there are no impersonal ways to eliminate the unwanted, as in the political sphere, hatchet men are well rewarded.” ^1/ In other words, a government with a policy of reducing expenditure is likely to reward bureaucrats who help to achieve that aim.

Prestige and salary are not necessarily correlated with bureau size. In many countries, the most prestigious (and well paid) public service positions are those heading relatively small departments, such as those in charge of foreign relations, or the central financial and personnel control agencies. ^1/ Furthermore, as Mueller points out, “salaries across government bureaucracies tend to be much more uniform than are salaries across companies…,” and with limited tenure in top positions “expanding the size of the bureau, even if size and salary were positively correlated, would not be likely to benefit directly the bureaucrat who brought about the increase.” ^2/

Musgrave criticizes the limitation of variables in the bureaucrat’s utility function to “personal economic gain and power.” ^3/ Considerations such as “duty, respect of one’s colleagues, realization of what one considers to be a ‘good society,’ and the satisfaction of having contributed thereto” may also be important. ^4/ (A possible counter to this criticism might be that even if bureaucrats are motivated by the desire to do a “good job,” they may still be led to compete for budgets and staff not for personal gains, but to better carry out the bureau’s function.) However, the dependence of the budget maximization assumption on the variables entering the utility function raises a more fundamental concern. Migue and Belanger note that “writers who have suggested these various independent variables in the utility function have allowed themselves the suspicious liberty of scrutinizing the manager’s psychology in a manner that economists have generally refrained from doing elsewhere.” ^5/ It may be difficult to design incentive systems for bureaucrats without engaging in some speculation on such variables, however.

The assumption that a bureau has significant bargaining power in exercising its monopoly position has been questioned by a number of writers. ^6/ Budgetary control procedures and the political incentives to prevent excessive growth of budgets are seen as being given too little weight. Miller and Moe suggest that if the government (represented by a legislative committee) conceals its demand function for a bureau’s output, a bureau’s best strategy may be to expose its true cost function for the supply of its output, with the government choosing a desired output level and budget. ^1/ Breton and Wintrobe point out that, while a bureau may be a monopoly, the bureau head is not in a monopoly position--if he refuses to supply the bureau’s output, he can be replaced. Rather, any bargaining power possessed by the bureau derives from the fact that information about the minimum cost of supplying the bureau’s services is not costlessly available to politicians.

The lack of differentiation between different types of bureaus and bureaucrats is also seen as a weakness. Incentives of bureau managers may differ from those of other bureau staff, and it may not be useful to draw inferences about the objectives of the bureau from the preferences of the bureau manager. ^2/

De Alessi’s suggestion that capital intensity in the public sector can be explained by bureaucrats’ rewards has been questioned by Margolis, who suggests that other explanations are more plausible. He puts forward the incentives created by an annual budget cycle, a zero cost of capital facing bureaucrats, and the opportunities to use capital projects for political patronage as being among such explanations. ^3/

Finally, the empirical support for the budget-maximization hypothesis, in the few attempts that have been made to test it, is slim. McGuire (1981) showed that, if agencies maximize their budgets, demand for their output always appears to be elastic with respect to the cost of the output. However, he presents the results of three empirical studies on the demand for state and local government services which indicate that, if anything, demand is relatively inelastic. The relatively inelastic demand for labor employed by state and local governments in another study cited by McGuire is also inconsistent with the budget-maximization hypothesis. Conybeare (1984) assesses this and other evidence and concludes that “in general, the empirical studies offer little support for the budget maximization hypothesis.” ^4/ A study by Hood, Huby, and Dunsire (1984) finds at best mixed support for the link between budget increases and bureaucrat utility, at least with respect to staff numbers and salaries. Data presented for the period 1972 to 1983 in the United Kingdom suggest that, at least in the aggregate, bureaucrats have been relatively unsuccessful in translating budget increases into staff and salary increases (one could, of course, argue that controls over these inputs prevented them from doing so). Survey data presented by Sigelman (1986) did not find that bureaucrat-managers generally supported major increases in their bureau budgets (although it could be argued that this is not inconsistent with budget maximization if bureaus are already in an equilibrium, extracting all government “surplus” in demand for the bureaus’ outputs).

