Base rate neglect
base rate fallacy · ignoring the base rate
Judging how likely something is from how well the case matches a picture, while ignoring how common the thing is to begin with. The picture is vivid and the frequency is not, so the frequency gets dropped.
In practice
A candidate who looks exactly like a great hire is still most likely an average one, because most people are. The resemblance is evidence and the starting frequency is the larger part of the answer.
The common mistake
Treating it as innumeracy. People who can do the arithmetic make the same error when the case is described vividly, which is why the fix is a habit of asking for the denominator rather than better mathematics.
The base rate is how common something is before you know anything specific. Most businesses are small. Most projects run late. Most candidates are average, since that is what average means.
Base rate neglect is judging from resemblance instead. This case looks like the picture of a successful launch, so it is probably one — with the underlying frequency of successful launches quietly dropped.
The number that shows how large the error is
Take a screening test for a condition that one person in a thousand has. The test is 95 percent accurate in both directions.
Someone tests positive. How likely are they to have it?
Most people, including trained ones, say around 95 percent. The actual answer is about 2 percent.
Run a thousand people. One has the condition and almost certainly tests positive. Of the 999 who do not, about 5 percent test positive anyway — that is roughly 50 false positives. So about 51 positive results, of which one is real.
Nothing about the test was misrepresented. The 95 percent figure is accurate. It is simply swamped by the base rate, and the intuition that discards the base rate is off by a factor of fifty.
Why it happens
Resemblance is immediate and frequency requires a denominator nobody has to hand.
Tversky and Kahneman called the underlying move the representativeness heuristic: judging probability by how well a case fits the prototype. It is a reasonable shortcut when the prototype is rare and informative, and it fails exactly when the prototype is common or the target is rare — which is most business situations, because success is rare and the description of a success is common.
Vividness makes it worse. A detailed story is more convincing and less probable than a vague one, because every detail added narrows the set of cases it can describe. That is why a specific, plausible plan feels more likely to work than a general one.
Where it costs money
Hiring. Someone who resembles the best person you ever worked with is still drawn from a pool where excellence is rare. The resemblance is real evidence and it is smaller than it feels.
Forecasting a project. The inside view assembles a timeline from the specific steps of this project. The outside view asks how long comparable projects actually took. When the two disagree, the outside view is usually closer, and the inside view is the one that feels informed.
Judging a new market. "This could be huge" is a claim about an outcome that is rare by definition. The base rate for a new line of business reaching significant revenue is low, and it applies to yours.
Reading a warning sign. If the signal is common among healthy companies too, its presence says much less than it seems to. A test only discriminates to the extent that it is rare among the cases you want to rule out.
The habit that fixes it
Ask for the denominator, every time. Out of how many? Of everyone who looked like this, what share turned out that way? The question is almost always answerable approximately and almost never asked.
Start from the outside view and adjust. Begin with what usually happens in cases like this, then move from there based on what is genuinely unusual about yours. Starting from the specifics and never reaching the base rate is the default, and reversing the order changes the answer more than any amount of extra analysis.
Distrust the vivid case. Survivorship bias supplies the vivid cases and removes the denominator at the same time, which is why the two errors travel together and compound.
Concept web
Open the full webQuestions
What is the base rate fallacy?
Judging probability from how well a case matches a picture while ignoring how common the outcome is to begin with. Because the underlying frequency is often the larger part of the answer, dropping it can be wrong by an order of magnitude.
What is an example of base rate neglect?
A 95 percent accurate test for a condition affecting one person in a thousand. A positive result means roughly a 2 percent chance of having it, because false positives among the many healthy people vastly outnumber true positives.
What is the outside view?
Estimating from what happened in comparable cases rather than from the specifics of this one. When the outside and inside views disagree about a timeline or a chance of success, the outside view is usually closer.