Misleading charts

A chart can show every number correctly and still create a false impression. That is what makes chart tricks so effective: nobody has lied, yet a distorted picture sticks in your head. The good news: most tricks follow a few patterns. Once you know them, you will spot them everywhere, in advertising, news and presentations.

The truncated axis

The classic: the vertical axis does not start at zero but just below the smallest value. The chart then shows only the tips of the bars, and tiny differences look enormous. If the axis starts at 90 and two bars reach 92 and 96, only 2 and 6 units stay visible: the second bar looks three times as tall, although its value is only about 4 percent larger. Truncated axes are not always meant to deceive, for temperature charts or share prices they often make sense. For bar charts meant to compare sizes, however, they are almost always misleading.

Inflated areas

Little pictures instead of bars are popular too: money bags, cars, figures. If a value doubles and the illustrator makes the symbol twice as tall and twice as wide, the area grows fourfold. Our eyes compare areas though, not heights. The impression: the value has quadrupled. With 3D charts and volumes it gets even more extreme, there the impression grows eightfold. Honest charts let symbols grow in one direction only, or line up equally sized symbols in a row.

Cherry-picking the time range

The section shown can deceive as well. Showing a share price only from its lowest point on presents an impressive rise and hides the crash before it. Choosing an unusually cold year as the starting point makes any warming look more dramatic, choosing an unusually warm one plays it down. This cherry-picking is especially insidious because every data point shown is correct. The deception lies entirely in what was left out. The counter question is always: what does the picture look like over a longer period?

How to expose the tricks

A short checklist helps: where does the axis start, at zero or somewhere in the middle? Are the axes evenly divided and labelled? Do symbols grow in more than one direction? Which time range is shown, and why this one of all ranges? And finally: does the headline match what the data actually shows? These five glances take less than a minute together and defuse most manipulations. Charts are a great tool for making data understandable. That is exactly why it pays to read them like a text.