Misleading charts

Intermediate

Introduction

Charts are persuasive because they look objective. That is exactly why a small design choice can mislead: a bar chart whose axis starts at 45 makes a 10% rise look like a doubling. Some misleading charts are made on purpose, many by accident. Knowing the common tricks protects you as a reader and keeps your own charts honest.

How it works

Truncated axes. In a bar chart, the reader compares the lengths of bars, so the axis must start at zero. With values 50 and 55 and an axis starting at 45, the bars are 5 and 10 units long, and the 10% increase looks like 100%. Line charts are different: they encode values by position, not length, so a zoomed axis is acceptable as long as it is clearly labelled.

Areas and bubbles. When a value is shown by the area of a circle or icon, the area must be proportional to the value, so the radius grows with the square root: a value four times larger gets twice the radius. Scaling the radius linearly makes it look 16 times larger. 3D effects and exploded pies distort areas further by adding perspective.

Dual axes. Two y axes with independent scales can make any two series appear to cross, diverge or move together, simply by choosing the ranges. Prefer two aligned charts, or index both series to 100 at the same start.

Cherry-picking. Choosing the start and end of a time range can turn a long decline into a rise. Show the full relevant period, or explain why a shorter one was chosen.

The lie factor. Edward Tufte defined the lie factor as the size of an effect shown in the graphic divided by the size of the effect in the data, where an effect is the relative change, |second − first| ÷ first. An honest chart has a lie factor close to 1 (Tufte suggested between 0.95 and 1.05). His famous example, a newspaper chart of car fuel standards, showed a 53% increase in the data with lines that grew by 783%: a lie factor of about 14.8.

Example

# How a truncated axis exaggerates a difference, measured as a lie factor.
before, after = 50, 55
for baseline in (0, 45):
    drawn_before = before - baseline
    drawn_after = after - baseline
    data_change = (after - before) / before
    drawn_change = (drawn_after - drawn_before) / drawn_before
    print(f"axis from {baseline}: bars {drawn_before} and {drawn_after}, "
          f"data +{data_change:.0%}, drawn +{drawn_change:.0%}, lie factor {drawn_change / data_change:.1f}")
Open in compiler

Line by line

  1. The data rise 10%, from 50 to 55.
  2. With the axis at 0, the bars are 50 and 55 long, and the drawn change equals the data change.
  3. With the axis at 45, the bars are 5 and 10 long: the drawn change is 100%.
  4. The lie factor divides the drawn change by the data change.

Expected result

axis from 0: bars 50 and 55, data +10%, drawn +10%, lie factor 1.0
axis from 45: bars 5 and 10, data +10%, drawn +100%, lie factor 10.0

In the real world

Political adverts and company reports are regularly criticised for truncated bar charts and selective time ranges. Newsrooms have style rules that bar charts start at zero and that bubbles are scaled by area, and fact-checkers redraw misleading charts with honest axes to show what the data really say.

Common mistakes

  • Starting a bar chart's axis above zero.
  • Scaling circle radius by the value, which squares the visual difference.
  • Using 3D bars or pies, whose perspective distorts sizes.
  • Comparing two series on dual axes whose ranges were chosen to make a point.

Best practices

  • Start bar axes at zero; zoom line charts only with clear labels.
  • Scale areas, not radii, and avoid 3D.
  • Show the full context and calculate the lie factor when in doubt.

Summary

Charts mislead through truncated bar axes, radius-scaled bubbles, 3D, dual axes and cherry-picked ranges. Bars start at zero, areas scale with the value, and Tufte's lie factor, the effect shown divided by the effect in the data, should stay close to 1.

Quick check

1. Bars for 50 and 55 are drawn from an axis starting at 45. The rise looks like…
2. A value is 4 times another. How much larger should its bubble's radius be?
3. A chart shows a 10% rise as a bar twice as long. Its lie factor is about…
4. What is the main risk of a dual axis chart?

Key takeaways

  • Bar charts must start at zero, because length carries the value.
  • Scale bubble and icon areas with the value: radius ∝ √value.
  • A lie factor far from 1 means the chart distorts the data.

Practice problems for this lesson

My notes

Write down what you want to remember from this lesson. Notes are saved on this device as you type, and appear on your dashboard.