Causes Trump Statistics

Kahneman returns to base-rate neglect with a crucial refinement: not all base rates are ignored equally. Purely statistical base rates — bare facts about how common something is — are routinely neglected, but causal base rates, which fit into a story about how the world works, are used. The mind disregards statistics that carry no causal meaning and embraces the same numbers when they can be woven into an explanation, revealing that our problem is not with numbers as such but with facts that lack a narrative.

His central illustration is the cab problem. A cab was involved in a hit-and-run; 85% of the city's cabs are green and 15% blue; a witness identifies the cab as blue, and the witness is correct 80% of the time. Most people judge it highly likely the cab was blue, fixating on the witness and ignoring that blue cabs are rare; the Bayesian answer, combining the witness's reliability with the base rate, is only about 41%. The statistical base rate — mere proportions of cab colors — is neglected because it tells no causal story.

Now Kahneman changes one detail: the two companies run equal numbers of cabs, but blue cabs are responsible for 85% of accidents. The numbers are identical, but suddenly people use the base rate, because it now reads as a fact about reckless blue-cab drivers — a cause. The statistical base rate became a causal base rate, fitting a stereotype about who drives dangerously, and once it carried causal meaning, intuition incorporated it. The same information is ignored as statistics and used as a cause.

The consequence is a deep asymmetry in how we reason. We are drawn to explanations and stereotypes — causal stories about kinds of people and how things happen — and we are nearly blind to dry statistical facts that do not slot into such a story. This makes us neglect the prevalence of conditions, the size of reference classes, and the role of chance, while readily absorbing any number that confirms a causal narrative, however loosely. Stereotypes, for better and worse, are causal base rates the mind willingly applies.

The applied takeaway is to deliberately attend to statistical base rates even when — especially when — they carry no causal story. When evidence points one way (a vivid witness, a striking case) but the underlying prevalence points another, force yourself to combine them rather than letting the causal-feeling evidence dominate. Ask whether you are ignoring a base rate simply because it is a bare statistic with no narrative attached, since those are precisely the facts intuition discards.

Kahneman's deeper observation is that this chapter completes the picture of why base-rate neglect is so hard to cure: the fix is not merely to 'use the numbers' but to recognize that our minds filter facts through a demand for causal sense. We reason in stories, and a fact that cannot be told as a story is, for System 1, almost invisible. Genuine statistical thinking requires the unnatural discipline of granting weight to facts that explain nothing — of letting causes and statistics each carry their proper share, rather than letting the appetite for a good causal story crowd the statistics out entirely.