The bottom line, however, is that the agency problem can never be 100% solved in a world where virtually everyone has a healthy regard for their own self-interest and the relevant information for evaluating performance is imperfect, costly to obtain and unequally distributed between the agent and his principals. Indeed, rational principals will only pursue the available techniques for control to the point that the marginal increment in "agency costs" rise to equal the marginal benefits to them of the additional increment in "faithfulness" that they produce. (That is to say, sometimes it is cheaper for principals to endure a certain amount of dereliction of duty by their agents than it is to pay for the precautions needed to prevent or punish it.) In some kinds of institutions -- especially those where results are not readily measurable with much precision, those where the nature of the agents' work is such as to require a very high degree of expert judgment, those where lines of responsibility and authority are very complex, those where agents work individually in widely dispersed work places, those where the agent's activities necessarily involve a lot of "judgment calls" to cope with rapidly changing circumstances and highly uncertain information, and those where large numbers of principals have only relatively small individual "stakes" at risk -- the incentives for agents faithfully to represent their principals may easily become so weak as to be largely ineffective. Experience demonstrates that these kinds of organizations often come to be run mainly for the benefit of the agents (managers and other employees, service professionals, politicians, officials) rather than their purported principals (stockholders, voters, taxpayers, clients, etc.). Two of the important tasks of the academic disciplines of business administration and public administration are to identify, and then to devise cheaper substitutes or remedies for, organizational arrangements that are characterized by costly agency problems.
[See also: incentive , marginal analysis , bureaucracy , bureaucratic politics ]