How to Lie with Statistics — and How to Spot It

Overview

Statistical deception occurs when numbers, graphs, or summary statistics are presented in a way that leads to a false or misleading conclusion. The deception can be intentional (disinformation, propaganda) or accidental (poor method, bias). This lesson focuses on the techniques and the critical questions you should ask when someone presents data.

Common Tricks (and how to spot them)

• Truncated or manipulated axes
  Bar/line charts that start at a value other than zero to exaggerate changes.
• Misleading averages
  Using mean when median is the better measure, or vice versa.
• Cherry-picking
  Selecting time windows, subgroups, or datasets that support the claim.
• Small or unrepresentative samples
  High margins of error or biased sampling make results unreliable.
• Correlation ≠ causation
  Two things moving together doesn't mean one causes the other.
• Misleading percentages
  Using percent-changes on tiny bases or omitting absolute numbers.
• Base-rate neglect
  Ignoring the overall prevalence when interpreting probabilities.
• Selective comparisons
  Comparing to the wrong baseline or to an outlier for dramatic effect.

Worked examples

Quick spotting checklist

1. Who produced the data and why? (sponsor bias)
2. What is the sample size and how was it chosen?
3. Which statistic is used (mean, median, mode) and is it appropriate?
4. Are time windows or subgroups cherry-picked?
5. Do graphs start at zero? Are axes labeled clearly?
6. Are percentages shown with absolute counts or bases?
7. Is causation claimed where only correlation exists?
8. Are confidence intervals, error bars, or margins of error reported?

Practice exercises

Try these on your students or readers:

1. Find a newspaper chart and re-plot it with a zero-based axis. Does the impression change?
2. Given the sample [2,3,3,4,50], compute mean and median and explain which is more informative.
3. Read a claim that says "X reduces risk by 50%" — ask for the absolute risk reduction and baseline risk.

Further reading & resources

• How to Lie with Statistics — Darrell Huff (classic, short)
• The Signal and the Noise — Nate Silver (methods, forecasting)
• American Statistical Association — guidance on misuse of statistics
• Online: FiveThirtyEight, Gapminder, Our World in Data (good practice examples)

Use these as starting points for lesson expansion, source critique assignments, or class activities.