A risk can leave one person's account without leaving the world. Insurance makes that movement its business: a household pays a certain sum so that someone with a wider pool will meet an uncertain loss. The transfer can make both parties better off. Yet the same change in the location of a loss can also separate the person who chooses a danger from the person who ultimately bears it. The return remains visible beside the chooser's name; the cost appears somewhere else. Before calling that return the wage of courage, the volume has to follow both sides of the account.
The relevant distinction begins with the terms on which the risk moves. An insurer can knowingly accept a specified exposure, collect a price for it and arrange capital against the claims. A lender can accept the possibility of default in return for a promised payment and protections over the borrower's conduct. Neither bargain requires that the original owner retain every risk. Specialization in bearing can be useful precisely because different people have different capacities to withstand a loss. The question is whether the danger accepted is the danger subsequently imposed, and whether the price received covers the undertaking that has actually been made.
Conduct can change after the bargain. A person whose losses are partly paid by another may choose differently from one who bears the whole consequence. This is the incentive problem called moral hazard. Its force depends on what can be observed, what can be contracted for and what the decision maker still has at stake. Michael Jensen and William Meckling made one version particularly clear in their account of the firm: shareholders with a fixed debt ahead of them may prefer an investment that raises their own claim's value while reducing the value left to creditors. The gain to the chooser can come partly from changing somebody else's exposure.
Suppose an imagined company owes eighty and its owners can choose between two projects of equal cost. The first pays one hundred and ten for certain. Creditors receive eighty and shareholders receive thirty. The second pays one hundred and fifty or fifty with equal probability. In the favorable outcome, creditors still receive eighty and shareholders receive seventy. In the unfavorable one, creditors receive the available fifty and shareholders receive nothing; the owners have no further liability. The risky project gives shareholders an expected thirty-five, five more than the safer project. It gives creditors an expected sixty-five, fifteen less. Total expected receipts fall from one hundred and ten to one hundred. The shareholders' preference and the loss to the undertaking are perfectly compatible.
The source of that preference becomes clearer if the owners finance the whole project themselves. They then compare the safe one hundred and ten with the risky average of one hundred and choose the former under the example's risk-neutral rule. Debt has changed the division of outcomes and, through that division, the choice. This does not make a hazardous project inherently wasteful. If its added possibilities produce enough additional value, taking it may improve the undertaking. The defect in this example is specific: an inferior total return becomes attractive to the decision maker because another claimant bears more of the downside.
Creditors who foresee the choice can respond before lending. They may offer less for the debt, demand a covenant or require information that makes deviation harder. Jensen and Meckling emphasize this response: anticipated behavior enters the terms, so their argument does not depend on creditors being indefinitely fooled. But an accurately priced incentive problem can still be costly. A lower price for the debt can return the loss to the original owner without making the inferior project superior; monitoring and contractual restrictions can themselves consume resources or prevent useful changes. A sound arrangement has to compare these costs with the loss it prevents.
A guarantee can move the same loss one stage farther. Imagine that an outside guarantor promises to make these creditors whole for eighty, charges nothing and leaves shareholders' project choice unchanged. Under the risky project, it pays thirty in the bad outcome and nothing in the good one: an expected payment of fifteen. Creditors are now protected, while shareholders retain their expected thirty-five. The missing fifteen appears in the guarantor's account. A risk-sensitive fee of fifteen charged to the shareholders would reduce their expected net receipt to twenty, below the safe project's thirty. Under these assumptions the fee restores the preference for the better project. Real guarantees bring questions of observability, financing and credibility that the simple arithmetic has deliberately held fixed.
The guarantor need not be a stranger, and the guarantee need not be written where the exposure first appears. A bank can move a loan into a separate vehicle and still promise that vehicle funding when its usual lenders withdraw. In his account of the credit crisis of 2007–2008, Markus Brunnermeier describes precisely this return path: liquidity backstops and holdings of structured claims left substantial exposures within the banking system. Selling a security changed the immediate owner of a claim. Establishing who remained at risk required reading the promises around the sale as well as the sale itself.
Some consequences also reach people outside the original bargain. A firm selling assets under pressure can depress prices used as collateral by others, tightening their borrowing constraints and inducing further sales. The resulting loss is more than a rearrangement of the first firm's contractual claims when it disrupts otherwise useful activity. Workers, suppliers and customers may encounter the consequence through a lost job, an interrupted payment or a service that disappears. Establishing such a cost requires the actual connection; the mere fact that an owner earned a return and somebody else later suffered does not identify one. But where the connection exists, an account confined to the owner's stake omits part of the risk the decision created.
Rescue brings the distinction between a present choice and an earlier incentive into view. Once a failure threatens a wider loss, support may prevent damage that letting the institution collapse would cause. The fact that support is useful at that moment does not determine who should have paid for the exposure beforehand or which claimants should retain value afterward. Conversely, preserving incentives does not require treating every avoidable disruption as a desirable punishment. These questions concern different margins: preventing a loss already in motion, allocating its burden, and changing the decisions that will generate the next exposure. They have to be considered together without pretending they are the same act.
The account of profit in the preceding plate already contained this difficulty. Knight recognized that contractual incomes can carry imperfect guarantees and that responsibility can be distributed among managers, financiers and the providers of productive services. A business can therefore offer real protection to others while leaving them with some uncertainty; it can also obtain protection for its owners while passing danger outward. Neither arrangement can be judged from the label on the income. The comparison must trace the enforceable commitments, the freedom to alter them, the resources available when they fall due and the consequences for people who supplied no consent at all.
A return can reward the bearing of risk only to the extent that the risk has actually been borne.
The bearer is thus a position to establish, not an honor bestowed by the size of a reward. A well-designed transfer can put danger where it is best managed and make an undertaking possible. An unpriced transfer can make the undertaking look safer to its chooser precisely by exposing someone else. The premium acquires its meaning from that difference. The next plate asks what happens when the price of bearing is itself encouraged by the calm it helps to sustain, and when a market's confidence grows faster than its capacity to meet the promises confidence has produced. The argument remains open.