The internal organization of government.

1. Introduction

ONE OF the accomplishments of economic theory has been the development of a theory of organizations. Three paradigms, adverse selection, moral hazard, and incomplete contracting, have been used to analyze how workers, managers, directors or investors respond to various incentives. Self-interested economic agents can be motivated in roughtly three ways. Formal incentives such as piece rate wages, bonuses, stock options and relative performance evaluation are based on verifiable measures of performance. Work inputs are monitored by foremen, fellow employees, bosses or boards of directors. Last, career concerns inside and outside the firm may encourage a forward-looking employee to work hard.

So far, incentive theory has been mainly motivated by and applied to private organizations. Yet, in view of the important role played by civil servants and politicians in our economies, one may wonder why limited attention has been devoted in this field to the internal organization of government. An answer to this question might be that there is little conceptual difference between governments and firms. Any distinction would be quantitative and left to empirical analysis. While this point of view has some appeal, there still seems to be some scope for a separate theoretical appraisal of the organization of government. The purpose of this paper is not to supply such an appraisal. Rather, its goal is to suggest some of the building blocks and some directions for research. I apologize to the reader for the lightness of the analysis, and just hope that this paper's only ambition, namely to encourage interest in the topic, will be fulfilled.

The general thrust of the paper is that the new methodology of incentive theory ought to enable economists to participate in and enrich a debate that has by and large been confined to other social sciences, in particular political science and sociology.

The first part of the paper (Section 2) discusses some specificities of the design of incentives in the public sector. While private enterprises are in a first approximation instructed to maximize profits, government agencies generally pursue multiple goals. Many of these goals are hard to measure. Furthermore, incentives based on measurable goals must be limited to not completely jeopardize the nonmeasurable dimensions of social welfare. Lack of comparison and heterogeneity of tastes of principals are identified as further factors leading to low powered incentives.

The second part of the paper discusses some implications of low powered individual incentives in government. First, career concerns, associated with the prospect of re-election, promotion or employment in the private sector, are at least as pervasive as in the private sector (Section 3). Career concerns are articulated around some mission that is followed by the government official. The mission can be simple--pursue goal 1--or composite--achieve a balance between goals 1 and 2. Neither the choice of the mission nor the intensity with which it is pursued need be socially optimal. Indeed, there is a potential multiplicity of missions that can be followed by rational officials. Also, several missions can be pursued by different officials of the same agency. Last, composite missions that reflect the several goals of social optimization may not fit with the officials' self interest. Our economic analysis here complements recent work in political science on government agencies.

Another topic that is particularly relevant under low powered incentives is the issue of regulatory capture and collusion (Section 4). It is argued that viewing intermediate layers of a hierarchy (such as government agencies) as being better informed than their principals lays the foundations for a theory of regulatory capture. The officials can manipulate information to favor specific interest groups. The civil service and the regulatory structure are then partly designed to limit such manipulations. The paper discusses a few implications of this view, concerning the stakes of the interest groups, the determinants of the influence of an interest group and the design of institutions.

While the second part of the paper analyzes individual incentives, the third part studies the division of labor within government. Section 5 points out that legal restrictions on commitment by government agencies can be viewed as a division of control rights between successive administrations. Balancing their well-known limitations, short-term commitments by the government have the benefit of allowing correction of wrongful policies (possibly due to capture of the current administration) by future administrations. Section 6 discusses a few elements of the division of labor between government and the private sector in the context of privatization.

Sections 7 and 8 investigate the following puzzle: why isn't government designed to behave as a coherent entity? Government agencies as well as politicians are not expected (individually) to maximize social welfare, but rather to pursue antinomic missions. Section 7 argues that the control of economic agents such as a public enterprise may be best performed by creating multiple principals with dissonant objectives. For instance, public enterprises are often subject to two masters with substantially different goals: A 'spending ministry' represents the 'technical point of view' and behaves rather softly with regard to the firm. When the firm runs a large deficit, this ministry must relinquish control to a more rigorous ministry of finance that is primarily concerned with the budget deficit. The basic idea of the section is that this division of labor within government promotes better behavior by the public enterprise through the threat of a shift of control to a tough ministry in case of financial hardship.

Section 8 arrives at a similar conclusion on the optimality of a division of labor in government from a quite distinct perspective. Its premise is that competition in government among advocates of specific interests or causes may give rise to good policy setting. Using enfranchised advocates generates precious information on the pros and cons of alternative policies, and creates a system of checks and balances. The idea can be applied to justify the existence and behavior of specialized ministries, biased representatives, multipartism or our democratic legal system.

