Frank Knight, then a 35-year-old economist at Iowa, published his Chicago PhD as a book in 1921 under the title Risk, Uncertainty, and Profit. The book drew a fundamental distinction — now called Knightian uncertainty — between two situations economists had been conflating. Risk is the situation where the probability distribution of outcomes is known: roulette wheels, fire-loss tables, casino odds. Uncertainty is the situation where the probability distribution itself is unknown — the situation entrepreneurs and innovators actually face. Profit, on Knight's account, is the reward for bearing uncertainty, not risk. John Maynard Keynes's General Theory (1936) reached a parallel conclusion in Chapter 12, with the operative driver named animal spirits. The 2008 financial crisis was a vast empirical vindication of the Knight-Keynes distinction.
Knight's distinction is operationally precise. Risk has three features: outcomes are clearly defined, probabilities are known or knowable through repeated trial, and frequency-based pricing is possible. Uncertainty lacks at least one: outcomes may not be enumerable in advance, probabilities may not exist in any frequency-based sense, or the structure of the problem itself may change. An insurance company can price fire-loss probabilities — that is risk. The same company cannot price the probability of a novel pandemic, systemic financial crisis, or new political regime — those are uncertainty. The distinction matters because probabilistic decision theory (von Neumann-Morgenstern expected utility, Savage's 1954 framework) assumes knowable probabilities; when they are not knowable, the framework must be supplemented with criteria like maximin or ambiguity aversion. Daniel Ellsberg's 1961 paradox — people prefer known probabilities over unknown ones even when the unknown distribution has higher expected payoff — established that uncertainty aversion is real and measurable. Hyman Minsky's financial-instability hypothesis builds on Keynesian uncertainty: stability breeds risk-taking, which produces fragility, which produces crisis. Robert Shiller's 1981 paper on excess stock-market volatility showed financial markets are uncertainty-priced, not risk-priced. The 2008 crisis was the consequence of risk-pricing mortgage-backed securities whose actual uncertainty had been bracketed by Gaussian-copula models that assumed fat tails were thin. Nassim Nicholas Taleb's The Black Swan (2007), published a year before the crisis, is the most-influential popularization.
Knightian uncertainty has moved from heterodox to mainstream after 2008. Central-bank communication now routinely distinguishes risk (model-priced) from uncertainty (requiring judgment and narrative). Climate-change economics is substantially organized around the distinction: Martin Weitzman's dismal theorem (2009) argued that fat-tailed climate distributions make standard cost-benefit analysis ill-defined, because the expected loss can be infinite under reasonable tail assumptions. Pandemic preparedness is similarly framed: pandemic risk was priced (small annual probability of a flu pandemic), but pandemic uncertainty (a novel coronavirus with specific transmission and lethality profile) was not pricable. AI safety and existential-risk economics — Toby Ord's The Precipice (2020) — operate substantially within Knightian uncertainty. Decision theory under deep uncertainty has emerged as a substantive subfield.