Plain-language answer
A base rate is how often something happens overall, across a whole population or over time, before you look at any specific case’s details. When people judge how surprising a coincidence is, they tend to reason almost entirely from the vivid specifics in front of them and barely adjust for how common that kind of event actually is. Psychologists call this base-rate neglect, and it is one of the best-documented biases in judgment research.(Kahneman & Tversky, 1973)
Why it matters
“What are the odds that I would think of my friend right before they called me?” sounds like it needs an estimate of one rare event’s probability. But the more useful number is a base rate: across a lifetime of thousands of days, how often does any person think of any friend or relative shortly before that person contacts them, purely because friends and relatives contact each other often and cross each other’s minds often? Once that base rate is large, a match on any given day stops needing a special explanation — it is closer to what the base rate predicts than to a rare event.
Worked example: the representativeness shortcut
In a classic experiment, people were told a short personality sketch — supposedly drawn at random from a pool that was either mostly engineers or mostly lawyers — and asked to judge the profession described. Told the pool was 70% engineers and 30% lawyers, or the reverse, people’s judgments barely changed: they matched the sketch to whichever profession it seemed to “represent,” largely ignoring the stated base rate of engineers versus lawyers in the pool.(Kahneman & Tversky, 1973) The base rate was handed to them directly, in the same paragraph as the question, and it still made little difference to their answers.
The same paper documents a related effect with base rates over time: flight instructors who praised a trainee after an unusually good landing often saw the trainee’s next landing get worse, and instructors who criticized a trainee after an unusually bad landing often saw the next landing improve. This looked like proof that criticism works and praise backfires — but it is close to what the base rate of ordinary performance variation predicts on its own (a topic covered in full on regression to the mean), with no need to invoke praise or blame as a cause.(Kahneman & Tversky, 1973)
Common misconception
“The details of my story are so specific that base rates don’t apply” gets the logic backwards. The base rate answers a different, prior question — how often does this class of coincidence happen to someone — and that number is exactly what tells you whether the specific story needs a special explanation at all, or whether it is the kind of thing that was always going to happen to somebody, sooner or later.
Limits and open questions
Good base-rate estimates require real data about how often an event type occurs, and for many everyday coincidences (thinking of someone before they call, dreaming about an event before it happens) no one keeps a systematic count of the misses — the times you thought of someone and they didn’t call. Without that denominator, even a well-intentioned base-rate estimate can be little better than a guess; the discipline is in naming the gap rather than pretending precision that the data doesn’t support.
Related
- Reference classes covers the closely related problem of which comparison group a base rate should be measured over.
- Regression to the mean works through the flight-instructor example above in full.
Key takeaways
- A base rate is how often an event type occurs overall — the number most coincidence stories skip past on the way to the specific, vivid details.
- People systematically under-weight base rates even when told them directly, in favor of how well a case matches a stereotype or a story.
- Before treating a coincidence as needing a special explanation, ask how often that kind of event happens to someone, somewhere, given how many chances there are for it.