1. Application of principal-agent theory to the public sector

Although the application of general agency theory to government has been quite recent, some of the writers we have been considering recognized the nature of the problem. Tullock, for example, invented the term “control loss” to refer to the discrepancy between management desires and subordinate actions in his 1965 work “The Politics of Bureaucracy.” ^2/ Breton and Wintrobe recognize a cost-benefit trade-off of “control devices” for politicians--the cost of control devices measured against their benefits in terms of the reduction in excess budgets their use makes possible. ^3/ From their analysis, Breton and Wintrobe infer that the cost of control devices is a function of the costs required to obtain information on the cost of production for different outputs of a bureau. If the government knew the minimum cost of producing the desired output, it could constrain the bureau at the budget and replace any bureau head who did not produce the output. Notice, however, that the term zero control costs, in the sense that Breton and Wintrobe use it, does not imply that agency costs are zero. There are agency costs associated with assessing output, since a bureau head could still use the budget to produce an output different from that desired by the government. ^4/ The limitation of their framework is its focus on the budget size as the principal’s control target.

De Alessi also recognizes the agency problem in his 1969 paper: “the weaker the correlation between d’s [the individual decision maker] welfare (as he sees it) and that of higher level decision makers (as they see it) the greater is the extent to which d is likely to increase his own welfare at the expense of the latter.” ^1/ However, he stops short of generalizing this approach by taking utility maximization to imply that the individual decision maker will seek to expand the resources under his supervision.

If we take the government as principal, and the bureaucrat-manager as agent, the budget maximization framework can be viewed as asserting that a wide divergence exists between the utility function of the government and that of the bureaucrat. ^2/ Input controls can be seen as constraining the use of inputs beyond the point where they would be negatively correlated with utility for the government but still positively correlated with utility for the bureaucrat. The turning point in the function relating input use to government utility--if the government had perfect and costless knowledge of the bureau’s production function--would occur when the marginal products of the bureau’s inputs were equal. In this case, however, there would be no need for the bureaucrat-manager, as all the decisions about input use could be placed in the hands of the principal. The whole point of the principal-agent relationship is that there are gains from specialization to be obtained from having different parties performing different roles.

Of course, it is impossible for governments to have costlessly available information about the cost-minimizing mix of inputs. Even if output could be perfectly measured and bureaucrats presented all relevant information to the government, the government would face other, unavoidable costs, as Downs points out:

a. it still requires time, effort, and sometimes money to comprehend the meaning of data, even if provided costlessly; and

b. decision makers have only limited capabilities regarding the amount of time they can spend making decisions, the number of issues they can consider simultaneously, and the amount of data they can absorb regarding any one problem. ^3/

The “optimal” degree of input controls therefore becomes a cost-benefit trade-off. The cost of input controls--that is, the degree of delegation from the principal to the agent--is the reduction in the gains from specialization, which provide the rationale for the principal-agent division. The benefits of input controls arise from the reduction in behavior by the agent that is utility-reducing for the principal. ^1/

Unfortunately, however, there is more to the problem than this “simple” cost-benefit calculation. The gains from specialization, and the divergence of the agent’s behavior from the principal’s desired behavior, are not exogenous--the principal has the power to make decisions regarding both variables. We spend some time developing these points before examining policy implications.

2. Aligning the interests of bureaucrats and government

In a principal-agent framework, we can view the implicit contract between the government (the principal) and a bureaucrat (the agent) as one in which the government assigns control of resources (including the budget) to a bureau in return for some output. The more the principal believes that the agent’s actions will diverge from the principal’s interests, the less decision-making power the principal will be willing to delegate, other things being equal. The principal can affect the likelihood of the agent’s actions diverging from the principal’s interests by the incentive, monitoring, and bonding systems that are set up.

3. Factors affecting agency costs

The size of the agency costs incurred for each bureau depends on the particular characteristics of the bureau, and the bureaucrats within it. Agency costs will be lower, for example, when the correlation is close between utility maximizing behavior for the bureaucrat and for the government. Jensen and Meckling describe this variable as “the tastes of managers.” ^3/ Peirce states that bureaucratic “failure” is more likely when “the course of action to which the bureaucrat is led by his personal interests conflicts with the legislated objective.” ^4/ Breton and Wintrobe discuss the importance of “loyalty” in bureaucracies as a means of mitigating the need for costly “formal control devices.” ^1/ Fama and Jensen, in discussing the case of nonprofit organizations, note that “the altruism of internal agents allows low cost control of agency problems and acts to bond donors and customers against appropriation.” ^2/

The pattern of prevailing values and cultural characteristics and the way in which they affect the behavior that is approved and disapproved of in a society has an important effect on the extent to which different variables enter managers’ utility functions. Obviously, in countries where corruption and bureaucratic nepotism are usual and acceptable, bureaucrats might be less concerned about the potential effects on their reputation, prestige, and job security of engaging in such practices, and agency costs would be high.

