Plate xii in print

The Secret

information as a good; the paradox of the thing you cannot value until you have it, and cannot be sold once you do.

The demonstration can transfer what the seller meant to sell.

— the argument of this plate, after Arrow (1962) and Anton–Yao (2002)

After the buyer has asked enough questions, the seller may discover that there is nothing left to sell. The explanation was the product. A method becomes intelligible through the steps that make it work; a useful fact establishes its usefulness by being revealed. The prospective customer who has learned enough to recognize the value may have learned enough to proceed without its owner. Throughout this volume, the missing information seemed to offer a remedy for the difficulty of exchange. Here the remedy must itself be exchanged. The seller needs to establish that he possesses something worth buying, while retaining something for which the buyer still needs to pay. The problem concerns the passage from one person's knowledge to another's, and the promises that can survive that passage.

Arrow's account of information as a commodity gives this problem its sharpest form. Inspection can transfer the information before payment; incomplete control over its use can then weaken the seller's ability to collect. The difficulty is conditional. A buyer who acquires a fact can still owe an enforceable payment, need further undisclosed knowledge, or require help putting it to work. Nor must every buyer learn the entire fact before assigning it any expected value. The question is how to make the transaction credible when the buyer cannot adequately assess the offered information beforehand and its transfer cannot be adequately governed afterward. Uncertainty about quality and difficulty enforcing exchange meet at the same point: the moment the secret becomes known.1

1. Kenneth J. Arrow, “Economic Welfare and the Allocation of Resources for Invention” (1962), “Information as a Commodity,” pp. 614–616, especially p. 615. His information-sale paradox expressly depends on incomplete appropriability; he also considers buyers judging a class of information from prior experience.

A small calculation separates valuation from delivery. Suppose a decision-maker can undertake a project costing 60. It will return 100 if conditions are favorable and nothing otherwise, with each state initially assigned probability one half. He is risk-neutral, can decline the project for zero, and must decide after receiving any report he buys. Without a report, the project's expected net return is minus ten, so he declines. A perfectly accurate report lets him undertake it only in the favorable state. Its expected gross benefit is therefore one half of forty, or twenty. He can value that report before knowing which answer it contains. A fee of five leaves an expected net benefit of fifteen. None of this arithmetic ensures that the seller really has an accurate report, can demonstrate its quality without disclosure, or can collect the agreed fee. Valuation supplies a reason to exchange; it does not supply the machinery that makes exchange possible.2

2. This is an illustrative value-of-information calculation with a stipulated prior, credible perfect signal and costless use of the result. Without information: max(½ × 100 − 60, 0) = 0. With information: ½ × max(100 − 60, 0) + ½ × max(0 − 60, 0) = 20. The realized value differs across states; twenty is the ex ante gain.
⁂

Showing part of a method can sometimes establish the value of what remains. Anton and Yao model a seller whose privately known stock of know-how improves buyers' prospects of commercial success. A partial disclosure is freely usable by the buyers, yet can signal how much further knowledge the seller possesses. The contract that follows is contingent on verifiable commercial outcomes. The seller's ability to post a bond matters: less wealth can require more disclosure and leave less value to sell. This is a particular equilibrium construction, involving competing buyers and specified contracting possibilities. It does not make every demonstration a reliable signal. It shows why giving something away can be part of a sale, and why an inventor's ability to finance a credible promise can affect what he receives for his invention.3

3. James J. Anton and Dennis A. Yao, “The Sale of Ideas: Strategic Disclosure, Property Rights, and Contracting”, Review of Economic Studies 69 (2002), §§1–2, pp. 513–518, and §5, p. 527. Their model distinguishes unverifiable transfers of know-how from verifiable downstream payoffs and imposes a limit on the seller's bond.

Other arrangements place the object of payment around the disclosure. A consultant can be paid for a continuing service whose value includes explanation and implementation. A seller with repeated customers can put future business behind the quality of the current report. A demonstration can reveal an output while withholding the procedure that produced it. A contract can specify payment before access, provided the purchaser has some adequate basis for trusting what will follow. Each arrangement moves the difficulty to a place where evidence or an obligation may be easier to establish. None makes those requirements disappear. A reputation must be informative about the particular work; an implementation service must remain useful after the explanation; a promised restriction must be capable of doing what the bargain assumes. The institution is part of the product's economic life, because it determines which parts of its value can support an actual exchange.

The calculation also shows why more knowledge need not always be worth its cost. In the same example, a person who already knows the state gains nothing from buying the perfect report. If the project pays 100 in the favorable state and costs nothing in either state, undertaking it is already an optimal choice without additional information; the report again adds no decision value. A fact can be true and newly learned while leaving the best action and expected payoff unchanged. Its worth depends on what the recipient already knows and what he can do differently. The distinction also applies to shared knowledge. One person's learning need not consume what another can learn, while teaching, transmitting, maintaining and applying it can still require scarce resources. Counting the information and counting the work needed to use it are separate parts of the same inquiry.4

4. These are boundary cases of the stipulated decision problem. With a zero project cost, acting without the report earns an expected 50; conditioning action on the report also earns 50. A fee then lowers net value. Non-rival knowledge does not by itself establish zero transmission or implementation costs.

The volume ends with a market trying to purchase the means of understanding its other purchases. The difficulty of that task explains neither every failure nor every success in the sale of knowledge. It tells us to follow the information through the transaction: what is revealed, what remains scarce, what can be promised, and who must bear the expense before a buyer can judge the result. Those questions continue after the first sale, when knowledge reaches people who never bargained with its producer. The next volume, The Commons, asks how shared resources are provided and governed. Information gives that inquiry a distinctive starting point. Learning can enlarge the company of those who possess something without diminishing anyone else's possession, while leaving the work of producing and sustaining it to be paid for.5

5. The transition concerns the distinction between use and provision. The value-of-information example measures one recipient's private decision gain; it supplies no general estimate of the social value of research, no universal case for secrecy, and no universal rule that an existing fact should carry a positive price.