John Kay and Mervyn King on Radical Uncertainty

Episode Summary

Executive Summary: Russ Roberts interviews Mervyn King and John Kay about Radical Uncertainty, arguing that many important real-world decisions cannot be reduced to precise probabilities or expected-utility calculations. The conversation critiques overconfident models, cost-benefit analysis, value-at-risk, and the misuse of “rationality,” while defending judgment, narrative, and abductive reasoning in uncertainty-rich settings like pandemics, finance, and public policy.

Main Topics: Radical uncertainty vs. measurable risk (Priority: 5/5): King and Kay distinguish quantifiable risk from uncertainty that cannot be assigned reliable probabilities, using COVID-19 as the central example. Limits of models in pandemics and policy (Priority: 5/5): They argue epidemiological and economic models are useful for framing questions but unreliable for precise prediction when key parameters depend on unknown human behavior and incomplete data. Critique of cost-benefit analysis and spurious precision (Priority: 4/5): The guests warn that precise-looking numbers in policy spreadsheets often rest on weak assumptions, hide uncertainty, and allow decision-makers to evade responsibility. Finance, value at risk, and regulatory failures (Priority: 4/5): They explain how value-at-risk and capital rules can create false confidence by relying on historical data that omit extreme events, especially before crises. Rationality, behavioral economics, and small-world models (Priority: 5/5): The discussion challenges the economics profession’s narrow definition of rationality as expected-utility maximization, arguing that many real decisions are better understood as adaptive responses to large-world uncertainty. Puzzles vs. mysteries; human judgment and adaptability (Priority: 4/5): They emphasize that many real problems are mysteries, not puzzles, and that humans succeed by using heuristics, rules of thumb, and judgment rather than exact optimization. Methodology and the future of economics (Priority: 4/5): The episode closes with a defense of broader reasoning styles—deductive, inductive, and abductive—arguing economics needs more realism and humility without abandoning mathematics or empirical work.

Key Arguments: Precise probabilities are often fabricated in situations where the underlying uncertainty is genuinely unquantifiable. Pandemic models can identify broad dynamics, but point estimates about deaths or lockdown timing are too dependent on unknown parameters to support confident policy claims. Cost-benefit analysis is useful for structuring thought, but becomes misleading when it turns unknowns into false decimals and single-number policy rankings. Value-at-risk failed in part because it used historical distributions that excluded the extreme tail events it was meant to guard against. Regulatory systems built around arbitrary thresholds encourage complacency once firms merely satisfy the metric. Behavioral economics correctly noticed that people do not literally maximize expected utility, but it wrongly re-labeled departures from the model as biases rather than adaptive behavior. Expected-utility theory may work in “small worlds,” but most consequential decisions occur in “large worlds” with incomplete, shifting information. Human beings are successful not because they calculate better than computers, but because they are adaptable, imaginative, and good enough under uncertainty. Economics needs deductive, inductive, and abductive reasoning; deductive modeling alone cannot handle policy, entrepreneurship, or personal life decisions.

Data Points: COVID-19 deaths forecast (US): 2.2 million - Referenced as the Imperial College estimate criticized for false precision COVID-19 deaths forecast (UK): 550,000 - Imperial College model estimate discussed alongside the US estimate Reduction in lives saved by earlier lockdown: 36,456 - Example of a highly precise estimate used to criticize misleading point estimates Lockdown timing comparison: 1 week earlier - Counterfactual used in the saved-lives estimate Precision example: 383.7 rounded to 384 - Russ Roberts’ illustration of absurd decimal precision in forecasts Banking stress-test concern: Not quantified - King argues future stresses are unknowable and cannot be reduced to a single test Macroforecasting performance: Poor when anything significant happens - General claim that economists predict little beyond stable periods

Pivotal Quotes: "what is going on here?" — Mervyn King: He describes the most important question to ask under radical uncertainty instead of over-relying on numerical forecasts "if we do not act in accordance with axiomatic rationality and maximize our subjective expected utility, it is not because we are stupid, but because we are smart" — Book quote discussed by Russ Roberts: Used to summarize the book’s argument that human cognition is adaptive rather than defective "uncertainty is what makes life interesting" — John Kay: He argues that uncertainty is not merely a problem to eliminate, but part of what gives life value and variety

Implications: Listeners should be skeptical of precise forecasts and single-number policy answers in crises. The episode urges decision-makers to focus on judgment, resilience, and honest acknowledgment of ignorance rather than false certainty.