Hofstede suggests other cultural characteristics that might help explain the differences in the degree of delegation observed in different countries. ^3/ For example, Ozgediz suggests that “in countries where strong uncertainty avoidance [risk and change aversity--the type of bureaucrat that Downs labels a “conserver”] is observed (for example, Belgium, France, Greece, Japan, Peru, and Portugal), individuals feel threatened by uncertain and ambiguous situations and avoid them by providing greater career stability and more formal rules. …In countries which tolerate large power distances between people [society’s attitude toward power and its distribution]…the use of hierarchical authority and rules are more readily accepted than in small power distance cultures (such as Ireland, New Zealand, and the Scandinavian countries).” ^4/

The ease with which the manager can pursue his own preferences, when they differ from those of the government, also affects agency costs. The need to obtain outside approval for expenditures that yield personal benefit--for example, for entertainment--can be seen as an attempt to make it more difficult for the manager to pursue his own preferences at the government’s expense. The government’s attitude toward risk may also affect its monitoring of the bureaucrat’s activities: if it perceives a high political risk from the exposure of “waste” in terms of perquisites appropriated by bureaucrats, it may control these areas beyond the point where the marginal costs and benefits to society are equal.

Agency costs also depend on the costs of monitoring and bonding activities, which in turn depend on such variables as the size of the bureau, ^1/ the complexity of its operations, ^2/ the geographical dispersion of operations, and the attractiveness of available perquisites. ^3/ The incentives for private interest groups to monitor the bureau’s performance, which affect the agency costs borne by the government, depend on the magnitude and the diffusion of costs and benefits of the bureau’s operations.

The cost of measuring and evaluating the manager’s (agent’s) performance, the cost of devising and applying an index for compensating the manager that correlates with the government’s (principal’s) welfare, and the cost of devising and enforcing specific behavioral rules or policies also affect firms’ agency costs according to Jensen and Meckling (1976, p. 328).

Finally, Jensen and Meckling point out that the market for managers and the cost of replacing them affect agency costs (1976, pp. 328-29). The cost of replacing the manager is smaller the less knowledge specialized to the bureau his responsibilities require, the easier it is to evaluate his performance, and the lower are replacement search costs. The government’s costs also depend on the legal framework under which bureau managers are hired and fired, and the costs of altering that framework.

It should be noted that encouraging competition among bureaus--as suggested by Niskanen and others--will not necessarily reduce waste. ^4/ The incentives that exist for the government to ensure that the bureaucracy minimizes costs are independent of the number of bureaus supplying a particular service, and the government could simply replace the manager rather than giving resources to a competing bureau to produce the desired output with less waste. ^1/ Only if the cost of setting up another bureau is less than the cost of replacing the manager with one who agrees to provide the agency’s output at lower cost, might it be worthwhile to set up bureaus with duplicating functions.

1. Rationales for input controls

Input controls are a way of dealing with the divergence of interests between the government and the bureaucracy, as principal and agent, respectively, in this relationship. The government could eliminate agency costs by making all decisions on input use itself, but not without (presumably) enormous efficiency costs caused by delays in the decision-making process--given the number of decisions it would have to make--and by its lack of expertise in some of its widely varied operations. Because of this cost-benefit trade-off, it generally makes sense for the government to delegate at least some of its decision making powers over input use.

The budget maximization literature examined in Section III provides what might be termed a “moral hazard” rationale for controls over certain inputs that have a higher marginal utility for the bureaucrat than for the government. For example, the ability of managers to set their own salaries is controlled in most organizations because moral hazard is particularly strong in this area. It may be more difficult to justify restricting the bureaucrat-manager’s ability to set salaries for other bureau employees on moral hazard grounds, although without cost-minimization pressures, managers might find it simpler to use remuneration increases in preference to other management techniques for retaining employees. While alternatives to salary increases might be more difficult and time-consuming for managers, they might also be more cost-effective in certain cases. The Williamson model examined in Section III provides a possible rationale for controls on staff numbers. It hypothesizes that bureaucrats derive personal benefits from staff size and would, if unchecked, expand it beyond the cost-minimizing level. Tullock’s public choice arguments, based on the power of the bureaucracy as a voting bloc, also suggest that staff numbers, in the absence of controls, would expand beyond the cost-minimizing level.

