A graph is not a picture of data. It is a set of choices about how to show data, and every choice is an argument.
Read the frame before the shape
Before you look at the line, read four things.
The title tells you what is claimed. The axis labels tell you what is measured. The units tell you the scale of what you are seeing. The source and date tell you whether to believe any of it.
Only then look at the shape. Readers who look at the shape first are being persuaded before they have been informed.
Where a chart can mislead you honestly
None of the following are lies. All of them change what you see.
A truncated axis. A bar chart whose vertical axis starts at 90 rather than 0 turns a rise from 92 to 96 into a bar that looks four times as tall. For bar charts, which encode value as length, this is genuinely misleading. For line charts showing change over time, a truncated axis is often reasonable — the convention differs by chart type, which is exactly why you must look.
A chosen time window. Any noisy series contains short stretches that rise and short stretches that fall. Show ten years and you see a trend; show eight months and you can see whatever you want.
Absolute counts versus rates. A city with more of something usually has more of everything, because it has more people. Per-thousand figures answer a different and usually better question than raw totals.
Aggregation. An average hides its distribution. "Average class size 24" is consistent with every class having 24 students and with half having 12 and half having 36.
Area instead of length. When a quantity is shown by scaling a circle or an icon, doubling both width and height quadruples the area. The eye reads area, so the difference looks far larger than it is.
Correlated lines
Two lines that rise together on the same chart do not establish that one causes the other. Both may respond to a third factor — population growth, seasonality, the economy. Charts that place two unrelated series on twin vertical axes, each scaled to fill the space, can make almost any pair look connected.
A checklist
- What is claimed by the title?
- What exactly is on each axis, in what units?
- Does the vertical axis start at zero, and does that matter for this chart type?
- What time period is shown, and what is left out?
- Counts or rates?
- Where did the data come from, and when?
- Who made this chart, and what would they like me to conclude?
A chart that survives all seven questions is a good chart. Learning to ask them takes about a week, and you never read a news graphic the same way again.