For 300 years, economists assumed that rational agents evaluate outcomes based on total wealth. Daniel Bernoulli proposed in 1738 that a person with $1 million and a person with $9 million experience $5 million differently — the first feels rich, the second feels impoverished. He was right about diminishing marginal utility. He was wrong about the carrier of value.
The Framework
Expected utility theory (EUT) holds that a rational agent evaluates options by multiplying the utility of each outcome by its probability and choosing the option with the highest expected utility. Bernoulli's key insight was that utility is not linear with wealth — the joy of gaining $1M when you have $1M is greater than gaining $1M when you already have $9M (diminishing marginal utility). This explained risk aversion for the first time: people prefer a sure $500K over a 50/50 chance at $1M because the first $500K has higher utility per dollar than the second $500K.
The error: Bernoulli defined utility as a function of absolute wealth states. Jack at $5M (up from $1M) and Jill at $5M (down from $9M) have identical utility because they have identical wealth. But Jack is elated and Jill is devastated — their reference points, not their wealth levels, determine their experience. Kahneman and Tversky's prospect theory corrects the 300-year error by defining utility as a function of changes from a reference point, not absolute states.
Where It Comes From
Kahneman devotes Chapter 25 of Thinking, Fast and Slow to Bernoulli's theory and its failure. The Jack-and-Jill thought experiment is the decisive counterexample. The chapter title — "Bernoulli's Errors" — is deliberately provocative: Kahneman argues that the error was not obscure or subtle; a first-year psychology student could construct the counterexample. The theory survived unchallenged for 300 years through theory-induced blindness.
> "His theory can handle the fact that a gift of $10 has a bigger impact on someone with $100 than on someone with $10 million, but it cannot handle the fact that Jack is much happier than Jill." — Thinking, Fast and Slow, Ch 25
Cross-Library Connections
Fisher's Getting to Yes implicitly corrects Bernoulli's error when it insists on understanding the other side's reference point (Currently Perceived Choice Analysis) rather than evaluating the deal's absolute terms. A $500K settlement feels different to someone who expected $1M (loss) versus someone who expected $0 (windfall).
The Implementation Playbook
Pricing: Don't evaluate your price against the customer's absolute wealth or budget — evaluate it against their reference point. A $997 product feels expensive to someone whose reference is "$0 for a free alternative" and cheap to someone whose reference is "$5,000 for a consultant."
Compensation: Same salary, different experience. An employee hired at $80K who gets a raise to $100K (gain of $20K) is happier than one hired at $120K who gets a cut to $100K (loss of $20K). Same wealth, opposite feelings. Bernoulli can't explain this; prospect theory can.
Policy Evaluation: Don't evaluate policy outcomes by final states — evaluate them by changes from the status quo. A tax reform that leaves everyone at the same final wealth can still create winners (those who gained) and losers (those who lost) whose experiences are asymmetric due to loss aversion.
Key Takeaway
Expected utility theory was the best model of rational choice for 300 years — and it was wrong about the most fundamental question: what do people evaluate? Not wealth states. Changes. The correction seems trivial but changes everything: it predicts loss aversion, the endowment effect, status quo bias, framing effects, and every phenomenon in Part IV of the book.
Continue Exploring
[[Prospect Theory Value Function]] — The S-curve that corrects Bernoulli by centering on changes from a reference point
[[Reference Dependence]] — The principle Bernoulli missed: carriers of value are changes, not states
[[Theory-Induced Blindness]] — Why Bernoulli's error survived unchallenged for 300 years
📚 From Thinking, Fast and Slow by Daniel Kahneman — Get the book