Other input controls can also be defended on moral hazard grounds. Without controls, according to Niskanen’s hypothesis about the elements in the bureaucrat’s utility function, bureaucrats would consume a higher level of current expenditures having some value as perquisites than they would if they were acting to minimize the costs of producing a given output--for example, travel and entertainment. The inability of bureaus to perform their own cash management may also be defended because the appropriation of perquisites is prevented. For example, a United Nations study defends the use of the central account as “a safeguard against corruption; no unused balances are lodged in departmental accounts which might be misappropriated or lent out for private benefit” (1987, p. 2).

Input controls can also be justified by the incentives, suggested in some of the literature we have examined, for bureaucrats to use too low a discount rate in making spending decisions. ^1/ De Alessi suggests that the bureaucrat uses these rates because he is interested only in maximizing pecuniary and nonpecuniary benefits to himself during his period in office, while Margolis has suggested he uses them because of the mechanics of the annual budget cycle and the zero cost of capital he faces. If either assumption is true, the use of too low a discount rate would justify having controls on any current spending with implications for future spending, including capital expenditure, the use of receipts, and controls on cash management. Given public service constraints on firing personnel, controls on permanent staff numbers might also be included (and might explain why the use of consultants to augment staff resources is tolerated, as future expenditure is not committed).

The literature examined in Sections III and IV provides, then, two main justifications for public sector input controls: bureaucrats derive utility from certain inputs aside from their contribution to bureau outputs; and bureaucrats have incentives to use too low a discount rate in making spending decisions.

2. Rationales for recent delegation

If the justification for input controls is valid, then why have some moves to greater delegation--examined in Section II--taken place? Consideration of the variables affecting agency costs between the government and bureaucrats (Section IV.3) suggests some possible reasons. First, increases in the size and complexity of government operations may have increased both the direct administration costs of input controls and the gains from specialization from allowing bureau managers more decision-making authority. Both variables would increase the amount of the residual loss (or appropriation of utility enhancing perquisites by bureaucrats) that it would be efficient for the government to tolerate. ^1/ In assessing the plausibility of this explanation, we question whether it suggests a way of discriminating between those countries examined in Section II that have undertaken moves to liberalize controls on current expenditure and those that have not. Although the data provided in Section II are not sufficiently detailed to provide a rigorous test of this hypothesis, casual observation of the differences in growth in expenditure on current goods and services and capital (so as to exclude nondiscretionary spending) between countries that have relaxed controls and those that have not suggests that the hypothesis does not explain the differences. ^2/

Another possible explanation for increased delegation of current spending is that in those countries where this has taken place, other mechanisms have been found to provide an adequate substitute for input controls. Alternatively, agency costs might simply be lower. The kind of cultural differences explored in the Hofstede study (see Section IV.3) might provide some explanation, as might the prevailing attitudes toward working in the bureaucracy, and the selection and performance-monitoring techniques of bureaucrats. While a more careful examination of this line of enquiry might explain differences in control regimes between countries, it may be less useful in explaining changes within countries.

Explaining intra-country changes would require a more detailed country-by-country examination than has been attempted in this paper. Such studies might look at whether other reforms, such as attempts to introduce a wider variety of output measures in the United Kingdom or the introduction of limited contract employment for senior Australian public servants, were connected to the decision to relax input controls.

A final possibility is provided by Downs’ laws of “Ever Expanding Control” and “Counter Control” (see Section III.2). If plausible, they might suggest that some governments have decided that some input controls are largely ineffective and can be removed without adverse effects.

Before concluding this discussion, we think it would be worthwhile to review those areas in which there have not been moves to greater delegation. They include the ability to set wages and salaries and to increase staff numbers, discretion over capital expenditure, and the ability to manage cash balances. It is significant that these controls could all be explained in the discussion above because of their implications for future expenditure, and for the hypothesized incentives for bureaucrats to use too low a discount rate. Although governments feel that satisfactory methods exist for controlling the moral hazard problems that arise in certain current expenditures, they have not been able to devise ways (especially in the noncommercial parts of government) of providing incentives for bureaucrats to use the social discount rate in making expenditure decisions.

