Guide
How to read statistics in the news
Numbers make a story sound authoritative, but they can mislead even when they are accurate. Here are the questions that matter.
Compared to what?
"Cases doubled" sounds alarming; going from 2 to 4 cases is different from going from 2,000 to 4,000. Ask for the absolute numbers and the baseline. A "50% higher risk" of something that affects 2 in 100,000 people means the risk goes from 2 to 3 in 100,000.
Average of what?
The mean can be pulled up by a few very large values (average income), while the median (the middle value) is often a better picture of a typical person. Check which is used.
Polls and margin of error
A poll of about 1,000 people carries a margin of error of roughly three percentage points. If one candidate leads 48-46, that gap is within the noise. Also look at who was asked, how (phone, online), and how the question was worded.
Correlation is not causation
Ice cream sales and drownings both rise in summer; ice cream does not cause drowning. When a study says two things are "linked", ask whether something else (like hot weather) could explain both.
Sample size and who was studied
Small studies and studies on mice or on a narrow group may not apply to you. A single study is a starting point; many consistent studies, or a systematic review, are stronger.
Charts
Watch the axis: a chart that starts at 50 instead of 0 makes small changes look dramatic. Check units, time ranges and whether the data were adjusted for population or inflation.
Quick checklist
- What is the source of the number?
- How big is the sample, and who is in it?
- Absolute or relative change?
- What is left out?