A stranger's promise becomes easier to buy when a familiar name stands beside it. The laboratory has tested the component; the inspector has examined the building; the auditor has considered the accounts. The buyer cannot repeat every inquiry, but he can recognize the signature of someone whose business is to inquire. A market threatened by the gap in knowledge builds witnesses. That was the hopeful answer to the lemon: the parties need not abandon a valuable exchange merely because one cannot see what the other knows. Yet the witness introduces a further dependence. Someone must decide whether the examination was competent, whether its result was faithfully reported, and whether the mark on the product means what the purchaser believes. The seal carries the question of trust one institution farther.
The seller's own name supplies the simplest version of this machinery. A business that expects to return tomorrow holds something it can lose today: the stream of future custom attached to its reputation. The brand is a hostage the seller gives to his own future good behavior. But a hostage disciplines only when it can be seized. The defect must become visible, the customer must connect it to the responsible firm, and future buyers must learn enough to change their conduct. A sound product can fail through misuse; a poor one can survive long enough for its maker to disappear. Reputation works through the prospect of a consequential judgment, not through the mere fact that a name has been printed on a box.
The arithmetic shows why a valuable name may still be an inadequate bond. Imagine a certifier expecting twenty units of net earnings each year, beginning next year, and valuing next year's unit at nine-tenths of today's. The continuing stream is worth one hundred and eighty today. An improper approval brings an additional one hundred and twenty immediately. Suppose detection occurs before next year's business and, if it occurs, destroys the whole future stream. With an eighty-percent chance of detection, the expected forfeiture is one hundred and forty-four, enough to outweigh the inducement. With a fifty-percent chance, the expected forfeiture is ninety, and the same inducement wins. These are invented terms, with ordinary current earnings unchanged and a risk-neutral certifier. Their point is that the size of the reputation and the chance of losing it are separate facts. A celebrated name can be badly disciplined when errors are hard to discover or easy to explain away.
A third-party certificate adds an investigation that the buyer could not cheaply conduct alone. Its advantage may come from specialized equipment, repeated examinations or the ability to compare many cases. The laboratory can discover a fault that one household would never learn to test for; an inspector can recognize a pattern that a purchaser meets only once. Sharing the cost of that competence can make otherwise doubtful trade possible. The mark nevertheless needs a defined object. A test of electrical safety establishes something different from an estimate of durability. A professional qualification records a person's training or attainment; a judgment about a particular service addresses what that person actually did. The authority of one finding should travel only as far as the finding itself.
Even an honest test leaves a question about what its result predicts. Suppose, in another invented example, a test passes ninety percent of sound units and ten percent of faulty ones. Among one hundred sound units and one hundred faulty units, it passes ninety of the first and ten of the second: nine-tenths of the approved group is sound. Apply the same test to ten sound units and ninety faulty units, treating the percentages as expected frequencies, and it passes nine of each: only half of the approved group is sound. The test has kept its stated performance while the meaning of its approval has changed with the population submitted. An institution that reports only the attractive success rate can leave the purchaser with the wrong inference without falsifying that particular number.
The person paying for the judgment can also shape the conditions under which it is made. A seller may reasonably pay for an examination because a trustworthy result makes the seller's product easier to buy. Payment by the inspected party therefore establishes an incentive to examine, as well as a possible incentive to flatter. The difficulty grows when that party controls a large stream of repeat business, can choose another certifier after an unwelcome result, and can keep the unfavorable assessment out of sight. The buyer sees the finished seal; the certifier sees the business that may be lost before the next seal is commissioned. Independence depends on the arrangements surrounding those two views.
The mortgage-securities examinations of 2008 made that dependence concrete. SEC staff examined Fitch, Moody's and Standard & Poor's work on subprime residential mortgage-backed securities and related collateralized debt obligations. Each used an issuer-pays model. The report identified weaknesses in separating rating work from fee discussions and business considerations, despite policies intended to manage the conflict. Its finding was more precise than a charge that every rating had been bought: in the market-share discussion, staff expressly reported no evidence that methodology or model decisions had been made on that basis. The record nevertheless showed why a written independence policy needed an operating structure behind it. An incentive may threaten a judgment before a particular corrupted judgment can be proved.
Competition among witnesses helps when purchasers can distinguish a rigorous judgment from a convenient one. It can work differently when the seller chooses which judgment the purchaser will see. Suppose two independent tests of the same unchanged object each have a one-half chance of returning a favorable result through noise. A seller allowed to publish only the better result can display a favorable result three-quarters of the time: only the case in which both tests are unfavorable prevents it. Neither examiner need have lied. The selection has changed the evidence reaching the buyer. A rule requiring all commissioned results to be disclosed would address that particular omission; it would still leave the quality of the tests to be examined. Counting the number of certifiers cannot settle which kind of competition they conduct.
Changing who pays can remove one pressure while leaving the work of verification unfinished. A purchaser-funded examiner has a clearer reason to serve the purchaser, but someone must collect the fee and prevent the information from being copied so freely that nobody funds its production. A public examiner can pool the cost, but requires competent staff, a usable standard and a means of correcting its own errors. Separating the assessor's compensation from the outcome of an assessment addresses a particular inducement; preserving inspection records makes another person able to challenge the reasoning. These arrangements deserve to be judged by the incentives and evidence they create. Calling the examiner independent supplies the beginning of that inquiry.
The seal can become a gate as well as a guide. Where buyers refuse to deal without an established mark, a new producer must obtain both the underlying quality and the accepted proof of it. Some of that cost pays for real knowledge; some may protect an incumbent standard against a better method that it was never designed to assess. Licensing goes further when it makes permission a condition of entry, while voluntary certification can leave an uncertified seller legally free to trade. Those arrangements answer different questions. A minimum competence requirement may prevent serious harm before reputation could form, yet the requirement still needs a defensible relation to the competence being protected. A valuable witness should be able to explain why its test bears on the transaction.
The seal earns its authority when its judgment can be tested.
The voucher is ultimately vouched for by a practice of correction that other people can use. A published standard, a preserved record, a separate review, a sanction that can actually be imposed: each gives the next observer something firmer than the last observer's assurance. The chain need not end in an infallible person. It needs enough opportunities for a mistake to be exposed and enough consequences to make concealment costly, while keeping the cost of scrutiny below the value it protects. The buyer will still rely on knowledge he does not possess. The achievement is to make that reliance answerable—to give the stranger's trust a place where reasons can be demanded and a wrong conclusion can be changed. The argument remains open.