3. How far should delegation of control go?

The final question to be considered is the extent to which governments should delegate control over public sector inputs to bureaucrats. First, since the government, as principal, bears the costs of, and derives benefits from, input controls (even if indirectly, through the preferences of voters), the government should decide how much decision-making power to delegate to each bureau. It will base the decision on how easily it can assess performance (which in turn depends on factors discussed in Section IV.3, such as bureau size and the complexity of its operations and objectives); on the extent to which its performance assessment will affect the manager (through formal and informal penalties and rewards--the incentive structure); and even on its assessment of the manager’s “loyalty”--the extent to which government thinks the manager will act in its interests. Such judgments, while qualitative, are made; according to the New Zealand Treasury,

“there is no reason to suppose that the performance of public sector managers is in any sense ‘unmeasurable.’ It is clear that Ministers do acquire an appreciation of the relative strengths and weaknesses of senior public service managers, through their individual contact with their own officials, consideration of a range of policy advice at Cabinet and Cabinet Committees, and experience during the budget review process.” ^1/

These factors are not independent--the assessment of the manager may itself depend on how he assists the performance assessment process, for example, by providing accurate information on the bureau’s activities.

The variability of agency costs between bureaus suggests that the optimal level of delegation is likely to vary from bureau to bureau. Although it is possible that the cost of deciding on and implementing different regimes of delegation outweighs the benefits, there is a case for presenting choices to government on varying degrees of managerial independence for different bureaus.

An example of the variability in delegation between bureaus is provided by state-owned enterprises. In some countries, managers of state-owned commercial enterprises have a much greater degree of decision-making power than their counterparts in other government agencies; they can control staff numbers and salaries, substitute between capital and current expenditures, and manage their capital stock and their cash. Agency costs in state-owned enterprises may be higher than for privately-owned firms because institutional arrangements to relate the manager’s utility to the performance of the enterprise have to be set up. In the private sector, the manager’s job security is related to the performance of the enterprise because of the possibility of takeovers or bankruptcy, and because the manager may have equity holdings in the firm, the value of which will also depend on its performance. ^2/ In both the private and public sectors, however, the performance of the enterprise can affect the manager’s reputation, future job prospects, remuneration, and personal satisfaction. Even the public sector manager’s short-term job security can be related to the performance of the enterprise if the government reviews the manager’s tenure based on performance.

The important point is that a manager’s incentive to act in the interests of the principal can be put in place in both the public and private sectors, even though the incentives created by stock holdings for the private sector manager are more difficult to replicate in the public sector. Rather, concerns about the efficiency of state-owned commercial enterprises relative to privately-owned enterprises arise because of the incentives of the principal to monitor the manager’s performance--there are good reasons why the government might have less incentive than a private shareholder to ensure a commercial enterprise operates efficiently.

We can defend the delegation of decision-making authority to the managers of publicly and privately-owned enterprises then, because it is possible to devise incentives that align managerial utility with performance in running the enterprise. A plausible reason for the difference in delegation between government-owned commercial enterprises and other public bureaus might be that it is more difficult for the government to assess managerial performance in noncommercial bureaus. While a good deal of effort has been put into developing quantified output measures for noncommercial bureaus in many countries, the experience with such measures has been mixed, at best. However, input controls do not substitute for output measures, and so output measures are not necessary prerequisites for further delegation of control over inputs. For example, if controls on travel expenditure require bureaucrats to submit travel proposals for approval, the government must still judge how the travel will contribute to the bureau’s output in deciding whether to approve them.

One can infer from the discussion in this paper that input controls are not the only method of dealing with the principal-agent problem. For example, an important alternative is to provide incentives to bureaucrats to act in the interests of the government. Designing incentive structures may require moving outside the realm of economics to “scrutinize the manager’s psychology,” in the words of Migue and Belanger (1974a, p. 28). However, incentive structures can be set up, based on a range of variables that might appear in the utility functions of bureau managers. Insofar as salaries vary between bureau managers, they should be influenced by the government’s assessment of performance, if not set by the government itself. Some method of periodically reviewing the bureau manager’s performance, with the assessment affecting future job prospects and continued tenure in the position would strengthen the principal-agent relationship between the government and its bureau managers. Such reviews could be performed by the government or by some independent body that took the government’s views into account.

Ex-post monitoring and disclosure requirements can substitute for ex ante controls if the assessment of a manager’s performance affects his utility. In New Zealand, for example, the details of overseas travel undertaken by politicians are regularly published, and scrutiny by the electorate provides a check on the appropriation of travel as a perquisite. Incentives linking the utility of government and bureaucrats more closely would allow greater reliance on ex post monitoring for bureau managers as well. One result might be that bureau managers would voluntarily seek approval for expenditures when they were unsure whether or not they would be utility enhancing for the government.

