A cab was involved in a hit-and-run. 85% of cabs in the city are Green; 15% are Blue. A witness identified the cab as Blue. The witness correctly identifies cab colors 80% of the time. What's the probability the cab was Blue? Most people say 80%. The correct answer is 41%. But the story changes completely when you reframe the base rate.

The Framework

Kahneman's Chapter 16 draws a critical distinction that most treatments of base rate neglect miss: not all base rates are created equal. Statistical base rates (population-level frequencies without causal context) are systematically ignored by System 1. "85% of cabs are Green" feels like background information — it doesn't change how you process the witness's testimony. But causal base rates (information that carries an individual, causal interpretation) are used. "The Blue cab company has been involved in a disproportionate number of accidents" changes the probability because it provides a causal mechanism — those drivers are reckless.

Both statements convey the same statistical information about Blue cabs, but they produce different judgments because System 1 processes causal stories and ignores statistical distributions. The cab problem is the cleanest demonstration: replace the population frequency with a causal mechanism, and suddenly people use the base rate correctly.

Where It Comes From

Chapter 16 of Thinking, Fast and Slow builds on the representativeness research from Chapters 14-15 to explain when base rates are used and when they're ignored. Tversky and Kahneman's cab problem (1982) is the canonical example. The causal/statistical distinction was identified when researchers noticed that simply giving the base rate a causal framing (reckless driving) produced dramatically better Bayesian reasoning than the statistical framing (proportion of cabs).

> "People who are taught surprising statistical facts about human behavior may be impressed to the point of telling their friends about what they have heard but this does not mean that their understanding of the world has really changed." — Thinking, Fast and Slow, Ch 16

The Implementation Playbook

Persuasion: When presenting data to support an argument, give it a causal frame. "30% of startups in this space fail within 2 years" (statistical) will be ignored. "Startups in this space fail because customer acquisition costs exceed lifetime value before the team can iterate" (causal) will be processed. Same information, different framing, dramatically different impact.

Risk Communication: Don't present risks as population statistics — present them as individual stories with causal mechanisms. "1 in 10 projects fail" is ignored. "Projects fail because scope creep exhausts the budget before the core features are delivered" activates System 1's causal processing.

Teaching: Abstract statistical principles are forgotten immediately. Case studies with causal narratives are remembered and applied. If you need to teach Bayesian reasoning, use causal examples (the reckless cab company) rather than statistical examples (the proportion of cabs).

Medical Communication: "The base rate of this condition is 1 in 1,000" will be ignored by patients. "People develop this condition when they carry a specific genetic variant that affects how the body processes X" will be processed — because it provides a causal mechanism that System 1 can use.

Key Takeaway

The causal/statistical distinction explains one of the most persistent failures of human reasoning: we have statistics, we know they're relevant, and we ignore them anyway. The solution isn't to present better statistics — it's to give statistics a causal costume. Transform "30% of X fail" into "X fails because of Y." System 1 processes stories, not numbers. Dress your numbers as stories.

Continue Exploring

[[Representativeness Heuristic]] — The mechanism that causes base rate neglect: judging by resemblance

[[Narrative Fallacy]] — Why causal stories are processed while statistical distributions are ignored

[[Denominator Neglect]] — A related mechanism: the individual case overwhelms the population statistics

📚 From Thinking, Fast and Slow by Daniel Kahneman — Get the book