Assessing the assumptions: a critical analysis of agency theory.

Principal-agent (agency) theory dominates the bureaucratic politics literature. Yet there has been very little effort devoted to assessing the assumptions of agency theory since the model was imported from economics. This article examines five major assumptions underlying agency theory. The authors suggest that the effort to translate and apply assumptions from economics to the study of bureaucratic politics misses much that is important They offer modifications to agency theory and a new direction for research.

In many ways, the work currently being conducted on the political control of bureaucracy is impressive. Deducing from a set of core assumptions and applying principal-agent (agency) models, analysts have arrived at positive conclusions concerning the efficacy of political control. Many of the empirical analyses have been quite sophisticated, adopting highly refined statistical tools to examine the "outputs" of a number of agencies. Much effort is devoted to modeling legislators' utility functions, the multiple dimensions of legislative choice, presidential strategies for control, and the like. Unfortunately, little attention has been devoted to theory development. The lack of attention is of more than passing concern. Without an accurate model and sound theory, the results of the most careful analysis may lack theoretical significance and empirical applicability. As Hanushek and Jackson (1977) caution,

If the model is doubtful prior to seeing the data, it is difficult to

see what purpose there is to the analysis. Since we cannot

"prove" explanations with observed data, simply showing that

behavior is consistent with a weak model is not useful, and

showing that observations refute a weak model is even more

useless. (p. 6)

Before initiating additional studies, it is useful to determine whether the principal-agent model is of sufficient strength to warrant further attention. If the current variant of the agency model is lacking, as we believe it is, then it is necessary to address the theoretical weaknesses and propose needed remedies.

This article proceeds in several stages. First, it discusses delegation and the resulting problems of political control--the central concern of agency theory. Second, it examines the core assumptions of the agency model used in political science. Finally, it seeks to replace some of the weaker assumptions of the model with more tenable propositions.

DELEGATION AND BUREAUCRATIC DISCRETION

Public policies commonly find their origins in legislation that defines objectives in broad terms and delegates the authority to implement them. Bureaucrats then must give ambiguous terms such as "the public interest" substantive content; they must establish rules to translate vague mandates into enforcement actions. Because legislative mandates can be extraordinarily vague, the distance between legislation and public policy can be great. Legislation, executive pronouncements, and orders can establish policy goals. They rarely dictate the series of actions that constitutes policy. This is left to be determined by administrators who become, by default, policymakers. The result is the classic problem of "agency" or unstructured delegation (Mitnick, 1984; Moe, 1984). Delegation then seems unavoidable. It is the product of politics, technical complexity, and the disjunction of demands and capacities.

The result of vague legislative mandates and broad grants of discretionary authority is a seemingly permanent state of tension in the American polity. When concerns over discretion become great enough, elected officials (principals) can choose to act to reign in the bureaucracy (agents). The question then becomes which actor or actors assume the role of principal. As pointed out by Hammond and Knott (1992, p. 1), presidents and Congress have fought for control of the bureaucracy since the inception of the republic. The history of this struggle is, in no small part, a significant factor in the development of American bureaucracy (see Nelson, 1982; West, 1995).

A central concern of all models of bureaucracy, then, is control. Agency theory offers a particular variant of the "delegation problem" premised on several key assumptions with regard to the motivation of actors, the role of institutions, and the nature of the political system. We now turn to a brief examination of the core assumptions of the agency model of political control.

ASSESSING THE ASSUMPTIONS

In this section, we identify and assess the core assumptions of the principal-agent model to determine whether it remains useful once the assumptions are relaxed to achieve greater verisimilitude. Agency theory is one of the many variants usually lumped together under the theory of rational action (see Monroe, 1991). It is not our intention to critique the entire field of rational action or the variety of subfields concerned with public choice, social choice, and the like. Rather, we restrict our analysis to principal-agent models.

Proponents of the agency model arrive at their conclusions by logically deducing from a number of simplifying assumptions. Such assumptions are integral to the process of deductive modeling. One often must sacrifice descriptive accuracy if one is to model a complex reality and identify critical relationships. Thus economists commonly assume that individuals are rational actors with complete information seeking to maximize wealth within free markets. Although these assumptions may not be descriptively accurate, they may suffice as devices of model building.

After a model is constructed, one must progressively modify core assumptions to arrive at an explanatory model that bears a relatively close relationship to the real world. The descriptive accuracy of the original assumptions becomes a matter of concern if, after they are relaxed to accommodate the empirical reality, the key relationships and conclusions derived from the model no longer hold. When this occurs, it is necessary to adopt new assumptions that stand a better chance of surviving this confrontation. Absent these refinements, one may remain wedded to a model of little explanatory value. Without sound theory, valid interpretation is impossible. With this in mind, in this section we consider the underlying assumptions of the principal-agent model as employed in political science in some detail. Before embarking on this journey, however, we offer the following snapshot of the principal-agent model for the uninitiated. Advocates of the model can benefit from the discussion in that it explicitly identifies underlying premises that often are merely assumed.

AN OVERVIEW OF PRINCIPAL-AGENT THEORY

According to agency theorists, a political system is composed of innumerable principal-agent linkages, both within organizations and between organizations. The most critical principal-agent relationships exist between elective institutions and administrative agencies. Congress delegates authority on the presumption that the administrative agency will exercise it faithfully. Similarly, a chief executive chooses agency heads and issues policy mandates with the expectation that the executive aim will be followed. Although delegation is necessary and commonplace, it creates great uncertainty.

The same asymmetry of knowledge and information that often necessitates delegation subsequently limits the capacity of principals to monitor and assess the behavior of their agents. The principals, as a result, are at all times vulnerable to distortion, shirking, and opportunistic behavior. The problems of control are real and find a parallel in most complex organizations. An asymmetrical distribution of information and knowledge characterizes any complex organization based on specialization and hierarchical modes of governance (see Barzel, 1989; Eggertsson, 1990).

Elected officials may find it difficult to ensure that administrators faithfully pursue established goals during implementation. This is of more than practical concern. Administrative autonomy can detract from representation and democratic accountability if administrators seek to exercise their discretion contrary to the wishes of elected officials. Many scholars suggest that democratic control prevails despite the factors just mentioned. Because bureaucrats seek to maximize their budgets and preserve their mandates, they are highly responsive to political demands and are sensitive to institutional forms created to enforce and convey these demands. Indeed, some research suggests that enforcement actions will automatically change to mirror changes in the preferences …