2. Specificities of incentives in government

Why do the incentives of a high official in a foreign ministry differ from those of a top executive at IBM? What distinguishes the task of a correctional officer from that of an AT & T sales representative? Such questions may seem trivial or irrelevant. Yet they condition much of what we perceive as a good organization of government. If differences exist between the public and private sectors, they must be traced either to differences in the measurement system (points 2.1 and 2.2 below) or to differences in the governance structure (points 2.3 and 2.4). Before proceeding, it is important to stress that the differences are differences in degree, not fundamental nature.(1)

2.1. Multiplicity of goals

The owners of a private corporation set the goal of 'maximizing profit' for the organization.(2) Some measurable variables, such as earnings or stock prices, are clearly related to this goal and can be used to build managerial incentive schemes. In contrast, the mandate of many government agencies is multidimensional. Indeed, the very intervention of government is often motivated by the idea that profit incentives by themselves would not yield socially optimal allocations. Other criteria such as consumer net surplus, pollution, development, or redistribution must also be taken into account.

The multidimensionality of goals per se does not hinder the construction of powerful incentive schemes. Such schemes can in principle carefully balance the use of measures of the various dimensions of performance. A clearly specified social welfare function with explicit weights on all dimensions of performance would be as implementable as profit. But the multidimensionality of goals often goes hand in hand with two difficulties.

First, several dimensions of performance are, unlike profit or cost, hard to measure. A regulator of a natural monopoly is supposed to ensure 'reasonable' prices, but even an econometrician may have a hard time measuring the regulator's contribution to the net consumer surplus. And who will put reliable numbers on the US Department of State's performance in 'promoting the long-range security and well-being of the United States' and on the US Department of Labor's success in 'fostering, promoting, and developing the welfare of the wage earners of the United States'?

Second, and relatedly, the multiplicity of goals raises the issue of their weights. The Environmental Protection Agency is instructed to curb pollution at a reasonable cost for the industries. Suppose, perhaps heroically, that the levels of pollution and the costs imposed on the industries are measurable. Setting up a formal incentive scheme for the EPA requires putting weights on these two measures. Yet, it is difficult to define what is reasonable and what is not. 'Optimal' pollution levels depend on available technologies, on the shadow cost of unemployment, on atmospheric conditions and so forth. The very contingencies that are supposed to condition the formal incentive schemes are hard to include in an incentive scheme.(3) It should also be noted, and we will come back to this point, that what is meant to be 'optimal' depends on what the EPA perceives to be its constituency.

2.2. Lack of comparison

A noisy observation of managerial performance reduces the efficacy of formal incentive schemes. One way of alleviating the imperfection of measurement of a manager's performance is to separate idiosyncratic risk from aggregate risk, that is the risk faced by the manager only from that faced by other managers in a similar situation. More prosaically, the performance of GM's managers ought to be compared to that of Ford's or more generally to that of the car industry before drawing conclusions on their efficiency. The feasibility of such 'yardstick competition' or 'relative performance evaluation' enhances the desirability and the strength of performance related incentives. In contrast, as Hicks (1935) pointed out, 'the best of all monopoly profits is a quiet life'. A modern version of this would be that the absence of yardstick is conducive to low-powered incentive schemes, where a low-powered scheme is one in which the agent bears only a small fraction of his performance.

That many government agencies have a monopoly position in their activity therefore suggests that their performance is hard to assess. True, elements of relative performance evaluation can be found at several levels of government. First, the performance of employees in an agency, for instance tax collectors with similar tasks, can be compared. But, at a higher level, the activity of the IRS as a whole can only be compared with that of its counterparts in foreign countries. Second, some government institutions such as hospitals or schools may face competition from the private sector. Third, there may be explicit competition among government organizations, as is the case among cities or states, or among agencies to gather intelligence or to catch drug dealers.

While this paper will discuss some aspects of competition in government, it will ignore some of the central issues in this regard.(4) The next two distinctions relate to qualitative differences in 'corporate governance' (the role of the organization's outsiders).

2.3. Heterogeneity of owners' tastes

A corporation's ownership in principle aims at maximizing total firm value. This goal is shared among investors and is stable over time. True, managerial incentives among other things require creating several constituencies, such as equityholders and debtholders, with somewhat conflicting goals. But the corporation issues heterogeneous securities in a controlled way. Contrast this with government agencies. The tastes of their principals, namely the people, are quite diverse and furthermore changing. While a corporation's goal is well defined(5) and time consistent and preference heterogeneity among claimholders is a deliberate construction, the goals of an agency are defined by a political process. And, because this 'aggregate goal' (if such a thing exists) changes over time in a noncontractible manner, incentives governing long-term choices by agency management that are deemed legitimate today may no longer be considered so tomorrow. This lack of time consistency of agencies' objective functions suggests that commitment possibilities in the public sector will be more limited than in the private sector. (Section 5 will study another reason why commitment is limited in government.)

2.4. Dispersed ownership

Corporations often face dispersed shareholders and creditors. Agencies are in this situation with a vengeance. Big shareholders, bank debt, and boards of directors, which alleviate the representation problem in corporations, have imperfect counterparts in government. Political parties and interest groups do coordinate subgroups of voters, but their incentives need not be perfectly aligned with the preferences of their constituents.(6) Agencies, like corporations, have their own boards of directors (e.g., congressional oversight committees), but the boards' incentives are different. Last, political takeovers also differ from private takeovers. Two limitations of political takeovers are, first, that they are a somewhat cumbersome way to …