Plate vii in print

The Warranty

warranties, brands, and reputation; the costly hostage a seller posts against his own cheating.

A promise gains weight from what its maker stands to lose.

— the argument of this plate, after Grossman and Klein–Leffler (1981)

The seller's assurance acquires weight when failure can cost him something. A buyer standing before a sealed machine cannot inspect the workmanship inside, but he can ask who will pay if it breaks. A warranty makes a promise about that later event. A familiar name offers a different assurance: the seller expects to trade again under an identity that dissatisfied customers can recognize. Each arrangement connects today's sale to a consequence beyond the receipt of its price. The connection gives the buyer a reason to trust information he cannot fully verify before paying. The details determine what has actually been pledged. A contractual payment, an asset held as security, and the value of future custom place different things at risk, with different means of making the loss occur.

A warranty can move uncertainty to the party better placed to know its source. Suppose two machines have probabilities of covered failure of one-tenth and four-tenths. A promise to pay one hundred if failure occurs has expected claims costs of ten and forty respectively. Assume identical claim expenses, verifiable failures, and a solvent seller who honors the promise. The same words then impose different expected costs on different qualities. This helps explain why coverage can support an assurance about workmanship. The calculation alone does not establish that buyers can identify quality from the length of a guarantee. Prices, outside opportunities, other terms, and buyers' beliefs also enter that inference. A seller may charge for bearing a large expected liability; a buyer may purchase the protection without learning the exact probability of failure. The guarantee changes the consequences of uncertainty even when it leaves some of the uncertainty itself intact.1

1. The probabilities and payment are illustrative: expected claims equal the failure probability multiplied by one hundred. Sanford J. Grossman, “The Informational Role of Warranties and Private Disclosure about Product Quality” (1981), §3, pp. 470–477, models a warranty contingent on an observable outcome when quality itself is costly to communicate or verify.

Insurance and disclosure can therefore accomplish different tasks. In Grossman's model, a full warranty equalizes a risk-averse buyer's payoff in the good and bad states. Under the model's assumptions, this can make the buyer's decision independent of the seller's undisclosed quality. Protection has removed the financial consequence of the unknown probability, without necessarily revealing it. The result depends on the benefit being compensable, the covered event being observable, and the promised payment arriving. The model initially holds quality fixed and rules out actions by the buyer that change the failure probability. If careful use affects breakdown, complete coverage can weaken that care; a deductible or an exclusion may address an incentive that full insurance would otherwise disturb. A short warranty accordingly needs to be interpreted with the product, its use, and its price. Coverage measures a promised allocation of loss, with quality inference depending on the surrounding arrangement.2

2. Grossman (1981), §3, pp. 471–477, and §4, pp. 477–480. The full-insurance result assumes risk-averse buyers, diversifiable seller payouts, known buyer benefits and an observable bad state. His §4 discusses consumer moral hazard and heterogeneous buyers as qualifications to extending the result. This plate makes no claim that every real loss can be fully compensated.
⁂

A reputation puts the seller's continuing income behind his performance. Suppose honest dealing earns ten each period, while cheating adds forty to today's profit. A breach becomes known after the current transaction and permanently ends the seller's business; he cannot return under another name. With an indefinite future and a discount factor of nine-tenths, honest dealing is worth one hundred including today. Cheating yields fifty and ends the stream. Honesty is preferable because the ninety of future profit it preserves exceeds the forty available from betrayal. At a discount factor of one-half, the honest stream is worth only twenty, and the same temptation wins. These are conditions on an incentive, rather than declarations about a seller's character. Benjamin Klein and Keith Leffler made the comparison central to their account of repeat purchases: losing future business disciplines a firm only when the business lost is valuable enough.3

3. Klein and Leffler, “The Role of Market Forces in Assuring Contractual Performance” (1981), §§I–II, pp. 615–625. In this plate's simplified example, honest value is 10/(1 − δ) and cheating value is 50. Honesty is weakly preferred when 10δ/(1 − δ) ≥ 40, hence δ ≥ ⅘. At the boundary both choices yield fifty. Outside earnings are normalized to zero; the action, detection and termination assumptions are stipulated.

The brand makes this future stake attach to an identifiable seller. A name carried from one sale to the next allows a buyer's experience to affect the purchases that follow. In Klein and Leffler's analysis, the relevant income is a continuing margin above costs that can be recovered or avoided when trade ends. Entry and investment can absorb the value of that margin while leaving it available to discipline later conduct. A firm may thus earn a normal return over its whole history and still have valuable future business to lose today. Money already spent on a sign or an advertisement is sunk. Its expenditure disciplines no later choice merely because it was expensive; the incentive comes from the future income that a breach would destroy. Visible investment can sometimes help buyers infer the size of that stake, provided the inference about future trade is sound.4

4. Klein and Leffler (1981), §III.A–C, pp. 625–633, especially pp. 626–627 on brand-name capital and pp. 629–631 on consumer cost uncertainty. They explicitly distinguish sunk expenditure from the future quasi-rents relevant to the seller's decision. Their advertising argument depends on an inference about those rents, rather than establishing that any conspicuous expenditure proves quality.

The hostage loses force when the loss cannot be imposed. A defect that appears long after customers have dispersed may produce little identifiable damage to a seller's trade. An easily replaced name weakens the connection between yesterday's sale and tomorrow's income. A business approaching a known last transaction has less future custom to protect. The example would offer no reputational restraint if the seller had no future profit to lose, however excellent his earlier record. An actual bond or escrow can instead place an asset within reach of an agreed enforcement process, though that process must still identify a covered failure. A warranty backed by neither recoverable assets nor a valuable continuing relationship may be difficult to collect. These arrangements can reinforce one another: the name encourages the seller to honor the written promise, while the promise specifies what performance the name stands behind.5

5. Klein and Leffler (1981), pp. 616–619, pp. 621–625 including notes 8 and 11, and §IV, pp. 633–637. Their model states the identity, communication and repeat-purchase assumptions; the conclusion compares private enforcement with explicit contractual enforcement and other institutions. The cash-bond comparison here distinguishes an asset recoverable on a specified event from a loss of anticipated custom.

The market builds trust by giving the knowledgeable party something to lose when his assurance fails. That loss may arise from a promised repair, a payment, or the disappearance of profitable relationships. Each device has a cost, and each leaves a question about who can recognize the failure and make its consequences follow. A brand can carry accumulated evidence and a continuing incentive together; neither feature makes its claim beyond examination. The buyer is judging an institution as well as a product. In The Signal, the inquiry returns to the price that carries knowledge across the market, and to the institutions whose cost helps make that knowledge usable.