Markets run on knowledge, unequally held, costly to gather and never complete. The buyer wants to know what the seller has; the insurer wants to know who is asking; the principal wants to know what the agent has done. Throughout this volume, the parties have devised ways of dealing across those gaps. A signal puts something at stake, a screen makes different choices revealing, a warranty connects a promise to a consequence. The price that emerges can reflect useful production and the cost of risk as well as the advantage of knowing more. To understand what it measures, one must understand the arrangement that produced it. The curtain between the parties changes the bargain without reducing every bargain to the same kind of concealment.
The obvious remedy is to produce the missing knowledge and sell it to whoever needs it. Inspection, research and advice can change a decision before the mistake becomes expensive. Arrow's difficulty begins when inspecting the proposed information also delivers it. A buyer may want the answer before deciding what the answer is worth; a seller who supplies it may find that little remains to exchange. The difficulty is conditional on what can be withheld, verified and protected. A buyer can still value a class of information from prior experience, and an enforceable right over its use can preserve something to sell after disclosure. The paradox identifies an obstacle to exchange, rather than proving that information has no assessable value until it is possessed.
The expected value of knowing can be calculated before the answer arrives when the alternatives and probabilities are specified. Consider the hypothetical project in the later Utility Secret. It costs sixty to undertake and returns a hundred in a favorable state, nothing in an unfavorable state; each is initially equally likely. Without further information, undertaking it has an expected net return of minus ten, so a decision maker who maximizes expected money chooses to abstain and earns zero. A perfect report permits undertaking the project only in the favorable state. That yields forty half the time and zero otherwise, for an expected gain of twenty. The answer is still unknown, but the value of the opportunity to condition the decision on it is already determinate.
The price of the report and its value to the decision need not be the same. A fee of five leaves the buyer an expected gain of fifteen in the example; twenty is the largest fee that leaves the buyer at least as well off as abstaining. Bargaining and alternative sources can determine where a price falls within the available range. The information also has limits. If the state is already known, the perfect report adds nothing. If the project costs nothing and can never lose money, undertaking it in either state already produces the same expected return as conditioning the choice on the report. Knowledge can be correct and still change no worthwhile decision. The missing comparison is therefore between actions made possible by the information, not between ignorance and omniscience in the abstract.
The exchange problem returns when the buyer asks the seller to prove that the report deserves trust. Revealing which state will occur may hand over everything useful before payment. Withholding the answer leaves the buyer to judge the seller, the method or the contract surrounding the offer. The seller can have an answer worth twenty to this buyer and still struggle to collect five for it. Conversely, an impressive presentation can command a fee without supplying the promised knowledge. Expected value and the ability to transact are separate parts of the problem. A calculation can establish the first under its assumptions while leaving the second dependent on institutions.
Partial disclosure can sometimes demonstrate knowledge while leaving valuable knowledge unsold. Anton and Yao model an inventor who exposes part of an idea to competing prospective buyers. That exposure is costly to the inventor because the recipients can use what they learn. What remains undisclosed can still command a payment, supported by contracts tied to verifiable commercial outcomes. The seller's capacity to post a bond matters to the signaling arrangement: different kinds of sellers must have different reasons to choose the offered disclosure and terms. The result gives a route through the dilemma under specified conditions. A demonstration alone is no universal proof, and an inventor with little wealth may be unable to support the same transaction as an otherwise similar inventor with more.
Verification can also be delegated, which moves the informational boundary without abolishing it. A buyer may rely on an inspector who tests a product, a professional who interprets a result or an institution that stands behind a certificate. The buyer need not acquire all the specialist's knowledge for the service to be useful. The terms must instead give the specialist a reason to perform the task and a consequence for misrepresenting it. This was the work of the Seal and the difficulty of Credence: some services are valuable precisely because the customer cannot reproduce the judgment being purchased. Making the intermediary answerable may be more practical than making every customer an expert.
The costs of producing knowledge and distributing it can pull the account in different directions. A research program may require years of work, yet an additional person can sometimes use the result at little expense once it exists. Another person's use need not consume the knowledge or deprive the first user of it. Restricting access can help the producer collect a return, while preventing applications whose benefits exceed the extra cost of supplying them. Free access can encourage those applications while leaving the original work unfunded. Arrow separates these margins because the incentive to create an answer and the efficient use of an existing answer are different decisions. Their institutional solution must connect them without pretending their costs are identical.
A simple production account shows the force of that distinction. Suppose making a reliable instruction costs twenty, delivering it costs one per user, and four potential users value it at twelve, ten, eight and four. If all four receive it, benefits total thirty-four and costs twenty-four, leaving ten. A uniform fee low enough for all four to buy cannot exceed four, so receipts cannot exceed sixteen. That fee cannot finance production and delivery. A fee of eight, with a user indifferent at that price choosing to buy, serves the first three and yields twenty-four against costs of twenty-three. It finances the work but leaves out a fourth use worth four that would cost only one more to supply.
The funding gap in that example permits several responses without selecting one by itself. A payment for creating the instruction, followed by distribution at delivery cost, could support all four uses if the creation payment can be raised and the promised quality secured. Different charges could also cover the total cost while admitting everyone, provided the seller can distinguish buyers and prevent profitable resale. Those conditions may be expensive or objectionable. The uniform-price arrangement is imperfect under the stipulated total-surplus criterion but has fewer requirements. Comparing institutions means including the cost of organizing the alternative and the effects of its payments, rather than treating a better allocation on paper as a free administrative achievement.
A private gain from learning also needs to be distinguished from a gain available to everyone together. Information can improve production, prevent waste or help match a product to the person who values it. It can also strengthen one party's position in dividing a fixed amount. If a fixed asset remains worth a hundred in the same use and advance knowledge lets a buyer negotiate its transfer for ten less, that ten moves between buyer and seller; it does not, merely through the changed price, create ten more output. Any real resources spent obtaining the bargaining advantage must be counted separately. An inquiry into knowledge therefore asks both what it changes and for whom. A private willingness to pay supplies evidence of value to the payer, not a complete social account.
The same discipline governs the institutional conclusions of the volume. A screen may preserve trade while burdening the people asked to distinguish themselves. A warranty may strengthen a promise while changing the care taken by its beneficiary. Required disclosure may supply facts that remain costly to interpret. These comparisons make the informational problem concrete; none licenses a presumption that every private response succeeds or every public response improves it. An institution's own information, incentives and means of enforcement belong in the account. The relevant alternative is an arrangement that could actually operate under those conditions, not a decision maker who acquires perfect knowledge without effort and uses it without conflict.
The price of knowing depends on how the knowledge can pass between us.
The curtain leaves the price with a history rather than depriving it of all meaning. Behind an observed sum may stand productive skill, protection from uncertainty, costly verification, bargaining power and a fact that only one side has seen. The work of the volume has been to separate those forces and follow the institutions that let people deal despite them. Its last question opens into the next. Knowledge can serve several people without being consumed, while the work of producing and maintaining it still needs support; access and contribution must somehow be arranged together. The Commons takes up that shared resource and the rules by which people may care for it. The argument remains open.