At its simplest, principal-agent theory examines situtations in which there are two main actors, a principal who is usually the owner of an asset, and the agent who makes decisions which affect the value of that asset, on behalf of the principal. As applied to the firm, the theory often identifies the owner of the firm as principal, and the manager as agent, but the principal could also be a manager, and an employee nominated by the manager to represent him in some aspect of the business could be the agent.
In this case the asset, which the agent’s decisions could enhance or diminish, is the manager’s reputation. To explain the relationship between principal-agent (or agency) theory, and other theories of the firm, we turn to Williamson’s (1985, pp. 23-29) categorisation of approaches in IO in terms of their views on contracts. There are two main such approaches or branches: monopoly, which views contracts as a means of obtaining or increasing monopoly power, and efficiency, which views contracts as a means of economising.
The early work on SCP and particularly on barriers to entry, for example, belong on the monopoly branch of contracts. Both transaction cost and principal-agent theories belong on the efficiency branch (together with most of what Williamson calls the New Institutional Economics). Thus, in Williamson’s perspective, agency theory is the theory that focuses on the design and improvement of contracts between principals and agents. Among the major concerns of principal-agent theory is the relationship between ownership and control2, and in this respect it can be seen to have emerged from the managerial theory tradition.
Indeed, in that it focuses on the contractual aspects of that relationship, and often adopts game-theoretic methods, principal-agent theory can be seen as a new IO version of a sub-set of managerial theory. Recent work in this area tends to be highly theoretical3. Principal-agent theory sees the firm – as does neoclassical theory – as a legal entity with a production function, contracting with outsiders (including suppliers and customers) and insiders (including owners and managers).
There is information asymmetry between principals and agents, but, unlike in transaction cost theory (which usually assumes bounded rationality) there is often assumed to be unbounded rationality. We will discuss this in more detail below; in the context of the design of contracts between principals and agents, unbounded rationality refers to the ability of those designing the contract to take all possible, relevant, future events into consideration.
The principal may know various things not known to the agent (in relation, for example, to the prospects of the firm), and vice versa (the agent may have a lower commitment to the firm than he leads the principal to believe), but if the obligations of both under the contract can be specified, taking into consideration the possibilites arising from private information, then there is unbounded rationality despite the information asymmetry.
The agency theorists’ concerns – and in this they are different from neoclassical theorists – are with “owners’ and managers’ problems of coping with asymmetric information, measurement of performance, and incentives” (Chandler, 1992b). The major difference between principal-agent and transaction cost theories is that the former focuses on the contract, the latter on the transaction.
The problem for principal-agent theory is how to formulate a contract such that the shareholders (the principal) will have their interests advanced by the manager (the agent), despite the fact that the manager’s interests may diverge from those of the shareholders4. Where objectives of the agent are different from those of the principal, and the principal cannot easily tell to what extent the agent is acting self-interestedly in ways diverging from the principal’s interests, then the problem of moral hazard arises.
The problem originated in the insurance industry, referring to the possiblity that people with insurance will change their behaviour, resulting in larger claims on the insurance company than would have been made if they had continued to behave as they did before they had insurance. This change in behaviour may, moreover, be known to the insurer, but may not be fraudulent – or, at least, may not be provably fraudulent. In the context of relations between principals and agents, moral hazard refers to the possibility that, once there is a contract, the agent may behave differently from how he would have behaved had he not had the contract.
It must, in addition, be difficult to determine whether his behaviour has conformed to the terms of the contract5. This arises particularly where the agent is a member of a team. Principal-agent theorists have attempted, by specifying conditions such as that the manager’s salary be equal to the expected value of his marginal product, to design contracts on the basis of which there will be an incentive for the manager to act in the shareholders’ interests.
However, the importance of the team element in managerial jobs discredits the notion of a manager’s marginal product (Aoki, 1984, Ch. 2 and p. 50). This team element6 is also present at the production level. Doeringer and Piore (1971, p. 277) emphasised the importance of “social cohesion and group pressure” in the establishment of work customs. The process whereby such routines are created, and their importance in the success or otherwise of firms, are central concerns of the evolutionary theory of the firm (Section 2. 4).
Principal-agent theory is more concerned with implications for shirking, that is, a reduction in effort by an agent who is part of a team. There may be a slight decline in total output as a result, but the cause will usually be unidentifiable. The shirking manager knows that his diminished effort is unobservable. Shirking is the moral hazard arising from the employment contract. What the principal can do, in the formulation of contracts, to offset shirking (and other types of management misbehaviour), is a key problem of principal-agent theory.
There are a number of ways of controlling moral hazard. Rather than attempting to calculate the value of each manager’s marginal product, managers could each be paid a salary plus a bonus based on the performance of the company. The problem here is that if the utility of leisure is different for different managers, then again some may work more and others less at maximising the long-run value of the firm. (On the other hand, where there is a great deal of cultural homogeneity, as can be argued to be the case in Japan, this salary plus bonus system seems to be effective.
