A diagnostic test with 95% accuracy for a condition affecting only 1% of the population returns a positive result for a specific individual. Intuitively, this feels like strong evidence the person has the condition. Properly accounting for how rare the condition actually is in the first place, a positive result from this same test is still more likely to be a false positive than a true positive — not because the 95% accuracy figure is wrong, but because the base rate fallacy describes exactly this tendency to ignore how rare the underlying condition is when interpreting what a positive test result actually means.
Working through why the math produces this counterintuitive result
Imagine 10,000 people, 1% of whom (100 people) actually have the condition. The test correctly identifies 95% of those 100 true cases, or about 95 true positives. Among the 9,900 people who don't have the condition, the test still incorrectly flags 5% as positive, producing about 495 false positives. Among the roughly 590 total positive results this test would produce across the full population, only 95 are true positives — meaning a randomly selected positive result is actually more likely to be a false positive than a true one, despite the test's genuinely high, correctly stated accuracy rate.
Why this is a general pattern, not specific to medical testing
The same underlying mathematics applies to any imperfect test or flag applied to a genuinely rare underlying condition — a fraud detection algorithm, a security screening system, a rare skill assessment — whenever the condition being tested for is rare enough, even a highly accurate test can produce a pool of positive flags where the majority are false positives, purely as a consequence of the base rate, independent of how accurate the underlying test genuinely is.
Why both the accuracy claim and the high false-positive share can be true at once
There's no contradiction between a test being genuinely, correctly described as 95% accurate and a majority of its positive results being false positives — both statements are simultaneously and correctly true, describing different things: the accuracy figure describes performance across the whole population, while the false-positive share among positive results specifically depends on how the rare true condition and the more common false alarms combine, a distinction the base rate fallacy specifically causes people to conflate.
What Bayesian reasoning actually adds to correct for this
Explicitly incorporating the base rate — how common the condition actually is before any test result is considered — alongside the test's known accuracy, using the mathematical framework of conditional probability, produces the correct estimate of how likely a positive result actually is to be a true positive, a genuinely different and generally more sobering number than the test's raw accuracy figure alone would naively suggest.
What this means for interpreting any test or flag applied to a rare condition
- Always ask how rare the underlying condition or event actually is, not just how accurate the test or flag is, before interpreting a positive result
- Recognize that a highly accurate test can still produce mostly false positives when applied to a sufficiently rare underlying condition
- Apply Bayesian reasoning explicitly, incorporating the base rate, whenever interpreting the practical meaning of a positive test result or algorithmic flag
- Be specifically cautious of decisions or interventions triggered by positive flags from a test applied to a rare condition, without first checking the actual expected false-positive share
The base rate fallacy is one of the more consequential and counterintuitive findings in probabilistic reasoning — a test's accuracy and the practical meaning of one of its positive results are related but genuinely different things, and conflating them produces a systematically overconfident interpretation of what a positive result actually tells you.