While incentive and monitoring systems may be able to deal with the moral hazard problems in current inputs, the problem of intertemporal resource allocation justifying controls on capital, management of cash balances, and staff may still remain. One solution would be to address the intertemporal nature of the inputs. An obvious example is in the staffing area, where bureau managers might be given greater power to hire and fire employees and adjust their pay so that labor could be treated as a current input, without involving any future spending obligations. The impact of such changes on the possible role of security of tenure as a bonding arrangement would have to be considered, however. Examining ways of confronting managers with the cost of capital (if Margolis is correct in viewing the zero cost of capital as part of the problem) is another possibility. Implicitly this might involve bureaus “borrowing” from the government to finance capital expenditure, which they would be required to service out of current expenditure.

4. Summary and conclusions

The basic question posed in this paper is whether economics has anything useful to say about how much decision-making power governments ought to delegate to public sector managers in the use of inputs. While far from providing all the answers, the literature that we have examined at least provides us with a useful framework for thinking about the problem and for asking the right questions.

In examining the types of controls being used in current budgetary systems, we noted both the moves away from some controls in a number of countries, and the control that governments, almost universally, retain over particular categories of inputs (except for state-owned enterprises in some countries). To explain these observations, we turned to theories of bureaucratic behavior. Not surprisingly, the more variables are included in these models, the more realistic they become, but the less clear-cut are the predictions they yield. We turned to principal-agent theory as a way of providing a useful framework for examining the problem. Among the lessons from this framework are that input controls are one of a number of ways--all of which involve some costs--of dealing with the agency problem, and that the optimal degree of delegation is a function of the correlation between the bureaucrat’s utility maximizing behavior and the government’s utility. The optimal level of input controls is, in theory, a standard maximization problem, but in solving that problem we have to cross the boundaries between economics and other disciplines.

From this approach, we learn it is necessary to examine the incentives facing bureau managers when deciding on the extent of delegation. That choice between input controls and other solutions to the agency problem must ultimately be made by the government, as the principal. The government’s decision will depend, in part, on its speculation about the utility functions of bureaucrats, which might help to explain the differences between countries and over time in the degree of delegation examined in this paper. The importance of prestige, reputation, and career prospects to bureaucrats, and the extent to which these depend on the government’s assessment of their performance depend on a diverse range of variables--for example, the size and career structure of the public sector, the size of the market for managers, and the prevailing ethics and extent of corruption in a country. In deciding how much decision making power to delegate to a particular manager, the government must make some judgment as to the relative costs and benefits to the bureau manager of seeking to maximize his own utility where his actions would be utility-reducing for the government.

As a result, realizing further gains from specialization in the principal-agent relationship may require improving the incentives for bureaucrats to act in the interests of government, so as to allow further delegation. Focusing on incentives facing the bureau-manager rather than the bureau cautioned us against accepting the suggestion of some writers that setting up competing bureaus provides an efficient method of reducing agency costs. It was also suggested that since detailed measurement of bureau output is just one of a number of monitoring mechanisms, and monitoring is itself only one of the possible solutions to the agency problem, output measurement is not a necessary prerequisite to further delegation.

The diversity of factors affecting the optimal degree of control over inputs led to the suggestion that investigation of a wider variety of bureau control regimes than is currently observed (largely limited to a split between-state owned enterprises and other bureaus) might be worthwhile.

One final caveat must be noted. It is accepted in this paper that the bureaucracy should do what the government wants it to and that it is appropriate to consider the bureaucracy as agent and the government as principal. It could be argued that part of the rationale for systems that allow for little formal intervention by the government--in appointing senior bureaucrats and in setting their salaries--is that the bureaucracy should provide some sort of “check and balance” on the government’s actions--in other words, that the bureaucracy has a valid objective other than to carry out government policies, broadly defined. In this case, the principal-agent model breaks down, as the problem is no longer to induce the bureaucracy, as agent, to act in the government’s interest. An example of where the principal-agent relationship is not appropriate is the judiciary, where in the United States for example, the lifetime tenure of Supreme Court justices and the inability of Congress to reduce the justices’ salaries can be seen as mechanisms to ensure that the Court is not induced to act in the interests of the government. Many countries have similar safeguards to ensure the independence of the judiciary from the government.