The proprietor can close the account book and turn toward the workshop whose decisions made its numbers. The distant shareholder receives the numbers after those decisions have been made. Between the two stands a manager who knows more about the work, chooses much of its direction, and receives only part of the gain or loss that follows. Delegation puts useful judgment where the work can be understood. It also separates the person making a choice from some of its consequences. The firm may have acquired the intelligence needed to act while losing the simple alignment of interest that came from an owner deciding over her own purse.
The separation creates a problem of incentives before it creates an accusation of misconduct. The manager may value a familiar project, a larger department, fewer difficult negotiations or the security of an undemanding routine. Some of these preferences can be accommodated cheaply; some can even improve production. Others cost the owners more than the manager is willing to count. The issue is the difference between the consequences she experiences and the consequences of the choice as a whole. Jensen and Meckling make that difference central to their account of outside ownership. A person who retains only a fraction of the financial claim bears only that fraction of an additional expense charged to the firm, even when she receives the whole personal benefit it buys.
A small private benefit can consequently outweigh a larger cost that falls mostly elsewhere. In a constructed firm worth one hundred before the disputed choice, a perk costs ten and gives the manager a personal benefit she values at three. Assume that it has no productive effect, that she can choose it without being penalized, and that nothing else changes. As sole owner she declines it: three is worth less than the ten she would surrender. Retaining one fifth of the shares, she bears only two of the ten and takes it. The firm’s financial value becomes ninety; her retained claim is eighteen instead of twenty, while she receives the benefit worth three. The choice turns on the marginal cost she bears, not on a supposition that owning a small share has made her a different kind of person.
The cost can reach the founder before the outside shareholders ever suffer a surprise. If buyers foresee that choice, they will pay seventy-two for four fifths of a firm they expect to be worth ninety, rather than eighty for four fifths of the imagined hundred. The founder then has seventy-two in sale proceeds and eighteen in retained equity: ninety in financial wealth, with the personal benefit worth three beside it. Anticipated discretion has lowered what she can raise. Unexpected diversion after shares have been sold can impose a loss on buyers; a cost incorporated into the original sale price has a different incidence. Jensen and Meckling’s argument gives a founder a reason to accept restraints even before investors demand them: a credible restraint can make the claim she sells more valuable.
Control has to be counted among the costs of obtaining a faithful agent. An audit consumes resources, a reporting requirement consumes time, and a restriction can prevent a valuable decision as well as an abusive one. Jensen and Meckling distinguish what principals spend to monitor, what agents spend to bind their own conduct, and the remaining loss from decisions that still diverge. Their agency cost is the sum of these three. The distinction prevents the comforting mistake of comparing an imperfectly supervised manager with perfect supervision supplied for nothing. It also prevents the opposite mistake of treating every expenditure on oversight as a successful reduction in the problem. A control deserves its place because of what it changes after its own burdens are counted.
The cheapest-looking arrangement can be the expensive one. Suppose a feasible control in the constructed firm costs four and prevents the ten-unit perk. The firm is now worth ninety-six after the control, and the manager loses the private benefit. Under the same anticipated sale, the founder’s financial wealth is ninety-six, against ninety plus a benefit valued at three without the control. She can prefer the restraint. A control costing eleven would leave only eighty-nine and would fail the comparison. The numbers put a limit on the common instruction to monitor more closely: better observation is useful when the loss it prevents exceeds the cost it introduces, with the parties’ actual valuations included. Spending until nothing escapes can cost more than the conduct that escapes.
The right to a return does not itself provide the knowledge or the means to govern how it is made.
Dispersed claims can make valuable vigilance difficult to supply. Suppose an investigation would increase the value of a company by a thousand and cost ten. A holder of one thousandth gains only one if she pays for it alone. A holder of one tenth gains a hundred from the same improvement. Both own claims on the same company; their private incentives to bear the cost differ. The illustration leaves open whether a board, a concentrated investor, a contractual reporting system or some other arrangement can organize the work more economically. Its point is narrower: the collective value of information does not ensure that a particular dispersed owner will purchase it. Governance must arrange for someone to undertake a task whose benefits otherwise spread beyond the person paying for it.
Giving the manager shares changes one margin without answering every question about performance. In the perk example, increasing her fraction makes the expense more costly to her. But a financial outcome can also depend on demand, other workers’ choices and events she could not control. A measure that rewards the outcome therefore needs to be judged by how it changes conduct, what information it conveys and what burdens accompany the reward. Monitoring can likewise miss the useful work it means to encourage or suppress initiative by restricting too much. Jensen and Meckling locate the scale of the problem partly in the cost of measuring performance, devising compensation and replacing a manager. An incentive is a designed relationship between action and consequence; its name on a compensation schedule cannot certify that relationship.
The principal’s purpose must be identified before a departure from it can be priced. An agreed expenditure on workers, a community obligation or a project with a long horizon is not shown to be an agency loss merely by lowering this year’s distribution. One must know the undertaking the manager was entrusted to carry out and distinguish an authorized choice from the substitution of her own preference. The same discipline applies when the claimant is a cooperative’s membership or the constituency of a nonprofit organization. Delegation poses a question about fidelity to a defined purpose. It does not, by itself, choose the purpose or prove that every person affected by the firm has been adequately heard.
The firm is run through the interests its arrangements make effective. Shareholders, managers, creditors and workers can each influence its choices, with different information and different powers to object. The presence of agency costs does not prove that delegation was a mistake: the manager’s expertise and the capital raised from others may make possible an enterprise that concentrated ownership could never have financed or operated. The comparison is with arrangements that could actually replace it. Yet the gain from delegation leaves the original question alive. Who can decide when the promise gives no clear answer, and what can others do if they disagree? The next plate moves from the manager’s reward to the assets over which that unfinished authority is exercised.