) Other examples of suggestions by principal-agent theorists for solving employment contract problems include the development of efficient ways of monitoring the performance of individual managers (or management teams), providing incentive contracts which reward agents only on the basis of results, bonding (where the agent makes a promise to pay the principal a sum of money if inappropriate behaviour by the agent is detected) and mandatory retirement payments. This last acts like a bond, in that there is a disincentive for the employee to misbehave because if he does misbehave he may be fired, and lose his retirement payment.
It should be emphasised that, to the extent that managers want to keep their jobs, the three markets (for corporate control, managerial labour and the firm’s products) can control moral hazard. In relation to the market for corporate control, for example, Many observers have interpreted the hostile takeovers [of the 1980s] as a corrective response to managerial moral hazard: The takeovers, it is claimed, were intended to displace entrenched managers who were pursuing their own interests at the expense of the stockholders (Milgrom and Roberts, 1992, p.
182). The fact that the acquisition share prices were higher than they had been in the market prior to takeover, may be evidence of management misbehaviour or moral hazard. This would be so if the original market value of the shares had been the equivalent of the company’s value (net present value of the future stream of profit that could reasonably be expected) under the original management, and the acquisition price was the company’s value under the new management.
It may, on the other hand, indicate an overestimation by the acquiring firm or individual of its/his capacity to improve the performance of the company. Milgrom and Roberts (1992, pp. 182-3) seem to conclude that the takeover premium was indicative of moral hazard when arguing that there is other evidence of management misbehaviour in the adoption during the 1980s by management of the poison pill defence against takeovers.
The poison pill is a special security, which gives the holder the right to acquire shares at very low prices in the event of a hostile takeover. Poison pills were created by management, in some cases without shareholder approval. If, as Stiglitz (1991) suggests, the acquiring firm in takeovers generally experiences no increase in its own share values, then it is more likely that there has in fact been an overestimation by the acquiring firm of its ability to improve the performance of the target company.
This is indicative, in other words, of an overestimation of the moral hazard of the managerial employment contract. The most obvious solution to the problem of conflict of interest between principal and agent is for the principal to become his own agent. Where there is team production, and the existence of a monitor can reduce shirking by enough to pay his own salary, then it may be appropriate for that monitor also to be the owner of the firm.
If he is not the owner, then there could be a need to monitor the monitor, to ensure that he does not shirk. This leads to the conclusion that the existence of firms in which there is an owner and a group of people working as a team for that owner, is a consequence of the need to monitor team production, and the need for the monitor himself to be the owner – with, for example, the power to fire shirkers, to pay each of the members of the team in accordance with his view of their productivity, to keep the residual and to sell the firm8.
We return to the question of the basis for the existence of firms in the next section, where transaction cost theory, among other things, takes exception to principal-agent theory’s conclusion about the significance of the need for monitoring. 1 This section draws in part on Milgrom and Roberts, 1992, Ch. 6. The reader is encouraged to read that chapter for more details, particularly on the relationship between moral hazard and performance incentives. For game theoretic perspectives on the relationship between principals and agents, see Gardner, 1995, Ch.
10. 2 See, for example, the article on the “separation of ownership and control” by Fama and Jensen, 1983. Other important articles on principal-agent theory include Mirrlees, 1976 and Fama, 1980. 3 See, for example, Maskin and Tirole, 1992, who analyse as a three-stage game the relationship between the principal and agent in which the principal has private information that directly affects the agent’s payoff. 4 or, to express it in terms closer to those of the theorists in this area, the problem is whether there exists any class of reward schedule for the agent (the manager) such as to yield a Pareto-efficient solution for any pair of utility functions both for the agent and the principal (Aoki, 1984, p. 49).
5 An example of moral hazard in employment contracts arises in universities, where there are two different groups of employees, those on short-term contracts, and those with tenure.
Tenure is supported by many, and not only those who have tenure(!), as a feature of the independence of the academic, and the need to protect the academic against political pressure. Tenure may perform this function to some extent but it also enables those who have it, to change their behaviour and shirk various duties. The academic on short term contract, it can be argued, works hard, prepares excellent lectures, volunteers for administrative duties, does above average research and publishing. Then he obtains tenure, relaxes more, gives last year’s lectures, avoids administration, and does less research and publishing.
In practice there is, no doubt, moral hazard in tenure, but given that the best teachers, administrators and researchers in academia have tenure, academics certainly do not always, or even usually, change their behaviour in the way predicted by moral hazard. 6 raised, as we saw in Chapter 2, by Alchian and Demsetz, 1972. 7 quoted in Aoki, 1984, p. 26. 8 For a more detailed discussion on the issue of team production and the monitor as owner, see Holmstrom and Tirole, 1989, or at a more introductory level, Douma and Schreuder, 1992, Chapter 6.