The Base Rate Fallacy: Start With How Common It Is

Someone describes a stranger to you: quiet, tidy, fond of detail, a little shy, keeps their books in careful order.

Are they more likely to be a librarian or a salesperson?

Almost everyone says librarian. The description fits the picture of a librarian well and fits the picture of a salesperson poorly, so the answer feels obvious.

Now add the missing information. In most countries there are perhaps a few tens of thousands of librarians and several million people working in sales. Even if only one salesperson in twenty is quiet and bookish, that is still far more quiet bookish salespeople than there are librarians of any personality.

So the correct answer is salesperson — comfortably — and the description, while genuinely informative, was nowhere near strong enough to overcome the difference in numbers.

The number we left out is called the base rate: how common something is before you learn anything specific. Leaving it out is such a reliable mistake that it has its own name.

Why the mind skips it

This is not carelessness, and understanding the mechanism makes it easier to catch.

Specific detail feels like strong evidence. A vivid, well-matching description is engaging and concrete. A background statistic is abstract and forgettable. When both are available, the detail dominates attention.

We reason by resemblance. The mind's quick method for "which category is this?" is to compare the case to a mental picture of each category and pick the better match. That method is fast and often works. It has one blind spot: it takes no account of how many members each category has.

Base rates are usually invisible. The description arrives; the population statistics do not. Nobody tells you how many salespeople there are. You would have to go and ask, and the question does not naturally occur — because the resemblance already produced an answer that feels finished.

That last point is the practical heart of it. The error is not that people weigh the base rate poorly. It is that the base rate never enters the calculation at all.

Where it shows up

Medical tests. This is the most consequential example and the same arithmetic as Bayes' theorem: screening for a rare condition with an accurate test still produces mostly false positives, because the healthy population is so much larger. A positive result raises the probability substantially and often still leaves it well below fifty per cent.

Judging people from a single impression. An interview candidate who resembles your idea of a great engineer is evidence, but a weak one, because the base rate of people who seem impressive in interviews and turn out ordinary is high.

Alarms and fraud detection. Any system searching for something rare will generate mostly false alerts. This is not a sign the system is broken — it is arithmetic — and it explains alert fatigue, where people start ignoring warnings that are usually wrong.

News and risk. Coverage selects for the unusual. A vivid report about a rare event creates a strong impression that competes with, and usually beats, the statistical reality of how rare it is.

Startups and investing. "This company reminds me of an early version of a famous success" is a resemblance judgement. The base rate — most companies that resembled that one still failed — is invisible, which is survivorship bias and the base rate fallacy operating together.

Building the habit

The correction is one question, asked before you commit to a judgement:

"How common is this in general?"

That is it. It sounds almost too simple to be worth naming, and it fixes most instances, because the failure is one of omission rather than of reasoning.

A few refinements make it more effective:

Think in counts, not percentages. "One per cent false positive rate" is hard to feel. "About a hundred of these ten thousand people will be wrongly flagged, and only ten are actually ill" is immediately clear. Converting to whole people is the single most reliable way to make these problems intuitive.

Ask what else produces the same evidence. A description that fits your favoured explanation may fit several others just as well. If the evidence would look the same either way, it is not doing much work.

Treat the base rate as your starting point, then adjust. Strong evidence should move you a long way from it; weak evidence a short way. Neither should let you forget where you started.

Be especially careful when the thing is rare and the story is good. That is the exact combination where the fallacy does the most damage — a compelling narrative attached to something that almost never happens.

None of this means the description of the quiet stranger was worthless. It genuinely made librarian more likely than it was before. It simply started from so far behind that a good match was not enough to close the gap — and knowing that gap exists, and asking about it, is most of the skill.