1.1The question
A person who takes a decision under uncertainty usually writes something down: an order ticket, a memo, an allocation rule, a trade log. What is written at that moment says what the decider thought the decision was a case of. A forecast of rain is a case of the forecaster's forecasts; a fire policy is a case of buildings of a certain construction in a certain district; an entry into a market is a case of entries made under a certain signal. If the written description fixes the class, a frequency can later be counted against it. If it does not, nobody will ever know how often decisions of that kind turn out well.
This article asks a narrow question of two such writings. When a decision record states what makes its decision hard, is the difficulty a missing frequency, because the case cannot yet be put in a class of like cases, or a missing market, because no one outside the decider can take the risk on at a fair price? The question comes from a long dispute over Frank Knight's Risk, Uncertainty and Profit (1921), and it is put here to two records that WSTiger, a studio in the RISCON family, publishes for study. Both records are synthetic. They are modern summaries written for the platform, not archival documents, and the firms, agents and traders in them are invented.
1.2The contribution
LeRoy and Singell (1987) read Knight's line between risk and uncertainty as a line about insurance markets: what cannot be insured, because of moral hazard and adverse selection, is uncertain. Langlois and Cosgel (1993) read it as a line about classification: what cannot be sorted into classes of like cases has no frequency, and so calls for judgment. Each side has its passages in Knight. What this article brings to the exchange is a different kind of material: documents written at the moment of decision, which can be asked which of the two difficulties the decider faced.
The contribution here is to supply that kind of evidence, in a limited and clearly labelled form. Two synthetic records written for a decision platform are read beside Chapters VII and VIII of Knight's book. The records were not written to test Knight, and their authors were not reading him. They were written to set a decision and its revisit conditions down before the outcome, which is the moment the dispute is about. Read this way, the records show the missing class at the time of writing and a proposal to build one by keeping a record; they show the threat to that class coming from the decider's own interest in how the case is later described. Chapter VIII of Knight's book already ties the two difficulties together, in a single passage about the typical business risk that no outside agency can insure. The records add a sense of order in time: the class is missing first, and the incentive to bend it arrives with the attempt to build it.
The placement of the secondary works in this debate, and their application to the two records, are this article's own. Each work is credited only with what its reading note in the parent study records.
1.3The debate joined
On one side stand LeRoy and Singell (1987), who hold that Knight took subjective probabilities for granted and that his uncertainty is the region where insurance markets fail. On the other stand Langlois and Cosgel (1993), who hold that Knight's uncertainty is the region where instances cannot be classed, and that the judgment it calls for is what a firm is built around. Two further positions frame the exchange. Keynes's A Treatise on Probability (1921), published in the same year as Knight's book, treats probability as a logical relation between evidence and conclusion that is not always numerical, and separates the weight of evidence from probability. Ellsberg (1961) showed that people prefer a bet with known odds to an otherwise equal bet with unknown odds, a pattern that gives Knight's distinction an experimental form. More recently Kay and King (2020) have argued that many decisions face a radical uncertainty that probabilities cannot capture.
The article proceeds as follows. Section 2 sets out the debate. Section 3 describes the primary text, the records, and the parent study's ledger. Sections 4 to 6 are three case sections, each opening with a claim about Knight's text, reading that text closely, and then bringing in a record. Section 7 discusses what the records can and cannot carry, and Section 8 concludes.