Online List Maker › Bar graph maker › Why bars start at zero
Why a bar graph has to start at zero
A bar shows its number by its length. Cut the bottom off the axis and every bar loses the same chunk, so small differences suddenly look huge.
Made-up example data.
What happened
In the second graph the axis starts at 90, so School A’s bar shows only the 2 points above 90 and School C’s only the 6. Your eye compares 2 with 6 and sees “three times as much”. The real comparison is 92 with 96: C is about 4% higher.
Edward Tufte called this ratio the lie factor: the size of the effect shown in the graphic divided by the size of the effect in the data. Here the bars show C as 200% bigger than A while the data says 4.3%, a lie factor of about 46. An honest graph scores 1.
How to spot it
- Look at the first number on the value axis. On a bar graph it should be 0.
- Watch for a zig-zag “break” mark at the bottom of the axis. It admits the cut, but the bars still mislead.
- Read the actual values. If two bars look wildly different but the numbers are close, the axis has been cut.
But line graphs don’t always start at zero
That’s right, and it isn’t a contradiction. A line graph shows each value by its position, not by a length from the baseline, so zooming the scale in on the data can be honest, as long as the axis is labelled. Body temperature is the usual example: a fever from 37 to 39 °C is serious, and on an axis starting at zero it would be a flat line. Bars are different: if the length doesn’t start at zero, it isn’t the value any more.
If the differences you care about are tiny compared with the values, a bar graph is the wrong picture. Plot the differences as bars (each school’s score minus the average, say), or use a line or dot chart.
The maker on this site never cuts a bar graph’s axis. The second graph above was drawn specially for this page.