The Ellsberg Paradox and Ambiguity in Trading Decisions

The Ellsberg paradox shows preferring known odds over ambiguous ones cannot be reconciled with any single subjective probability, a bias with trading implications.

The Ellsberg paradox describes a pair of choices that most people make consistently, yet no single set of subjective probabilities can justify both choices at once, which means the standard expected utility framework, maximize probability times payoff using some fixed belief, cannot explain typical human behaviour under ambiguity. The result matters for trading because it shows that aversion to unknown probabilities is a distinct force from ordinary risk aversion over known ones, and desks price that distinction directly into spreads on instruments with uncertain models.

An urn contains 90 balls, 30 are red, and the remaining 60 are some unknown mix of black and yellow, in a ratio nobody is told. A ball will be drawn at random. In the first choice, a person picks between betting on red, a known one in three chance, or betting on black, an unknown chance somewhere between zero and two in three.

In the second choice, the same person picks between betting on red or yellow together, an unknown chance, or betting on black or yellow together, a known two in three chance, since black and yellow always add up to exactly 60 of the 90 balls regardless of the unknown split.

Most people, when tested, prefer the known bet in both choices, red over black in the first pair, and black-or-yellow over red-or-yellow in the second pair. Suppose a person is a standard expected value maximizer holding some fixed personal probability for how many of the 60 balls are black, call the implied probability of black p. Preferring red over black in the first choice requires one third to be greater than p.

In the second choice, the probability of red-or-yellow equals one minus the probability of black, which is one minus p, while the probability of black-or-yellow is fixed at two thirds. Since p is less than one third from the first preference, one minus p is greater than two thirds, meaning the red-or-yellow bet actually has the higher implied probability of winning. A consistent expected value maximizer with belief p should therefore prefer red-or-yellow over black-or-yellow in the second choice, the opposite of what most people actually pick.