The base rate

A striking result still sits inside a base rate. If a condition is rare, a test that is often right will still raise many false alarms, because there are so many healthy people to flag. A raw count can mislead in the same way: more accidents in a large age group does not mean that group is riskier per driver. Ask how common the thing was before this story, and whether you are looking at a count or a rate. The card is base-rate neglect.

Cognitive bias

Base-rate neglect

Base-rate neglect grabs a vivid likelihood and ignores how common the thing was to begin with. A test can be quite accurate and still be wrong most of the time it raises an alarm, if the condition is rare.

Open the full card

Mini check

A condition affects 1 in 100 people. A test catches 90% of real cases, but also wrongly flags 9% of healthy people. You test positive. Roughly how likely is it that you have the condition?

About 1 in 11, roughly 9%.About 90%, because the test catches 90% of cases.About 91%, because only 9% of results are false.Exactly 50%, because you either have it or you do not.

More road accidents involve drivers aged 30–50 than drivers over 80. Someone concludes that drivers over 80 are safer. What is missing?

The rate: accidents per driver or per mile driven, not the raw count.Nothing, because a smaller count always means lower risk.A count of accidents involving cyclists instead.The age of the cars rather than the drivers.

Try it today

Before you trust a dramatic percentage, write down how common the event was to start with.

Next step