“9 out of 10”: numbers that mislead

Three questions for every number

1. Out of how many? “9 out of 10” can mean nine people out of ten, but also 900 out of a thousand. That's a big difference. It's the same with percentages: “a 200% increase” can also mean that something happened once last year and three times this year.

2. Who measured it – and who paid for it? Did they ask randomly chosen people, or only happy customers? Was the survey done by the company that sells the product?

3. Compared with what? “The best result ever” – and what was it like at the same time last year? And elsewhere?

Charts that exaggerate

A chart looks like precise proof. But if the vertical axis doesn't start at zero, a small difference looks huge. Say one bar shows 92 and another 95. If the axis starts at 90, the second bar is two and a half times taller in the picture – even though the difference is only 3 points.

The UK's Office for National Statistics spells it out in its rules for charts: a bar chart's axis should always start at zero. When you see a chart, look at the numbers on the axis first.

Together isn't because

“Pupils who eat breakfast get better grades. So breakfast improves grades!” Maybe. But maybe the pupils who eat breakfast also sleep more or have more peace and quiet at home for studying. Two things going together doesn't mean that one causes the other.

A classic example: when it's hot, more ice cream is sold and more people get sunburnt. Ice cream doesn't cause sunburn – summer is behind both.

Find out more in the Manipulation atlas: Percentages without a base, Misleading chart and Correlation ≠ causation.