Percentages without a base
'Risk doubles' sounds alarming, but doubling a one-in-a-million chance gives two in a million. Always ask for the actual numbers behind a percentage. A change in relative terms can look huge while the absolute change is tiny.
Average can mean three things
The mean, the median and the mode are all called 'average'. A few very high earners pull the mean income up while the median, the middle value, stays low. When a story gives an average, ask which one it is and whether a few extreme values could be distorting it.
Correlation is not cause
Two things rising together does not mean one causes the other. Ice cream sales and sunburn both go up in summer, because of the weather. Be suspicious when a headline says one thing 'leads to' another on the basis of a pattern alone.
Polls and sample size
A poll samples a small group to estimate what everyone thinks. The margin of error shows how much the result could move, so a lead of two points inside a margin of three is not a real lead. Check how many people were asked and how they were chosen.
Charts can be stretched
A chart whose vertical axis starts at 90 instead of zero makes a small change look dramatic. Check the axis labels and the time period shown. Choosing a start and end date can make almost any trend look upward or downward.
Questions to keep handy
Compared with what? Out of how many? Over what period? Who collected the data, and why? Four short questions will catch most misleading statistics.