The Unknown

The insurer does not know which house will burn. The founder does not know whether anyone will buy the thing he proposes to make. Each commits resources before the answer arrives, yet the uncertainties need not be alike. The insurer may have a population of comparable houses, a record of losses, and grounds for estimating how much variation the coming year will bring. The founder may have a handful of promising conversations and no close precedent for the market he hopes to open. Calling both situations risky leaves out the difference in what their numbers mean. Frank Knight made that difference central to his account of profit: the return left after the contracted claims have been met depends on judgments whose reliability cannot itself be read from a table. The preceding plates have priced hazards through their distributions. This one asks what becomes of the price when the distribution is among the things in doubt.

Knight distinguished probabilities derived from a known arrangement, probabilities estimated from a sufficiently comparable class, and estimates for which no such classification could do the required work. A fair die belongs to the first kind; a record of repeated industrial losses may support the second; the prospective sales of a particular new enterprise depend heavily on the third. The distinction concerns the basis of the judgment, rather than whether the person making it can write down a number. Someone may forecast a venture's success at sixty in a hundred and still have much less confidence in that estimate than in the chance of drawing a specified ball from a counted urn. Knight expressly allowed gradations: actual cases are neither perfectly homogeneous nor entirely without resemblance.¹ His unknown is therefore more demanding than a blank page. There may be experience, analogy, and trained judgment, but no sufficiently stable class of repetitions to turn them into the objective frequency the calculation appears to promise. An estimate can be useful without acquiring the authority of a count.

The economic force of the distinction appears in the treatment of costs. Where losses can be grouped with sufficient independence and regularity, an enterprise can make provision for them as it provides for materials. The bill for cover becomes one of the sums that expected receipts must meet. But the judgment that receipts will meet all those sums remains exposed. In Knight's enterprise, the entrepreneur offers workers and other suppliers agreed payments and takes the residual after they are paid. If receipts exceed the commitments, there is something left; if they fall short, there is a loss. This is a description of how the uncertain result is allocated, not a schedule promising a positive wage for courage. Knight allowed entrepreneurs as a class to lose, and he distinguished the residual from payment for the owner's ordinary labor and property.² Nor does an agreed wage make a worker's future perfectly secure. A guarantee is only as good as the means behind it; suppliers may accept incomplete security and share the exposure. The firm organizes uncertainty. It cannot abolish uncertainty by assigning its remainder a name.

Daniel Ellsberg gave the uncertainty of the odds a small and exact test. One urn contains fifty red balls and fifty black; another contains a hundred in an undisclosed mixture. Offer the same prize for a red draw from either urn, then offer the same prize for a black draw. A person who prefers the counted urn in both comparisons cannot be represented as maximizing expected utility with one fixed probability for red in the unknown urn. The first preference requires that probability to be below one-half; the second requires it to be above one-half. Changing the curvature of utility cannot repair the contradiction, because the prizes are identical. Ellsberg reported such preferences from informal interrogation, including responses that survived reflection; he did not report them as the choices of every person.³ The comparison isolates a reluctance to rely on the undisclosed composition. It is distinct from a reluctance to bear a known gamble, and it need not disappear merely because someone supplies a best guess about the missing odds.

For a price, this creates a second inquiry beside the cost of a known spread. Imagine a buyer offered a claim that pays one hundred if a proposed product succeeds and nothing if it fails. A forecast of one-half gives an expected payment of fifty. That arithmetic does not establish what the claim is worth to the buyer: the timing of the payment, its connection with his other resources, and his confidence in the forecast still matter. If he pays forty for a claim with a well-supported one-half chance but only thirty for this one, the difference may express a discount for ambiguity. Yet those illustrative prices cannot be carried into another transaction as a universal ten-dollar charge for ignorance. Another buyer may know the industry better, trust the forecast more, or believe success more likely. A third may decline to buy at either price because failure would coincide with the loss of his own job. Observed discounts mix beliefs, exposure, and attitudes. The missing probability does not prevent an offer; it prevents the offer from serving, by itself, as proof that the probability has been discovered.

Judgment also changes what can be done before the final wager. A founder can test a product on a smaller scale, seek commitments from customers, or arrange expenditure in stages. An investor can ask which assumptions drive the estimate and how the result changes when they move. Such actions have costs, and postponement may surrender an opportunity, so information is neither free nor an instruction to wait forever. Their purpose is to alter the decision's exposure to what is poorly known. These are applications of Knight's distinction, rather than a formula he supplied for every enterprise.⁴ They also show why uncertainty cannot simply be identified with merit. A person may incur avoidable ignorance, make a disciplined judgment, or take a claim whose losses others ultimately bear. The realized profit alone does not tell us which occurred. Even a carefully chosen venture can fail, while an ill-considered one can prosper. To assess the contribution requires the decision, the available information, and the allocation of consequences; a surviving fortune is only the last of those records.

The answer to the plate's question is consequently not one additional premium to be entered beside beta. Where the odds themselves are unsettled, the terms depend on whose judgment is trusted, whose resources are committed, and who can absorb the error. Some uncertainties can be reduced; some can be distributed; some remain in the decision after every available calculation has been made. Knight explains why an enterprise can leave a residual that was never contractually determined in advance. Ellsberg shows why a chooser's response to incomplete odds may resist representation by a single probability distribution. Neither result requires that every unfamiliar prospect command a higher return, nor that no price can be agreed. The next plate turns to a premium whose past magnitude can be counted, yet whose explanation is still difficult.⁵ There the number observed is much larger than a particular account of the dangers borne can explain. Knowing the price and understanding what it pays for remain separate achievements.